|
|
|
Student Projects:
What makes for a cradle-to-cradle building?Project Code:AMM_01Supervisor:Dr Haley JonesOutline:The cradle-to-cradle paradigm considers the whole-of-life design of products and processes with the four basic tenets:
Such an approach lets us tackle not only climate change but environmental sustainability as a whole. It is one way we could be approaching our interactions with the world that just could help us to, not only survive, but thrive along with the rest of the planet, rather than at the expense of it (and, eventually, ourselves). Goals of this projectThis project will allow the student to become familiar with the paradigm and to explore the potential for its application.Other similar project ideas are welcomed. Contact:E: haley.jones@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Haley.Jones/ What makes a cradle-to-cradle city?Project Code:AMM_02Supervisor:Dr Haley JonesOutline:The cradle-to-cradle paradigm considers the whole-of-life design of products and processes with the four basic tenets:
Such an approach lets us tackle not only climate change but environmental sustainability as a whole. It is one way we could be approaching our interactions with the world that just could help us to, not only survive, but thrive along with the rest of the planet, rather than at the expense of it (and, eventually, ourselves). Goals of this projectThis project will allow the student to become familiar with the paradigm and to explore the potential for its application.Other similar project ideas are welcomed. Contact:E: haley.jones@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Haley.Jones/ Towards cradle-to-cradle for the XXX industryProject Code:AMM_03Supervisor:Dr Haley JonesOutline:The cradle-to-cradle paradigm considers the whole-of-life design of products and processes with the four basic tenets:
Such an approach lets us tackle not only climate change but environmental sustainability as a whole. It is one way we could be approaching our interactions with the world that just could help us to, not only survive, but thrive along with the rest of the planet, rather than at the expense of it (and, eventually, ourselves). Goals of this projectThis project will allow the student to become familiar with the paradigm and to explore the potential for its application.Other similar project ideas are welcomed. Contact:E: haley.jones@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Haley.Jones/ Computation of heat conductivity of metallic glassy alloysProject Code:AMM_04Supervisor:Dr Zbigniew StachurskiOutline:A significant research effort is continuously being applied to understand the structure and behaviour of amorphous materials. Thermal, optical, magnetic and electronic properties of amorphous materials hold great promise towards current and emergent technologies. A computer model of an ideal amorphous solid (IAS) of a Zr-based metallic glass has been defined and simulated. Next stage is to compute the physical properties of such a body.In this project will use the approximate theoretical approach for calculating heat conductance. Following the Green-Kubo formalism in linear response theory, the thermal conductivity of the IAS model will be determined by using classical molecular dynamics simulation to calculate the heat current correlation function. Goals of this project1.Learn to do molecular dynamics on the IAS model using the existing software.2.Compute thermal conductivity or other physical properties of the IAS model of a Zr-based metallic glass. Contact:E: zbigniew.stachurski@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Zbigniew.Stachurski Magnetic and dielectric properties of metallic glassy alloysProject Code:AMM_05Supervisor:Dr Zbigniew StachurskiOutline:A significant research effort is continuously being applied to understand the structure and behaviour of amorphous materials. Thermal, optical, magnetic and electronic properties of amorphous materials hold great promise towards current and emergent technologies. A computer model of an ideal amorphous solid (IAS) of a Zr-based metallic glass has been defined and simulated. Next stage is to compute the physical properties of such a body.In this project we will use the microscopic theory of magnetism and dielectrics. Goals of this project1.Learn to do molecular dynamics on the IAS model using the existing software.2.Compute thermal conductivity or other physical properties of the IAS model of a Zr-based metallic glass. Contact:E: zbigniew.stachurski@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Zbigniew.Stachurski
Software Development for a Mobile Communications SystemProject Code:AS&T_01Supervisor:Dr Mark ReedOutline:This project involves working with a team on the continuing software development for a communications system. The software will be part of the software stack used for modem operation as well additional software for testing the modem. The software will be a combination of C++ code running on embedded linux as well as C/C++ code running on a PC. There may be a need for GUI development and Java tools are expected to be also used.Goals of this projectThe goals of this project are to develop a software test environment and platform to show a compelling case for the wireless modem functionality. The software is an important part of this platform.Requirements/PrerequisitesThe student is expect to have good software skills and problem solving skills. The ability to program in C/C++ and Java are essential. Communication system understanding is helpful but not essential.Student GainThe student will get exposure to a commercially oriented project while working with a team of highly motivated engineers. The project will involve working with hardware platforms for testing and face challenging problems including real-time issues and writing efficient software for embedded systems.Contact:E: mark.reed@nicta.com.auW: http://users.rsise.anu.edu.au/~mreed/ Scattered sensor systems for bushfire monitoringProject Code:AS&T_02Supervisor:Dr Kim BlackmoreOutline:An aerially scattered sensor system for monitoring bushfire conditions is proposed. The objective in this project is to determine the likely locations of sensors under various scattering conditions relevant to bushfire situations. The effect of helicopter flight behaviour, updrafts, extreme wind and heat will be investigated.Requirements/PrerequisitesThe student should be familiar with MATLAB programming and mathematically adept.Contact:E: Kim.Blackmore@anu.edu.auW: http://engnet.anu.edu.au/DEpeople/Kim.Blackmore/ Wireless Modem RealisationProject Code:AS&T_03Supervisor:Dr Mark ReedOutline:This project entails building a wireless modem for a wireless communication system. The Modem (transmitter and receiver) will be developed in MATLAB and will follow a known wireless standard. The project will consist of working with a team and becoming intimately familiar with modem realisations for real-world systems.The depth of this project will only be limited by the students ability. The work will start by developing a known transmitter design and validating and testing this. The design will be to the teams coding and realisation standards. The work will then move to the receiver and detection and decoding of the signal. Finally real-world wireless channels will be used to verify performance. Goals of this projectThe goal of the project is to build a modem that can be tested and validated for potential use within a much bigger project.Requirements/PrerequisitesThe student is expected to have used MATLAB and have some understanding and desire to work on communication systems.Student GainThe student will work with a number of researchers to perform the task and get a much better idea what is involved in real communication systems design.Contact:E: mark.reed@nicta.com.auW: http://users.rsise.anu.edu.au/~mreed/ Graphical User Interface DevelopmentProject Code:AS&T_04Supervisor:Dr Leif HanlenOutline:In conjunction with the Australian Institute of Sport, NICTA has developed a machine learning algorithm that automatically detects and classifies swimming activities based upon sensor measurements. The sensor being used is a tri-axial accelerometer worn externally proximate to the lumbar spine. The results of this classification are entered into a database that acts as a log of all the swimmer's activity, and is available to coaches and sports scientists via a custom graphical user interface (GUI).Goals of this projectAt present, the GUI requires some training to be used effectively. The goal of this project is to redesign this GUI using current best design practices to make it intuitive, "user-friendly" and suitable for non-technical users.Requirements/PrerequisitesThis project will require researching the current state of the art in GUI design and implementation of these ideas onto an existing application.Ideally, the student should be familiar with C#/.Net programming. However, this is not a requirement; Familiarity with C++ (or similar) would be beneficial. If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Contact:E: leif.hanlen@nicta.com.auInertial Measurement Units (IMU's) using only AccelerometersProject Code:AS&T_05Supervisor:Dr Leif HanlenOutline:Inertial measurement units (IMU's) use "dead-reckoning" to estimate the location and movement of an object. Typically IMU's use a single 3D accelerometer (measures acceleration in x,y,z directions) a gyroscope (measures angular velocity) and other sensors. Typically, IMU's have one of each type of sensor: the iPoD may be considered a simple IMU since it has a single 3D-accelerometer.Estimating position and direction using these IMU's is difficult, and noisy since the sensors are co-located. There has been some recent work on using multiple accelerometers. The accelerometers can be spatially separated by a small distance in x,y,z directions, and using 6 or more simple accelerometers, most other sensors may be replaced. There are a number of questions to be addressed:
Requirements/PrerequisitesThe project will require understanding of Matlab, for simulation focused results. Some simple control- or comms- theory will help but is not essential. Hardware engineers are available for discussion in regard to the capabilities of the sensing devices.web ref: http://www.mycoordinates.org/april09/cali.php">multiple The student should be familiar with MATLAB programming. Hardware development is not necessary If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Contact:E: leif.hanlen@nicta.com.auCloaking DevicesProject Code:AS&T_06Supervisor:Dr Leif HanlenOutline:Is it really Science-Fiction? Active noise headphones are available off-the-shelf (at a price). These devices essentially work by detecting a wavefield (a sound wave) and applying a negative wave so that when the original sound and the negative wave enter the ear canal, the user does not detect the original sound. Can the principle be applied to a remote region - so that the "active headphone" cancels out sounds "over there" rather than only inside the ear?Theory says yes - since we can reconstruct a soundfield remotely using beam-forming and we can generate sounds remotely using directional speakers (also beam-forming). The problems are: (1) How do we get the sound cancelled where we want it without being horribly loud somewhere else and (2) How do we cope with noise? Requirements/PrerequisitesThe project will require understanding of Matlab, for simulation focussed results. We also have a number of Texas-Instruments signal processing boards and a hardware design lab if students wish to develop a hardware demonstrator.The student should be familiar with MATLAB programming. Hardware development is not necessary. If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Contact:E: leif.hanlen@nicta.com.auIndependence of physical measurements for swimmingProject Code:AS&T_07Supervisor:Dr Leif HanlenOutline:NICTA currently has over 100hours of swimming data obtained from inertial measurement devices. The inertial devices include acceleration and gyroscope as well as a number of other type of sensor. This project will evaluate the independence of the inertial data over different time scales. we will examine common measures of independence such as correlation, and look at less common measures such as mutual information and differential entropy.Goals of this projectAdvanced students may be able to use the material to develop an activity prediction and/or recognition algorithm based on the sensor data.Students will have access to data gathered from Olympic swimmers as well as a wider interaction with members of the Human Performance Improvement project team. Details of the HPI project are given on the web.http: // nicta. com. au/ research/ projects/ human performance improvement Requirements/PrerequisitesA NICTA assignment of IP and a NICTA confidentiality agreement will be required for this projectIf you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Background Literature
Related NICTA project:Human Performance Improvement Contact:E: leif.hanlen@nicta.com.auCorrelation of Perceived Performance, and Physical MeasurementsProject Code:AS&T_08Supervisor:Dr Leif HanlenOutline:There is a loose relationship between the physical performance of an Olympic athlete and their perceived performance [1, 2, 3]. A number of rating systems exist, in particular, the Relative Perceived Effort (RPE) score. The score rates (on a scale of 0-10) the effort involved in a given task: 0 denotes no effort at all, while 10 denotes near collapse.Currently, at the Australian Institute of Sport, Olympic basket-ballers measure the effort involved in training via the RPE score. Similar schemes are used around the world [4]. The basket-ballers complete a training exercise and fill in a form of their perceived effort. Is there a relationship between RPE and physical effort? NICTA and the AIS have recorded heart-rate levels and accelerometer readings from a number of activities, and the athletes have put in their perceived effort [5]. The objective of this project is to evaluate the the conjecture that some as yet unknown combination of accelerometer data and heart-rate data might be used to predict the RPE score of an athlete. Goals of this projectThe objective of this project is to determine if there is a relationship between the two types of data: a subjective performance measure and physical (acceleration and heart-rate) measurement.Students will have access to data gathered from Olympic basketballers as well as a wider interaction with members of the Human Performance Improvement project team. Details of the HPI project are given on the web. http: // nicta. com. au/ research/ projects/ human performance improvement Requirements/PrerequisitesThis project is ideally suited to students who have an interest in mathematics and statistics. Some Matlab experience will be helpful. A student who has completed 2nd-year mathematics will have sufficient skill to complete the tasks.If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Background Literature
Related NICTA project:Human Performance Improvement Contact:E: leif.hanlen@nicta.com.auLognormal-Rayleigh combination model for human shadowing of MIMO radio signalsProject Code:AS&T_09Supervisor:Dr Leif HanlenOutline:There has been some conjecture in the communication theory research world as to whether or not radio signals around the human body can be statistically characterised by combining two well known statistical distributions:
Can the two models be combined to give shadowed-non-line-of-sight radio model for people? Requirements/PrerequisitesA NICTA assignment of IP and a NICTA confidentiality agreement will be required for this projectStudents will have access to data gathered from several wireless network measurement campaigns as well as a wider interaction with members of the Human Performance Improvement project team. Details of the HPI project are given on the web. http: // nicta. com. au/ research/ projects/ human performance improvement This project will require some basic wireless communications knowledge and statistics. Matlab programming knowledge will be a significant help. If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Background Literature
Related NICTA project:Human Performance Improvement Contact:E: leif.hanlen@nicta.com.au3D Space-Time model for wireless body area networksProject Code:AS&T_10Supervisor:Dr Leif HanlenOutline:Body Area Networks (IEEE 802.15.6) represent the next generation wireless personal area networks. They will provide networking for small smart sensors, with up to 10 piconets per person, and 256 nodes per net (ie, up to 2000 nodes per person).The technical requirements of the new standard are given online [1]. Details of the channel modelling campaigns may be found in: [2, 3, 4] and the complete wireless channel model given by [5]. So far, there are no statistical models for the propagation of a wireless signal around the human body from any point to any other point. Goals of this projectThis project will develop a model, in 3D space and over time to give a spatial Jakes-model [6, 7, 8] for point-to-point communication around the human body.Requirements/PrerequisitesThis project will require some basic wireless communications knowledge and statistics. Matlab programming knowledge will be a significant help.Students will have access to data gathered from several wireless network measurement campaigns as well as a wider interaction with members of the Human Performance Improvement project team. Details of the HPI project are given on the web. http: // nicta. com. au/ research/ projects/ human performance improvement A NICTA assignment of IP and a NICTA confidentiality agreement will be required for this project If you are the successful applicant for this project, you will be located off campus at NICTA's Canberra Research Lab. Background Literature
Related NICTA project:Human Performance Improvement Contact:E: leif.hanlen@nicta.com.auAutomatic Fuseball playing systemProject Code:AS&T_11Supervisor:Dr Leif HanlenOutline:Fuseball is the table-top equivalent of soccer (football). This project will develop a computer player to challenge human fuseball players. The project will appeal to students who are interested in robotics and embedded system design. There are a large number of independent components to this project which may be undertaken by separate individuals. Topics of note are the visual system to track the status of the table in real time, a control system to drive the computer player's team and a strategic system to make the computer unbeatable! It is expected the project can be developed to a demonstrable state in 3 months by a team of 6-10 dedicated students.This project will be an excellent introduction to the complex world of real-time embedded system design and should also be great fun! This is a project for approximately 6-10 summer scholar students. In an extension we would like to incorporate the sensor glove into the game, so that the computer player may "learn" good strategies. This will require students who understand signal processing - to extract the relevant information form the glove - and students who understand online machine learning (how to use the data to make a better player). Realistically, the sky is the limit for what you can do with this project. Requirements/PrerequisitesDepending upon the particular aspect of the project students will need to be familiar with one or more:
DSP programming, Graphics processing unit programming and FPGA design may be helpful, but are not essential. Contact:E: leif.hanlen@nicta.com.auWireless Modem RealisationProject Code:AS&T_12Supervisor:Dr Mark ReedOutline:This project entails building a wireless modem for a wireless communication system. The Modem (transmitter and receiver) will be developed in MATLAB and will follow a known wireless standard. The project will consist of working with a team and becoming intimately familiar with modem realisations for real-world systems.The depth of this project will only be limited by the students ability. The work will start by developing a known transmitter design and validating and testing this. The design will be to the teams coding and realisation standards. The work will then move to the receiver and detection and decoding of the signal. Finally real-world wireless channels will be used to verify performance. Goals of this projectThe goal of the project is to build a modem that can be tested and validated for potential use within a much bigger project.Requirements/PrerequisitesThe student is expected to have used MATLAB and have some understanding and desire to work on communication systems.Student GainThe student will work with a number of researchers to perform the task and get a much better idea what is involved in real communication systems design.Contact:E: mark.reed@nicta.com.auW: http://users.rsise.anu.edu.au/~mreed/ GPU Implementation of Important Decoding Algorithms in Communication SystemsProject Code:AS&T_13Supervisor:Dr Parastoo SadeghiOutline:Reliable transmission of infomation over communication channels often requires sophisticated algorithms to be implemented at the transmitter and receiver ends. Fast simulation of such algorithms during design stages of a communication system is vital for a successful practical implementation. The new generation of powerful and general-purpose graphic processing units (GPU) provides new opportunities for very efficient communication system evaluation and analysis.For example, to combat channel noise, powerful channel error connection codes are used at the transmitter. The receiver needs to accordingly employ appropriate decoders to choose the 'most likely' transmitted signal. It is important to understand how a coding/decoding technique performs in a given channel condition. Two other main tasks of a receiver (especially in wireline ADSL and wireless fading channels) are to combat inter-symbol interference (ISI) and also possible time variations in the channel. Goals of this projectIn this project, the student will gain a good understanding of important and widely used coding and decoding techniques, such as convolutional codes and Viterbi decoders by first implementing and testing them on the CPU. Meanwhile he/she will aquire programming skills on NVIDIA CUDA platform. The next step is to devise the best strategy to make the best use of parallel processing capabilities of GPU and to transform the code onto the GPU. The code will then be tested and compared with standard CPU implementation. If the time permits, other major decoding algorithms such as Turbo decoders and even Turbo equalizers will also be investigated.The project will provide an excellent opportunity for the student to work in a research team with researchers in the school of Engineering at the ANU. It is expected that the outcome of the research be submitted as a technical paper to a national or international conference for publication. Requirements/PrerequisitesA good understanding of C programming language is required. Familiarity with digital communications and Metlab programming will be a benefit.Background Literature
Contact:E: parastoo.sadeghi@rsise.anu.edu.auW: http://users.rsise.anu.edu.au/~parastoo/ Optimizing Channel Estimation for Higher Data Rate in Wireless Communication SystemsProject Code:AS&T_14Supervisor:Dr Parastoo SadeghiDr Tharaka Lamahewa Outline:In wireless communications, the signal transmitted through the wireless environment (i.e. the channel) experiences distortion in its amplitude and phase. The receiver needs to have an accurate estimation of what has been done by the channel in order to decode the signal. Conventional wireless networks usually spend a considerable amount of transmit resource on channel estimation. For example, GSM system sacrifices about 22% of the data transmission time to perform channel estimation. However, this results in a direct reduction in the actual data rate. Therefore, optimizing the amount of resource spent on channel estimation is crucial for high data rate transmission, which has been an active area of research in recent years.In this project, the student will gain a good understanding on transmission design and performance analysis for future wireless communication systems. After an introduction to the topic and relevant background studies, the student will choose either of the following options:
Requirements/PrerequisitesFamiliarity with wireless communications and Matlab will be a benefit.Student GainThe project will provide an opportunity for the student to work in a research team with researchers in the School of Engineering at the ANU. It is expected that the outcome of the research be submitted as a technical paper to a national or international conference for publication.Background LiteraturePilot symbol transmission for time-varying fading channels: an information-theoretic optimization," Proc. Int. Conf. on Signal Processing and Commun. Syst., Dec. 2007. W:http://users.rsise.anu.edu.au/~parastoo/papers/2007/ICSPCS_2007.pdfContact:E: parastoo.sadeghi@rsise.anu.edu.auW: http://users.rsise.anu.edu.au/~parastoo/ E: tharaka.lamahewa@anu.edu.au W: http://users.rsise.anu.edu.au/~tharaka/ Performance Evaluation of Wireless Communication Systems in a Temporally Correlated Channel EnvironmentProject Code:AS&T_15Supervisor:Dr Parastoo SadeghiDr Tharaka Lamahewa Outline:Due to the nature of the wireless channel, the design of the wireless systems fundamentally differs from wired system designs. The wireless channel is much more unpredictable than the wired channel because of factors such as multipath, mobility of the user, mobility of the objects in the environment and delays arising from multipaths. In this project we particularly aim to investigate/evaluate the performance measures such as packet error rate (PER) and packet error density (PED) of a wireless communication system when the receiver is moving (mobile) and the environment surrounding the receiver is non-uniform.In this project, the student will be exposed to following research areas: wireless channel modelling, finite state Markov channel (FSMC) modelling and, PER and PED calculations. Requirements/PrerequisitesFamiliarity with wireless communications and Matlab will be a benefit.Student GainThe project will provide an opportunity for the student to work in a research team with researchers in the School of Engineering at the ANU. It is expected that the outcome of the research be submitted as a technical paper to a national or international conference for publication.Background Literature1. "Finite-State Markov Modeling of Fading Channels", IEEE Signal Processing Mag., 2008, W: http://users.rsise.anu.edu.au/~parastoo/papers/2008/SPM_2008.pdfContact:E: parastoo.sadeghi@rsise.anu.edu.auW: http://users.rsise.anu.edu.au/~parastoo/ E: tharaka.lamahewa@anu.edu.au W: http://users.rsise.anu.edu.au/~tharaka/ Wireless tweetingProject Code:AS&T_16Supervisor:Dr Leif HanlenOutline:We have developed a number of wireless signal transmitters and have a number of iPods (with developer access). We are looking to develop a twitter app which wirelessly tweets details of the subject's activities through the day.Simple examples are: based on a number of wireless locators, tweet the where-abouts of the subject in the office every N (eg. 30) minutes. Include the accelerometer app to tweet the activity level of the subject Requirements/PrerequisitesThis project will require a solid ability to program C++ and app's in iPod as well as develop custom PHP scripts to access the tweeting facility. You will have a support with firmware engineers, but will need to be self motivated for the coding side.Contact:E: leif.hanlen@nicta.com.au
Universal Artificial IntelligenceProject Code:AI_01Supervisor:Dr Marcus HutterOutline:The dream of creating artificial devices that reach or outperform human intelligence is an old one. Most AI research is bottom-up, extending existing ideas and algorithms beyond their limited domain of applicability. The information-theoretic top-down approach pursued in [Hut05] justifies, formalizes, investigates, and approximates the core of intelligence: the ability to succeed in a wide range of environments [LH07]. All other properties are emergent.Goals of this projectThe fundamentals of UAI are already laid out, but there are literally hundreds of open questions (see the exercises in [Hut05]) in this approach that have not yet been answered. The complexity ranges from suitable-for-short-projects to full PhD theses and beyond.Requirements/Prerequisites
Student Gain
Background Literature
Contact:E: marcus.hutter@anu.edu.auW: http://www.hutter1.de/rsise/index.htm Human Knowledge Compression ContestProject Code:AI_02Supervisor:Dr Marcus HutterOutline:Being able to compress well is closely related to intelligence as explained below. While intelligence is a slippery concept, file sizes are hard numbers. Wikipedia is an extensive snapshot of Human Knowledge. If you can compress the first 100MB of Wikipedia better than your predecessors, your (de)compressor likely has to be smart(er). The intention of the Human Knowledge Compression Prize [Hut06] is to encourage development of intelligent compressors/programs.Goals of this projectSome of the following four subgoals shall be addressed:
Requirements/Prerequisites
Student Gain
Background Literature
Contact:E: marcus.hutter@anu.edu.auW: http://www.hutter1.de/rsise/index.htm On the Foundations of Inductive ReasoningProject Code:AI_03Supervisor:Dr Marcus HutterOutline:Humans and many other intelligent systems (have to) learn from experience, build models of the environment from the acquired knowledge, and use these models for prediction. In philosophy this is called inductive inference, in statistics it is called estimation and prediction, and in computer science it is addressed by machine learning. The problem of how we (should) do inductive inference is of utmost importance in science and beyond. There are many apparently open problems regarding induction, the confirmation problem (Black raven paradox), the zero p(oste)rior problem, reparametrization invariance, the old-evidence and updating problems, to mention just a few. Solomonoff's theory of universal induction based on Occam's and Epicurus' principles, Bayesian probability theory, and Turing's universal machine [Hut05], presents a theoretical solution [Hut07].Goals of this project
Requirements/Prerequisites
Student Gain
Background Literature
Contact:E: marcus.hutter@anu.edu.auW: http://www.hutter1.de/rsise/index.htm Universal Induction versus No Free LunchProject Code:AI_04Supervisor:Dr Marcus HutterOutline:Solomonoff's theory based on Occam's razor provides a rigorous and formal method for inductive inference, prediction, and time series forecasting, by essentially uniquely specifying a universal model class and prior. This model seems to solve the long outstanding (philosophical) problem of induction and many other deep statistical questions [Hut07]. On the other hand, the No-Free-Lunch theorem(s) [WM97] state that all optimization or search algorithms are on average equally good/bad, if a uniform average over the space of all functions is taken. So NFL believers conclude that we need a prior, biased to our particular problem class at hand, that is, there is no universal (problem independent) solution to the induction problem. There is an ongoing battle between believers in Occam's razor and believers in no-free-lunches [Sto01,SH02].Goals of this project
Requirements/Prerequisites
Student Gain
Background Literature
Contact:E: marcus.hutter@anu.edu.auW: http://www.hutter1.de/rsise/index.htm Projects in Automated PlanningProject Code:AI_05Supervisor:Dr Patrik HaslumOutline:Planning is a branch of AI concerned with automating the reasoning that goes on in formulating plans, in the sense of a series of steps to be taken, each of which has some effect on the state of (a highly abstract model of) the world, so as to bring about some desired end state. As a prototypical example, one may take single-player games ("puzzles") such as the Freecell card game or the old Rubik's Cube. Possible applications of automated planning methods (aside from solving puzzles) are many and diverse: examples include airport ground traffic control, computing genome edit distances, testing protocols for logical flaws, or controlling a printer.Goals of this projectThe aim of research in planning, as in many other branches of AI, is to construct domain-independent ("universal") solutions for this kind of problem. That is, rather than solving each application problem individually, a general AI planning system should be able to solve any one of them, provided a formal specification of the problem as input. Several approaches to achieving this have been tried, such as different variations of search, or recasting the problem as another kind of general reasoning (e.g., a CSP, or logical deduction). Yet, efficient general automated planning remains a challenge.Within the area of AI planning, a wide variety of projects are possible, ranging from highly theoretical to practical, implementation-oriented, and anywhere in between. New projects based on student's ideas can also be considered. Students interested in AI planning are encouraged to enquire. Contact:E: patrik.haslum@rsise.anu.edu.auW: http://users.rsise.anu.edu.au/~patrik/ Various Combinatorial Optimisation ProblemsProject Code:AI_06Supervisor:Dr Patrik HaslumOutline:Although research in AI planning aims to construct fully general, domain-independent, planning systems, systems are often evaluated and compared by testing them on a collection of benchmark domains, accumulated over time by the research community. Many of these benchmarks resemble well-known combinatorial optimisation problems, such scheduling/sequencing, allocation, etc., or at least contain parts that do, but each also has its own peculiarities and tweaks. For a majority of benchmark domains, it is known that finding optimal solutions is difficult (NP-hard or worse) while just finding any solution is relatively easy (there exists efficient, and often simple,domain-specific algorithms), providing further evidence that the underlying problems encoded in these benchmark domains are, at heart, about optimisation. This project consists in picking up one (or more, depending on project time frame, problem difficulty, etc) of the commonly used planning benchmark domains that encode optimisation problems, and attack the underlying problem using whatever methods/tools seem most suitable, including for example LP/MILP, CSP/COP/SAT, local search methods, or anything else. Goals of this projectThe aim is to obtain a better understanding of the problems that underlie planning benchmarks -- how hard are they, in a practical, rather than complexity-theoretical, sense, and what instance parameters determine that hardness? how well dodomain-independent planning systems compare, in terms of effectiveness and solution quality, to a domain-specific solution? -- but also to learn something about the different combinatorial optimisation technologies -- how difficult are they to apply to a given problem? which is more suited to what kind of problem? Contact:E: patrik.haslum@rsise.anu.edu.auW: http://users.rsise.anu.edu.au/~patrik/ Natural Language Query Answering from WikipediaProject Code:AI_07Supervisor:Dr Scott SannerOutline:Have you ever wanted to ask your computer a question in natural language and have it respond with the answer you wanted? It sounds like a daunting task, but many of the foundations of such a technology are well understood and Wikipedia provides a wonderful open source natural language knowledge base to support this task.Goals of this projectIn this project, you will combine various off-the-shelf natural language processing (NLP) tools including part-of-speech taggers, parsers, and co-reference resolvers to build up a small demonstration system to answer simple natural language queries from Wikipedia.Contact:E: scott.sanner@nicta.com.auW: http://users.rsise.anu.edu.au/~ssanner/ Deep Opinion and Sentiment Mining from BlogsProject Code:AI_08Supervisor:Dr Scott SannerOutline:A fast-growing topic of interest in text mining is the automatic mining of opinions and sentiment from the blogosphere. What do Australians think about the emissions tax or the Kyoto protocol? What are the basic topics that people care most about in Australia ? Most current approaches to opinion and sentiment mining treat a document as an unordered "bag of words"; in this project, you will determine if a more accurate decomposition of natural language documents into structured features will yield better results with state-of-the-art opinion and sentiment mining techniques.Goals of this projectThe ultimate goal will be to see if you can can improve on our current opinion and sentiment mining algorithms applied to a few Gb of blog data.Contact:E: scott.sanner@nicta.com.auW: http://users.rsise.anu.edu.au/~ssanner/ Automated Feature Induction in Reinforcement LearningProject Code:AI_09Supervisor:Dr Scott SannerOutline:Reinforcement learning is the task of how to learn to make optimal sequential decisions when you can only sample experience from acting in a (simulated) environment. For example, the world's best backgammon player TD-Gammon taught itself to play Backgammon via reinforcement learning techniques. One of the greatest obstacles to practical reinforcement learning is finding the right features to use.Goals of this projectIn this project, you can examine your choice of automated feature induction methods ranging from functional gradient boosting to artificial neural networks to kernel machines. Application domains can range from learning to play a game (Othello, Backgammon, Go) to learning to control a small mobile robot. Your goal is to design a learning algorithm that not only learns to perform its assigned task but also learns important structured features when given only a set of primitive observations (e.g., game board state, tactile input, vision, etc...).Contact:E: scott.sanner@nicta.com.auW: http://users.rsise.anu.edu.au/~ssanner/ Mars Rovers and Traffic ControllersProject Code:AI_10Supervisor:Dr Scott SannerOutline:Markov decision processes (MDPs) are a theoretical tool for modelling sequential decision making problems and their optimal solution. Recent advances in the theory of MDPs permit efficient solutions to problems with both continuous state and action spaces. Such models are highly appropriate for planning in both Mars Rovers and Traffic Controllers (just to name two examples).Goals of this projectIn this project, you would choose one of these problem domains (or perhaps another you can suggest) and implement an (approximately) optimal planning system for this task. This project offers the chance for the student to learn about the theory of optimal sequential decision making and its application to practical problems.Contact:E: scott.sanner@nicta.com.auW: http://users.rsise.anu.edu.au/~ssanner/ General Named-entity Recognition using the WikipediaProject Code:AI_11Supervisor:Dr Wray BuntineOutline:A number of named-entity recognition systems are available in open source or "free for research" distributions. Most however, work better when augmented somehow with dictionary/gazeteer resources. In this project we wish to explore the use of the Wikipedia name resources, augmented with DBpedia (http://dbpedia.org/About) resources. We'd like to develop a general purpose, robust name recogniser with these.Requirements/Prerequisites
Background LiteratureSoftware discussed at http://en.wikipedia.org/wiki/Named_entity_recognitionContact:E: wray.buntine@nicta.com.auAutomated Hierarchical Classification of Web PagesProject Code:AI_12Supervisor:Dr Wray BuntineOutline:Classifying web pages according to the ODP Web directory data http://www.dmoz.org/, or alternative according to the hierarchy in the Wikipedia, is a useful task for many kindsof document analysis. The challenge is in processing the HTML and text content, in applying supervised learning (we at the SML group, http://sml.nicta.com.au/, have many such tools), in scaling to large hierarchies (100's to 1000's) and in scaling to an acceptable performance on a standard server. This is an open ended task that could use theoretical work (methods to address scaling and/or hierachical issues) or applied work (exploring the use of existing techniques), or both. Requirements/Prerequisites
Background Literature
Contact:E: wray.buntine@nicta.com.auImage Search via Kernelised SortingProject Code:AI_13Supervisor:Dr Wray BuntineOutline:Kernelised sorting does visual layout of images according to visual structures such as hierarchical plates, rotating spheres and fish-eye style graphs. We have a basic system developed but wish to further improve or create different visual structures, the subsequent user interface, and then to perform user testing.Requirements/Prerequisites
Background LiteratureNovi Quadrianto, Le Song, Alex J. Smola.Kernelized Sorting. Advances in Neural Information Processing Systems NIPS 22, Vancouver, B.C., Canada, 2009. Contact:E: wray.buntine@nicta.com.auMapping Machine LearningProject Code:AI_14Supervisor:Dr Mark ReidOutline:Machine learning is a very diverse field that studies a wide variety of problem types (e.g., classification, regression, dimensional reduction, ranking, density estimation, clustering) in a variety of modes (e.g., batch, online, active, supervised, semi-supervised, unsupervised). There is an increasing body of literature exploring how these various problems are related and, in some cases, how one type of problem can be reduced to another.Goals of this projectThe aim of this project is to build a "map" that presents an overview of machine learning problems and the relations between them. This will be achieved by identifying the most important types and modes of machine learning problems and then summarising and synthesising key papers for each and the results that relate them.Requirements/PrerequisitesAn interest and basic understanding of machine learning and good writing skills.Student GainThis project will provide the student with: an excellent overview of modern machine learning from a theoretical perspective; experience in understanding and synthesising existing research; and writing literature surveys. There may be an opportunity to publish the survey if it is comprehensive and well written.Contact:E: mark.reid@anu.edu.auW: http://mark.reid.name/work/ Efficient Algorithms for Spatial ReasoningProject Code:AI_15Supervisor:Assoc Professor Jochen RenzOutline:In computer science and engineering, spatial information (such as the spatial description of a room or a city) is often represented using coordinate systems, but this is not the way humans deal with spatial information. We often describe spatial features by specifying the relationship between objects in space (for example, the book is "on" the table, "to the left of" the screen). This is what is done in the area of Qualitative Spatial Representation and Reasoning, an important subfield of Artificial Intelligence. It is possible to define many different spatial relations for different aspects of space such as direction relations, distance relations, size relations etc. and to express spatial knowledge using these relations.Goals of this projectSince reasoning with these relations is a very complex problem, it requires a detailed theoretical analysis of the different relations in order to find efficient reasoning algorithms for different sets of relations. However, recent advances in the area allow to make this theoretical analysis and to find efficient reasoning algorithms automatically. In this project we look at different spatial calculi, test if we can find efficient algorithms automatically and try to find out what makes reasoning hard.Contact:E: jochen.renz@anu.edu.auW: http://rsise.anu.edu.au/~jrenz Spatial Representations for BiotechnologyProject Code:AI_16Supervisor:Assoc Professor Jochen RenzOutline:There are several ways of representing biological entities like proteins, molecules or cells, which are often too complex for applying data-mining, machine learning or other computational techniques. Many of the functional properties of biological entities and of their interactions depend on spatial properties, e.g. a protein might only interact with another protein if it is folded in a way that some of its parts are accessible from the outside to other proteins.Goals of this projectThe idea of this project is to find examples of biological properties and interactions where spatial relationships are important, identify and categorize these relationships, and based on this find simple functional representations of biological entities which could be used for applying computational techniques.Contact:E: jochen.renz@anu.edu.auW: http://rsise.anu.edu.au/~jrenz Trust and ReputationProject Code:AI_17Supervisor:Assoc Professor Jochen RenzOutline:In online transactions we often deal with people we do not know. How can we possibly be sure about who we can trust and who we should better not trust. A popular method is to use a reputation system which keeps track of the past performance of every user of the system and which assists users in forming their opinion about whether another user can be trusted. Even though many reputation systems are being used (the most prominent is probably the eBay Feedback System), their performance is not satisfactory at all and many fraudulent transactions are still happening.Goals of this projectThe aim of the project is to develop and implement a reputation system and to compare its performance with existing systems.Interested students should have experience with online transactions and existing reputation systems such as eBay. Contact:E: jochen.renz@anu.edu.auW: http://rsise.anu.edu.au/~jrenz
Human head-neck system modelingProject Code:CS_01Supervisor:Dr Karim SeghouaneOutline:In order to have a better understanding of the physiological human head-neck stabilization system which controls themotion of human's head in response to daily movements, the aim of this project is to study the sensorimotor systems responsible for stabilizing the head relative to space and assessing the relative importance of visual and vestibular reflexes. Models of the dynamics of the somatosensory, vestibular and visual feedback systems involved in the human head-neck stabilization system will be derived and its parameters identified based on experimental data. These data will be generated from an experimental setup that is currently under developement. Requirements/PrerequisitesThe project will require understanding of Matlab, for results analysis. The student will also be involved in organizing and developing the experimental setup.Contact:E: abd-krim.seghouane@nicta.com.auW: http://nicta.com.au/director/research/programs/seacs/people/abd_krim_seghouane.cfm Implementation and Control of robotic SWARMsProject Code:CS_02Supervisor:Dr Brad (Changbin) YuOutline:Inspired from biological swarms of bees, birds, ants, fish, etc., many researchers today study efficient coordination of teams of aerial, ground and underwater vehicles for various cooperative missions. A typical mission may involve forming a certain team structure (formation), maintaining or changing such a formation, intelligent guidance for cohesive motion of the whole team, etc. without a master device to drive the vehicles in the team. Efficient coordination of vehicles in such missions has various defence and civil applications.This project aims to utilize a testbed to build SWARM robotic applications involving teams of up to 10 wheel robots involving Surveyors robots, E-puck and Wifibots. The student will normally develop algorithms in Matlab (separately to this task) and validate using Webots Simulator before implementing with the testbed. Requirements/PrerequisitesInterested students should be capable of programming in Matlab and C/C++.Contact:E: Brad.Yu@anu.edu.auW: http://cecs.anu.edu.au/~bradyu Modelling of Complex Systems with Graphs and MatricesProject Code:CS_03Supervisor:Dr Brad (Changbin) YuOutline:Starting from real world observations about formations and abstracted them using graphs, be it undirected or directed, one obtains a direct view of the communication and information architecture among different nodes/agents/vertices. But how do we compare two graphs modelling two similar-looking system, we want to tell ultimately which one is better than the other. Various matrices, but not those of topological sort like an incidence matrix, are used to best capture the tiny difference that are visually negligible but potentially differ drastically.This project aims to study the transposition of the knowledge between two fundamental fields of mathematics. A number of former developed graph theoretical results are to be verified or advanced using theories and procedures in which matrices are extensively involved. Requirements/PrerequisitesInterested students should be capable of programming in Matlab, and have good knowledge in linear algebra and discrete mathematics.Contact:E: Brad.Yu@anu.edu.auW: http://cecs.anu.edu.au/~bradyu Switching or blending? - A comparative study on multiple model based observer designProject Code:CS_04Supervisor:Dr Weitian ChenOutline:Observer design aims to provide state estimation for dynamic systems based on limited information (measurements). For systems whose state variables are not available, observers are often required to be designed for the purpose of solving control problems. Therefore, observer design is an important research area in systems and control. Although it is solved well for an important class of systems called linear time invariant systems, observer design has experienced many difficulties in the course of moving toward complex systems such as time-varying systems and nonlinear systems.Multiple model based approach has a great potential to deal with those difficulties encountered in the observer design. Two different multiple model based methodologies have been proposed in the literature. One switches amongst observers designed for all models, while the other blends all those observers designed to form an overall observer. The question now is: which is better: switching or blending? This project is to carry out a comparative study on the above mentioned observer design strategies through extensive simulations on different complex systems such as time-varying systems, nonlinear systems, and switched control systems. Those simulations will be carefully designed to reveal the advantages and disadvantages of each method. Furthermore, the possibility of new observer design will be explored. Requirements/PrerequisitesInterested students should have some knowledge of observer design and be familiar with programming using Matlab.Contact:E: weitian.chen@anu.edu.auGeneralized Dynamic Factor Models for Time Series with Multiple Frequencies - A literature review projectProject Code:CS_05Supervisor:Dr Weitian ChenOutline:Generalized dynamic factor models is a recently developed tool in econometric modeling, which has been used in modeling of national economies.The tool has its root in factor analysis, which was developed by psychologists commencing round the beginning of the 20th century to test the mental ability of a person (leading to the now well known IQ test). Goals of this projectThe goal of this project is to introduce the summer scholar into a fascinating new interdisciplinary research problem and to equip the student with some basic knowledge related to the subject and some deeper knowledge about the past and present research in the area of generalized dynamic factor models for time series with multiple frequencies.Requirements/PrerequisitesThe student is required to collect those publications in the subject area, to read and understand as much as they can, and to provide a literature review report on the subject with the significant help from the supervisors. We will also be seeking to have the student trial algorithms on a data set.No particular preliminary knowledge or skill is required except the passion to learn and explore new research problems. Every student interested is welcome to apply. Contact:E: weitian.chen@anu.edu.auUser friendly navigation with landmarksProject Code:CS_06Supervisor:Dr Lars PeterssonAssoc Professor Jochen Renz Outline:Everyone that have been using a GPS system knows that some of the navigational descriptions are very ambiguous and does not always help very much. Using landmarks such as churches, petrol stations etc seems to be a much more natural way of describing a route. For example, turn left at the petrol station. In this project we combine computer vision with route representation to come up with a more intuitive way of describing routes.The project has two parts. 1) Use a vehicle mounted camera to detect useful landmarks and their visibility along a route. 2) Develop a route representation that allows us to compute optimal route descriptions based on landmark visibility. The project can be carried out by one student or two that collaborate. Contact:E: lars.petersson@nicta.com.auW: http://nicta.com.au/director/research/programs/asst/people/lars_petersson.cfm E: jochen.renz@anu.edu.au W: http://rsise.anu.edu.au/~jrenz Symbolic Tools for Quantum Feedback NetworksProject Code:CS_07Supervisor:Professor Matt JamesOutline:Future quantum technologies will need methods for describing largecomplex networks of quantum components. The aim of this project is to use the symbolic capabilities of Mathematica or Maple to develop tools for describing and simplifying quantum circuits. Contact:E: Matthew.James@anu.edu.au
Visualizing System Health via Continuous Regression DataProject Code:C&PS_01Supervisor:Dr Stephen BlackburnOutline:One of the greatest challenges to a large, continuously evolving software system lies in efficiently identifying regressions in correctness or performance. To combat this, we typically employ a wide range of continuous regression tests. These tests evaluate the system continuously, every day of the year, on a range of hardware and operating systems, across a range of scenarios, testing for both regressions correctness and performance. While such systems can generate vast amounts of data over time, this data can be hard to meaningfully mine, and in particular, may not directly help our goal of associating regressions with particular commits. This is particularly true in the case where the regression is subtle or intermittent.Goals of this projectThis project will focus on data presentation and data mining. The goal will be producing graphics and visuals that allow a large open source research project to effectively and quickly identify regressions and associate them with particular commits. Experience shows that such tools can greatly improve the productivity of a large project.Contact:E: Steve.Blackburn@anu.edu.auW: http://cs.anu.edu.au/people/Steve.Blackburn PC-Relative Branching For a High Performance CompilerProject Code:C&PS_02Supervisor:Dr Stephen BlackburnOutline:The optimizing compiler used by Jikes RVM (formerly known as Jalapeno) is a leading-edge compiler, use in many research publications, and the basis for Jikes RVM's excellent performance (performing as well as Sun, IBM and Oracle JVMs on many important benchmarks). However, the compiler currently uses indirection rather than PC-relative branching for calls. This design is a legacy of the VM's flexibility and a now-deprecated feature that allowed code to be moved by the system's garbage collector.Goals of this projectThis project will implement PC-relative branching in the IA32 version of Jikes RVM. The project will likely involve collaboration with colleagues at IBM's TJ Watson Research Labs. There is good reason to expect that this will yield a noticeable performance improvement to this already competitive JVM. We expect the student to enjoy working with a high performance compiler, interacting with top people from industry and academia, and with some luck, being rewarded with a noticeable performance improvement.Contact:E: Steve.Blackburn@anu.edu.auW: http://cs.anu.edu.au/people/Steve.Blackburn Faster Dynamic Execution through Smarter Adaptive CompilationProject Code:C&PS_03Supervisor:Dr Stephen BlackburnOutline:High performance runtime systems use adaptive compilation to perform feedback-directed optimization (FDO). One of the original, and more advanced such systems is the Adaptive Optimization System (AOS) used by Jikes RVM (formerly known as Jalapeno), and developed by Matthew Arnold et al. Such systems observe execution of a program, and based on those observations, a) select particular code for optimization, and b) target specific optimizations and levels of optimization at that code according to its dynamically observed characteristics. Jikes RVM's AOS is very aggressive, allowing speculative optimization and on-stack replacement (OSR), which means even a very long running hot loop can be optimized and replaced on the fly without breaking out of that hot loop.Goals of this projectThis project will explore and implement a number of strategies for improving the effectiveness of the AOS, with guidance from Matt Arnold and other colleagues of ours at IBM Research. These strategies include different strategies for sampling (which is used to drive the AOS's decisions), including sampling driven by hardware events rather than a timer, and sample rates changing over time to maximize the effectiveness of the tradeoff between the cost of sampling and the quality of the information gleaned by sampling. This project has the potential to improve the performance of a high performance research JVM used by researchers all over the world.Contact:E: Steve.Blackburn@anu.edu.auW: http://cs.anu.edu.au/people/Steve.Blackburn Porting Open JDK's Libraries to Jikes RVMProject Code:C&PS_04Supervisor:Dr Stephen BlackburnOutline:A key element of any modern runtime system is its class libraries. For many years, Jikes RVM, a high quality research JVM used by researchers worldwide, has used the Classpath open source class libraries. More recently, Sun has made a version of its class libraries publicly available under an open source license. These libraries are more complete and most likely better performing than the Classpath libraries.Goals of this projectThis project will involve porting the Open JDK libraries to Jikes RVM. We expect that this will allow Jikes RVM to run important workloads that it currently cannot, and we hope that it will further improve the performance of Jikes RVM. The project will be technically challenging, but should be very rewarding and will involve working with top researchers and JVM developers from industry and academia.Contact:E: Steve.Blackburn@anu.edu.auW: http://cs.anu.edu.au/people/Steve.Blackburn Optimizing Reference Counting Garbage CollectorsProject Code:C&PS_05Supervisor:Dr Stephen BlackburnOutline:An important class of garbage collection algorithms use reference counting. For each object, a count is kept of all incoming references. When the count of incoming references reaches zero, the object is clearly unreferenced and may therefore be collected. Reference counting has a number of interesting advantages, including that it depends only on local information to determine the liveness of an object. However, there are also a number of challenges to using reference counting, including the problem of detecting and collecting cyclic structures, and minimizing the overhead of reference counting operations and the space required to keep each object's count.Goals of this projectThis project will implement and evaluate a number of novel optimizations for reference counting in the context of MMTk, the memory management toolkit within the high performance Jikes RVM runtime system.Contact:E: Steve.Blackburn@anu.edu.auW: http://cs.anu.edu.au/people/Steve.Blackburn A Computer Vision Java Library for the CellProject Code:C&PS_06Supervisor:Dr Eric McCreathOutline:Computer vision libraries are important foundations for developing applications in areas such as: motion tracking, face recognition, object identification, gesture recognition, etc. This project aims at developing part of a computer vision library in Java which uses the Synergistic Processing Elements(SPEs) of the Cell Broadband engine. This will enable much of the computational burden of image processing to be offloaded to the SPEs freeing up the Power Processing Element(PPE) for other tasks.Goals of this projectThis project will focus on understanding the overheads in interfacing Java with the SPEs of the Cell processor for computer vision applications. The goal is to produce a low latency efficient library that works for multi-threaded applications.Contact:E: ericm@cs.anu.edu.auW: http://cs.anu.edu.au/people/Eric.McCreath Fast Multipole Algorithm vs the Particle Mesh Ewald methodProject Code:C&PS_07Supervisor:Dr Eric McCreathOutline:The simple n-body algorithm for calculating electrostatic interaction is O(n^2) where n is the number of particles. The particle mesh Ewald method enables the same computation to be done in O(n lg n) time, whereas, the fast multipole algorithm reduces this to O(n). Interestingly, the Ewald technique is more commonly used by molecular dynamics systems even though it is inferior in terms of time complexity. This project would involve implementing directly comparable versions in Java of these three algorithms to help understand their performance differences.Goals of this projectThe goal would be to obtain a deeper understanding of the performance of these algorithms such that a more informed choose between them can be made. Also this project would aim to understand how such problems perform in a high level language such as Java. This project would feed into other projects which involve implementing these approaches on novel hardware such as the Cell or GPUs.Contact:E: ericm@cs.anu.edu.auW: http://cs.anu.edu.au/people/Eric.McCreath
Performance Modelling Service Oriented Architecture (SOA) applications on Enterprise Service Buses (ESB)Project Code:EG_01Supervisor:Dr Paul BrebnerOutline:3 sub-projects: Performance data monitoring approaches, performance modelling application variations, and Virtualization and modelling experiments.Following the success of our last years summer scholar project (a trial of NICTA's service-oriented performance modelling technology applied to the MULE Enterprise Service Bus (ESB) LoanBroker Application), we propose to extend this work as follows. Performance data monitoring approaches Investigation of requirements and approaches for capturing performance data from SOA/ESB infrastructures and applications. This sub-project involves understanding the requirements for capturing performance data for performance modelling, and investigating solutions for capturing data in experimental contexts from one or more open source ESBs and applications. We may trial some open source and commercial SOA/ESB monitoring tools as part of this project. Performance modelling application variants Conduct experiments and develop performance models of different architectural styles/variants of a ESB application. This follows on from last years work which was successful in building a model of one variant of the MULE LoanBroker application. This year we aim to build models for a number of variants of the application, and investigate the accuracy and power of the performance models to predict performance and scalability for known and hypothetical variants. Virtualization and modelling experiments This sub-project will involve deploying the SOA/ESB infrastructure and applications on virtualized servers and conducting experiments to determine if the performance models developed above correctly predict scalability and performance. We also plan to investigate the ability of virtual servers to isolate service consumers, and provide specified Quality of Service levels (such as maximum service response times) for different consumers. Requirements/PrerequisitesWe require 3 students with interest in, and experience with software engineering, software architectures, web services, SOA, ESB, Java, software performance and scalability.Background Literature
Contact:E: paul.brebner@nicta.com.auW: http://www.nicta.com.au/people/brebnerp
Image Interpretation Based on Human-Computer InteractionsProject Code:HCC_01Supervisor:Dr Jun ZhaoOutline:Current image processing and pattern recognition algorithms are not robust enough to make automated image interpretation feasible in many real-world applications. For this reason, we need to develop image interpretation systems that rely on human guidance. It is important to retain the 'human in the loop' where a human operator can aid the automatic image interpretation through human-computer interactions (HCI).Goals of this projectThis project aims at developing a semi-automatic image interpretation algorithm for feature detection. The tasks include collecting human inputs using an already developed graphical user interface, and then implement and extend existing online machine learning algorithm for feature tracking to 2D feature selection and image segmentation problems. Experiments will be performed on two datasets that involve remote sensing imagery and hyperspectral images.Requirements/PrerequisitesInterested students should have experiences in programming with Matlab and have basic knowledge on image processing and pattern recognition. The student will work closely with the supervisors and postgraduate students in NICTA.Contact:E: jun.zhao@anu.edu.auRelighting for face swappingProject Code:HCC_02Supervisor:Dr Roland GoeckeOutline:In [Blanz04], a 3D Morphable Model (3DMM) based system that exchanges faces across large differences in viewpoint and illumination was proposed. On similar lines, a complete model-free approach for automatic face replacement in images was proposed in [BKD08]. Both of these "Cool!" applications of state-of-the-art computer vision research were highly appreciated in the graphics community, but have a number of drawbacks. Recently, we proposed a novel regression based technique [Asthana09] to synthesize the face of any unseen person at any random pose and expression using just a single frontal image. The approach is completely 2D and real-time.Goals of this projectThe overall aim of this project is to build a system using the already existing framework that can swap faces between any two people at any random pose and illumination. The summer scholar is expected to work on the subproblem of "Face Relighting". Face relighting is one of the most exciting and important modules of the project for producing realistic looking results. In this, given a source and destination face images, the goal is to map the illumination from the source to the destination face image.Requirements/Prerequisites
Student GainBackground LiteratureW:research.microsoft.com/en-us/um/people/zliu/cvpr2003.pdf W: Blanz, K. Scherbaum, T. Vetter and H.P. Seidel, Exchanging Faces in Images. In Proceedings of EUROGRAPHICS 2004, Grenoble, France, 2004. [Blanz04] W: [BKD08] D. Bitouk, N. Kumar, S. Dhillon, P. N. Belhumeur, and S. K. Nayar, Face Swapping: Automatically Replacing Faces in Photographs. ACM Trans. on Graphics (also Proc. of ACM SIGGRAPH), Aug, 2008. W: [AAsthana09] A. Asthana, R. Goecke, N. Quadrianto and T. Gedeon, Learning Based Automatic Face Annotation for Arbitrary Poses and Expressions from Frontal Images Only. In The IEEE Computer Society Conference on Computer Vision and Pattern Recognition [ CVPR 2009 ], Florida, USA, June 2009. Contact:E: roland.goecke@anu.edu.auW: http://users.rsise.anu.edu.au/~roland Face generation at any poseProject Code:HCC_03Supervisor:Dr Roland GoeckeOutline:In [Blanz04], a 3D Morphable Model (3DMM) based system that exchanges faces across large differences in viewpoint and illumination was proposed. On similar lines, a complete model-free approach for automatic face replacement in images was proposed in [BKD08]. Both of these "Cool!" applications of state-of-the-art computer vision research were highly appreciated in the graphics community, but have a number of drawbacks. Recently, we proposed a novel regression based technique [Asthana09] to synthesize the face of any unseen person at any random pose and expression using just a single frontal image. The approach is completely 2D and real-time.Goals of this projectThe summer scholar will have an opportunity to work on the subproblem of "Generating a Face at Any Desired Pose" using the 3D model constructed by our existing system. Estimated pose information will be provided by FaceAPI (Seeing Machines software). Using this pose information, the parameters of the 3D model can be varied to generate the face at any random pose. The goal here is to develop an algorithm to "optimally" vary the 3D model parameters to generate the face image at a desired pose and perform face swapping between the source and destination face images, ignoring factors such as illumination and expression variations.Requirements/PrerequisitesStudent GainBackground LiteratureW: Blanz, K. Scherbaum, T. Vetter and H.P. Seidel, Exchanging Faces in Images. In Proceedings of EUROGRAPHICS 2004, Grenoble, France, 2004. [Blanz04]W: [BKD08] D. Bitouk, N. Kumar, S. Dhillon, P. N. Belhumeur, and S. K. Nayar, Face Swapping: Automatically Replacing Faces in Photographs. ACM Trans. on Graphics (also Proc. of ACM SIGGRAPH), Aug, 2008. W: [AAsthana09] A. Asthana, R. Goecke, N. Quadrianto and T. Gedeon, Learning Based Automatic Face Annotation for Arbitrary Poses and Expressions from Frontal Images Only. In The IEEE Computer Society Conference on Computer Vision and Pattern Recognition [ CVPR 2009 ], Florida, USA, June 2009. Contact:E: roland.goecke@anu.edu.auW: http://users.rsise.anu.edu.au/~roland Facial Expression Animation using Active Appearance ModelsProject Code:HCC_04Supervisor:Dr Roland GoeckeOutline:The use of Active Appearance Models (AAM) is becoming popular in face recognition and facial expression recognition systems. AAMs are a computer vision technique for matching statistical models of shape and appearance to new images. The model is first built or trained using a number of facial images and then it is used to deform or warp to new and previously unseen images, thus providing information such as the location of the eyes, nose, mouth etc. The model itself, built from multiple facial images from the training phase can be visualised as a combined image and a "morph-like" affect can be produced by altering the shape and appearance parameters.If we analyse a video, capturing perhaps 30 frames per second and then, using the model to find the facial features in each frame, we can save away the shape and appearance parameters that were used to warp to each facial image. If we later apply the collected shape and appearance parameters to the model in the same order they were captured then, using just the model and the parameters, we should be able to simulate the video. This has some interesting possibilities. For example, the model and the parameters would typically be much smaller in size that the sum of the images it represents. The shape and appearance parameters can be applied to a different set of facial images thus simulating facial expressions on a different subject. Goals of this projectThis project aims at developing a proof of concepts system to reconstruct video based on replaying the previously extracted shape and appearance parameters.Requirements/PrerequisitesIdeally, the system will be build in C++ and interested students should have experience or a very good ability in programming C++ together with a basic knowledge of image processing and pattern recognition. The student will work closely with the supervisors and postgraduate studentsStudent GainThe project will involve working with researchers in a very innovative and emerging field of study. The student will gain hands-on experience in building this system and in working with C++.Background LiteratureW: Blanz, K. Scherbaum, T. Vetter and H.P. Seidel, Exchanging Faces in Images. In Proceedings of EUROGRAPHICS 2004, Grenoble, France, 2004. [Blanz04]W: [BKD08] D. Bitouk, N. Kumar, S. Dhillon, P. N. Belhumeur, and S. K. Nayar, Face Swapping: Automatically Replacing Faces in Photographs. ACM Trans. on Graphics (also Proc. of ACM SIGGRAPH), Aug, 2008. W: [AAsthana09] A. Asthana, R. Goecke, N. Quadrianto and T. Gedeon, Learning Based Automatic Face Annotation for Arbitrary Poses and Expressions from Frontal Images Only. In The IEEE Computer Society Conference on Computer Vision and Pattern Recognition [ CVPR 2009 ], Florida, USA, June 2009. Contact:E: roland.goecke@anu.edu.auW: http://users.rsise.anu.edu.au/~roland Fuzzy Signature BuilderProject Code:HCC_05Supervisor:Prof Tom GedeonDr Sumudu Mendis Outline:The fuzzy signature concept has been enhanced and popularized by the iHCC research group at the SoCS in ANU. Fuzzy signature concept is a non-traditional computational intelligence approach to solve complex structured data problems. It has been successfully applied in medical diagnosis, communication in cooperative robots, cognitive modeling, intelligent data analysis, human eye gaze path analysis to predict their cognition, etc. This project involves working with a sub group of iHCC research group on developing an automated fuzzy signature builder. This involves developing a front end GUI to input manual fuzzy signatures and use genetic programming toolbox (Matlab) to automated generation of fuzzy signatures into the existing fuzzy signature inference engine (Matlab). In overall this project need research and development in interfacing Matlab programs with Java or any other multi-platform software, computational intelligence topic of genetic programming, and HCIGoals of this projectThe goals of this project is to develop a multi-platform software package, which embedded Matlab's powerful technical computing ability, for fuzzy signaturesRequirements/PrerequisitesEnthusiasms for research and development in computational intelligence topics, and solving new and in depth issues in programmingStudent GainThe student will get benefits to interact with one of school's highly publishing research team. In addition, this is a application oriented research that benefit you to enhance your software engineering, HCI, and mathematics skills.Background LiteratureMendis, B. S. U. & Gedeon, T. D. (2008) Aggregation Selection for Hierarchical Fuzzy Signatures: A Comparison of Hierarchical OWA and WRAO. 1-8.Mendis, B. S. U. (2008) Fuzzy Signatures: Hierarchical Fuzzy Systems and Applications (PhD thesis). College of Engineering and Computer Science, The Australian National University, Australia. Gedeon, T. D., Zhu, D. & Mendis, B. S. U. (2008) Eye gaze assistance for a Game-like interactive task. International Journal of Computer Games Technology, vol 2008, 1-10. Contact:E: tom@cs.anu.edu.auW: http://cs.anu.edu.au/people/~Tom.Gedeon E: sumudu@cs.anu.edu.au W: http://cs.anu.edu.au/~Sumudu.Mendis/
Visualising Proof SearchProject Code:LC_01Supervisor:Dr John SlaneyOutline:Automated reasoning systems search for proofs by repeatedly making inferences from what they already know, thus collecting a pool of known consequences that can be used as input for more inferences. In order to understand this process better, we need to apply visualisation tools to produce animated pictures of what is going on. Little research has been done on this, so it is very much an open area.Goals of this projectWe shall use a generic information visualisation tool currently under development in NICTA's Canberra Lab, applying it to runtime data extracted from automated reasoning software to reveal interesting features of the proof search process. One outcome will be better insight into the process of reasoning. Another will be that we serve as a case study for improving the visualilsation tool.Requirements/PrerequisitesImagination, creative thinking, willingness to give ideas a go. Some competence with scripting languages or Java programming skills would be useful. Deep knowledge of automated theorem proving is not required, though some familiarity could help. Many theorem provers are in C or C++, so those languages would be relevant.Contact:E: john.slaney@anu.edu.auW: http://arp.anu.edu.au/~jks An Automated Theorem Prover That can do ArithmeticProject Code:LC_02Supervisor:Dr Peter BaumgartnerOutline:An automated theorem prover is a program that tries to automatically prove valid (or disprove) formulas given to it. Many applications of automated theorem proving require reasoning modulo some form of (integer) arithmetic, e.g. for software verification. Unfortunately, theory reasoning support for the integers in current theorem provers is too weak for practical purposes. This summer scholar project is about implementing a theorem prover that can do better. The theory behind it has already been developed (by and large). A nice outcome would be a prototypical implementation and running some experimentswith it. There is ample room to refine the theory and bring in own ideas. Contact:E: Peter.Baumgartner@nicta.com.auW: http://rsise.anu.edu.au/~baumgart Logical Analysis of Business RulesProject Code:LC_03Supervisor:Dr Peter BaumgartnerOutline:Business Rules are widely used in industry to specify the conditions or constraints that control the operations of a business. For instance, business rules might reflect legislation governing visa regulations, or conditions under which social benefits are granted.Goals of this projectThe background of this project is a running project between NICTA and a Canberra-based SME. We are building a prototype system that allows to analyze sets of business rules for logical errors, such as inconsistencies and functional dependencies. Spotting such errors is of great importance in the development phase of business rules. The goal is to help with extending our already existing prototype by enhancing its functionality (both the user front-end and the internal reasoning engine).Requirements/PrerequisitesGood Java programming skills. Some background in mathematical logic and/or constraint processing would be useful.Student Gain
Contact:E: Peter.Baumgartner@nicta.com.auW: http://rsise.anu.edu.au/~baumgart Advances in Practical SAT SolvingProject Code:LC_04Supervisor:Dr Jinbo HUANGOutline:Satisfiability (SAT) is the problem of determining whether a Boolean formula can evaluate to true (i.e., be satisfied) under some assignment of truth values to its variables. For example, the formula ((X or (not Y)) and (X or Y or (not Z)) and ((not X) or (not Y) or (not Z))) can be satisfied by assigning false to all three variables.Problems in many areas of AI and computer science can be reduced to SAT. Although SAT is NP-complete, problem instances arising from real-world applications often have special structure allowing them to be solved efficiently. Current SAT solvers are frequently able to solve instances with over a million variables. Goals of this projectThis project will involve a review of the latest SAT technology, particularly a class of algorithms known as clause learning, followed by an attempt to advance the state of the art by proposing new algorithms or improvements to existing algorithms. The final product could be either an extension of an existing SAT solver or an entirely new solver, which may then be considered for entry into the next International SAT Competition.Contact:E: Jinbo.Huang@nicta.com.auW: http://rsise.anu.edu.au/~jinbo Logical Analysis of Software Package ManagementProject Code:LC_05Supervisor:Dr Jinbo HUANGOutline:In software package management a major task is to maintain dependencies among packages when installing or removing packages while at the same time retaining certain desirable properties of the package repository. For example, installing a package X may require the removal of an existing package Y, but Y may be depended on by another existing package Z which cannot be removed. A possible solution may then lie in the fact that Z actually depends on the disjunction of Y and some W (that is, Z can survive with either Y or W). Hence replacing Y with W will resolve the dilemma.Desirable properties of a package repository include, for example, upgradeability, which ensures that the user can always upgrade a package to a newer version without having to remove previously installed packages that they need to use, although it may require the installation, upgrade, or replacement of other packages. Goals of this projectFor large repositories and complex dependencies, such analysis can be a challenge. Some of these tasks are now well understood in formal terms. For example, package installability is known to be NP-complete and an encoding to SAT is known. The aim of this project is to extend the scope of this knowledge by formulating additional queries and properties encountered in package management in formal logic, and studying their theoretical complexity as well as practical solution methods.Contact:E: Jinbo.Huang@nicta.com.auW: http://rsise.anu.edu.au/~jinbo Automated Reasoning for Artificial IntelligenceProject Code:LC_06Supervisor:Dr Rajeev GoreOutline:Much current research in Artificial Intelligence currently utilisesnon-classical logics called modal logics. These logics can capture notions like time, space, obligation, knowledge, belief etc. which classical logic cannot do as easily. But the interests of AI researchers rarely looks at the task of actually building efficient reasoners for such logics. We have recently developed automated theorem provers for very expressive modal logics. Are they any good for AI research ? Goals of this projectThe goal of the project is to build more specialised automated reasonersfor logics with applications in Artificial Intelligence. We have a framework, we need help to tailor it to the needs of AI. Requirements/PrerequisitesA strong background in maths or logic would be extremely useful but isnot necessary. Of more importance is enthusiasm and an interest in the formal side of AI. You will need to learn the theory of modal logics and how to automated them. Then you will need to learn the uses of these logics in AI. Finally, you will need to use our technology to build the provers that are needed by AI researchers. This project will introduce you to an area of theoretical computer science which mixes theory and practice and may lead to a publication. It will form a good basis for doing a Phd, either with us, or at another university. Contact:E: rajeev.gore@anu.edu.auW: http://arp.anu.edu.au/~rpg
Properties of Flexible Solar ModulesProject Code:PVS_01Supervisor:Dr Vernie EverettDr Elizabeth Thomsen Outline:Micro-modules of flexible silicon solar cells are being developed for applications such as powering small portable electronic devices. The requirements for these modules are that they are highly efficient, lightweight, flexible, and able to operate under a range of light conditions. Monocrystalline silicon is capable of high efficiencies; however at the thicknesses used in most photovoltaic (PV) applications it does not display the necessary flexibility. If the silicon is thinned to less than 50 microns, it can provide high flexibility such that the cells can be wrapped around a finger. Hence, the modules being developed require very thin silicon cells along with flexible connections and packaging. This project involves testing the modules under a range of light conditions, as well as exploring mechanical properties such as flexibility.E: Dr Elizabeth Thomsen Dr Vernie Everett Contact:E: vernie.everett@anu.edu.auE: elizabeth.thomsen@anu.edu.au Iron imaging in multicrystalline silicon wafers for solar cellsProject Code:PVS_02Supervisor:Dr Daniel MacDonaldOutline:Iron is one of the most important metallic impurities in multicrystalline silicon solar cells. It has a negative impact on solar cell performance because it acts as a strong recombination centre, causing the loss of photo-generated electrons. We have recently developed a method to create very sensitive and high spatial resolution images of the iron content across a silicon wafer, based on a pair of photoluminescence (PL) images. This project would involve applying this new technique to a range of multicrystalline silicon wafers that have undergone different treatments to try and remove or neutralize the iron.Goals of this projectFirstly, to implement the iron imaging technique into our new PL imaging system, which may involve some development of computer applications to manipulate the raw PL data and display the results. The main part of the project would then be to apply this technique to various multicrystalline silicon wafers, in order to better understand the behavior of iron in this material, and how its negative impact can be reduced. If there is sufficient time, the project could be extended to develop chromium imaging, another important impurity, and which in principle can be done in a similar fashion to iron imaging.Requirements/PrerequisitesAn interest in photovoltaics and solar energy. Some knowledge of semiconductors would be helpful. Some experience in developing measurement software applications would also be beneficial.Student GainYou would learn a lot about silicon solar cells, and state-of-the-art characterization techniques.Background LiteratureD. Macdonald, J. Tan and T. Trupke, Journal of Applied Physics, volume 103, article 073710 (2008).Contact:E: daniel.macdonald@anu.edu.auW: http://engnet.anu.edu.au/DEresearch/semiconductor/ Assessing cheap silicon for solar cellsProject Code:PVS_03Supervisor:Dr Daniel MacDonaldDr Keith McIntosh Outline:Most solar cells are fabricated from silicon. The silicon is crystalline and relatively pure, which makes it expensive to produce, but some manufacturers are reducing its cost by crystallising the silicon more rapidly and by using less pure reactants. One problem faced by these manufacturers is how to assess their silicon. Usually, silicon is assessed by measuring its conductance under illumination to calculate the "lifetime" of its electrons. The cheap silicon prevents this measurement at room temperature because crystal defects "trap" electrons preventing them from contributing to the silicon's conductance. At the ANU, we have recently developed an instrument to circumvent the inhibitive "trapping" by performing the measurements at high temperatures.Goals of this projectThe project involves the assessment of cheap silicon over a range of temperatures. It would suit a student who enjoys experimentation and physics. The student would learn theory relevant to solar cells and semiconductors, and would gain experience in experimental physics.Contact:E: daniel.macdonald@anu.edu.auW: http://engnet.anu.edu.au/DEresearch/semiconductor/ E: keith.mcintosh@anu.edu.au China project – 1: Study of cooling liquidsProject Code:PVS_04Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To determine a group of cooling liquids suitable for a concentrator receiver and the appropriate sealants.Goals of this projectSelection of cooling liquids with adequate optical properties.Selection of cavity sealants. - Measurement of the refractive index. - Measurement of the variation of the refractive index with temperature. - Transmittance and absorbance of the fluids – wavelengths. - Accelerated tests – UV light exposure. - Measurement of the chemical compatibility between the cooling liquid and the sealant. Accelerated tests (thermal expansions of the liquids). Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au China project – 2: Electrical and thermal characterization of immersed cellsProject Code:PVS_05Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To characterise the electrical and thermal performance of liquid immersed cells.Goals of this projectSelection of cooling liquids with adequate optical properties.Immersed solar cells. - Electrical characterization of cells immersed in a cooling liquid (preparation of a box to immerse the cells, well sealed, to conduct the tests). I-V curves. - Thermal characterization of the cells immersed in a cooling liquid (preparation of experiment, variation of Tliquid variation of Tcell). - Transmittance and absorbance of the fluids – wavelengths. - Electrical and thermal characterization (different tests at different temperatures, Flash I-V). Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au China project – 3: Accelerated and outdoors tests on liquid immersed cellsProject Code:PVS_06Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To determine the chemical compatibility of the liquids and cells in the long-term.Goals of this projectSelection of cooling liquids with adequate optical properties.Immersed solar cells. - Thermal cycling tests. - Damp heat tests. - Humidity freeze tests. - UV light exposure tests. - Initial and post-characterization (attention to electrical insulation tests). Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au China project – 4: Thermal modelling of a liquid immersed-cells receiverProject Code:PVS_07Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To model the thermal performance of the liquid immersed-cells receiver, with different light profiles and temperatures using a finite-element method (software).Goals of this projectCreate a Theoretical model of the receiver: materials, properties, dimensions.- Heat transfer in the receiver under different light profiles and intensities: distribution of temperatures. - Thermal expansions and mechanical effects. Requirements/PrerequisitesExperience with finite-element method software.Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au APP project – 1: Accelerated tests of linear CPV receiversProject Code:PVS_08Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To determine the long-term performance of the receiver.Goals of this projectTesting:- Thermal cycling tests. - Damp heat tests. - Humidity freeze tests. - UV light exposure tests. - Initial and post-characterization. Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au APP project – 2: Low concentration photovoltaic systems for Building Integrated Photovoltaics (BIPV)Project Code:PVS_09Supervisor:Dr Vernie EverettDr Marta Vivar Outline:To determine the opportunities in the photovoltaic market for low concentration PV systems. To identify the aspects that must be considered for the building integration.Goals of this project- Architectural aspects for BIPV.- CPV requirements. - Low CPV systems vs. BIPV: differences with ground installations and differences with flat panels (quantification of losses vs. aesthetics req.). - Analysis of the Australian photovoltaic market for BIPV (feed-in-tariffs, regulations, applications, etc). - Specific analysis for thermal, PV or Hybrid. Requirements/PrerequisitesMicroconcentrator project background. Thermal, PV or Hybrid (PV/Thermal).Contact:E: vernie.everett@anu.edu.auE: marta.vivar@anu.edu.au
How to Make a Robot Walk/Run/Hop/Jump/..Project Code:R&CV_01Supervisor:Dr Roy FeatherstoneOutline:How can we make a robot walk? What does the control system actually do? What are the scientific principles of walking, running, etc.?Walking is something that humans do very easily, but robots find rather hard. When a human walks, every footstep is different: the exact timing and placement of each foot, and the thrust delivered by each foot on the ground, are modified at each step in order to maintain balance, accommodate uneven terrain, adjust overall speed and heading, and so on. Robots can do this too, but not very well. Goals of this projectThis project is a simulation study of robots with legs. It includes accurate models of robot bodies, the surfaces they walk on, and the control systems that make then walk, run, etc. The objective is to experiment with different control systems, different walking/running styles and different designs of robot legs, in order to see which ones work best. Another objective is to make some cool movies. You can see some movies made by other students here: Project PageRequirements/PrerequisitesTo do this project, you will need a basic knowledge of how physical processes (like motion) are simulated on a computer: the process is modelled as a differential equation, and the computer solves the equation by numerical integration. (Essentially, it's an initial-value problem.) You should alsobe acquainted with the concept of feedback control systems, or, failing that, be acquainted with the use of feedback in electronic circuits. The project will make use of Matlab and Simulink, and it is desirable that you be familiar with both. Graphics programming skills would also be useful. This project offers you the chance to learn about advanced methods in modelling physical reality on a computer, and to learn how a control system can make a robot walk/run/etc. without falling over. It is likely you will be part of a small team working on this project. Contact:E: roy.featherstone@anu.edu.auW: http://users.rsise.anu.edu.au/~roy A Visualization Tool for Hyperspectral ImagesProject Code:R&CV_02Supervisor:Dr Antonio Robles-kellyOutline:A hyperspectral image contains much more bands than the traditional RGB color image. It provides us with richer information of the world through the "fingerprints" of objects across the electromagnetic spectrum.Goals of this projectThe aim of this project is to develop a visualization tool for hyperspectral images. Tasks include displaying the "fingerprints" in 3D space, as well as computing and visualizing feature extracted from these fingerprints.Requirements/PrerequisitesThis project requires experiences in Matlab or C/C++, and basic knowledge on linear algebra and statistics. The student will work closely with the supervisors and postgraduate students in NICTA.Contact:E: antonio.robles-kelly@nicta.com.auVisual Processing for the Bionic EyeProject Code:R&CV_03Supervisor:Dr Paulette LiebyOutline:Researches and develops vision processing algorithms to assist vision-impaired individuals with ambulatory navigation and general interaction with their environment.Requirements/PrerequisitesThe VIBE project at NICTA are seeking three summer scholars to help us in this endeavour.http://nicta.com.au/research/projects/visual_processing_for_the _bionic_eye Metlab and/or C/C++ (preferably both) and a basic knowledge of computer vision. Contact:E: Paulette.Lieby@nicta.com.auContent-based Image Retrieval with local invariant featuresProject Code:R&CV_04Supervisor:Dr Lei WangOutline:Extensive usage of digital imaging equipments has lead to an explosion of the number of images. This can be easily seen from the Internet, public digital libraries and personal photo albums. How to efficiently access such a large number of images has become a problem. As a solution to this problem, the content-Based Image Retrieval (CBIR) aims to enable the computer to automatically find the images that are similar to auser's query from an image database. The retrieval is based on the visual content of images, for instance, color, texture or shape. It has many advantages over the traditional text-based retrieval methods. In recent years, the local invariant feature has proven to be a powerful tool to representing images and measuring the similarity between two images. This technique can be readily applied to Content-Based Image Retrieval. Goals of this projectThis project addresses the Content-based Image Retrieval by using the local invariant features. Also, given some online feedback from the user (known as Relevance Feedback mechanism), how to manipulate these local invariant features to refine the retrieval result will be explored. This project involves extracting local invariant features and implementing the retrieval algorithms. The final retrieval algorithm will be incorporated into an CBIR demonstration system to display its effectiveness and efficiency.Requirements/PrerequisitesThe student will have experience on C/C++ and Matlab programming. Some knowledge on computer vision and pattern recognition will be a plus. The student will work closely with the supervisor on a day-to-day basis. Also, this project provides opportunities to interact with the excellent researchers and students in RSISE and NICTA.Background Literature
Contact:E: lei.wang@anu.edu.auW: http://users.rsise.anu.edu.au/~wanglei/ Patch-based object recognition with discriminative visual codebooksProject Code:R&CV_05Supervisor:Dr Lei WangOutline:Nowadays, computers have been able to easily "see" this world through digital imaging equipments. However, enabling a computer to "understand" what it has seen still remains a difficult problem. Playing a critical role in solving this problem, object recognition aims to recognize the object (for instance, a motorcycle or an airplane) in an image through its appearance. Recent years have witnessed the significant progress of patch-based object recognition. In this approach, a number of small patches are cropped from an image, each of which is characterized by a local invariant feature. Thus, each image is represented by a bag of features. This process can be intuitively understood by viewing an image as a "document" and the small patches as "words". For recognition, a visual codebook (a "dictionary" of visual words) is often needed to be created. With the codebook, each image is represented by the number of occurrence of each visual word in this image. The design of visual codebooks has significant impact on the recognition performance.Goals of this projectThis project involves image patch sampling, local invariant feature extraction and particularly the visual codebook design. Since the ultimate goal of object recognition is to predict the type of the object in a given image as accurately as possible, a discriminative visual codebook will be desired. A good web resource and some recently published papers on patch-based object recognition are listed in the reference part below.Requirements/PrerequisitesThe student will have experience on C/C++ or Matlab programming. Some knowledge on computer vision and pattern recognition will be a plus. The student will work closely with the supervisors and the post-graduates on a day-to-day basis. Also, this project provides opportunities to interact with the excellent researchers and students in RSISE and NICTA.Background Literature
Contact:E: lei.wang@anu.edu.auW: http://users.rsise.anu.edu.au/~wanglei/ Where am I? ---- Visual Localization on the ANU CampusProject Code:R&CV_06Supervisor:Dr Lei WangOutline:"Where am I" is a question we seldom ask others because usually our eyes can give the answer. However, imagine that it is the first time for you to visit the campus of ANU. Can you find out where you are by simply looking at the buildings that you have never seen before?Recent development in computer vision technology has equipped modern computers with the capability to "see" and recognize objects. One interesting application of this capability is a visual localization system: the user takes a photo of the environment in which he/she is (for example, with a digital camera or simply a mobile phone), sends the photo to the system and the system will tell the user where you are. To implement such a system, we need a database to store the photos of the buildings on campus. Also, a algorithm is needed to extract visual information from each photo. Then, given a query image, the system will compare the visual information extracted from the query image with those stored in the database to identify which part of the campus has been captured in the query image. The question of "where am I" is then answered based on this result. Goals of this projectThis project is composed of several interesting steps such as building the database, visual feature extraction, defining the similarity measure to evaluate whether a buldiing in a query image is same to those in the database, and designing an indexing mechanism supporting efficient search in the database. Plenty of resource for each step can be found from the literature. The student may focus on one of them or build up the whole system.Requirements/PrerequisitesThe student will have experience on C/C++ or Matlab programming. Some knowledge on computer vision and pattern recognition will be a plus. The student will work closely with the supervisors and the post-graduates on a day-to-day basis. Also, this project provides opportunities to interact with the excellent researchers and students in RSISE and NICTA.Background Literature
Contact:E: lei.wang@anu.edu.auW: http://users.rsise.anu.edu.au/~wanglei/ Optimising computer vision applications for resource constrained devices (3 student projects)Project Code:R&CV_10Supervisor:Dr Lars PeterssonOutline:Project #1: Design and implement a capture and processing scheduler on the Asus R50A for optimizing application resource utilisationProject #2: Design and implement a synchronisation manager on the Asus R50A for optimising wireless communication network resource utilisation Project #3: Design and implement a coordination systems for coordinating the processing activities of multiple Asus R50A platforms This series of projects deals with a real-world computer vision application to be used in vehicles on the road. The challenge is that the computer vision algorithm to be used is computationally very resource intensive. When the target deployment platform is resource constrained, for example a handheld device, intelligent decisions need to be made about how to schedule, forward plan and parameterise the computer vision algorithms in order to maximise the useful work performed by the application given the available resources. The goal of these projects is build systems which enable and carry out these decisions. Development is being broken down into three smaller, but interrelated projects (outlined above). Each is challenging but manageable within the summer scholar program. Goals of this projectThe target platform for our computer vision application is resource constrained. The platform is the Asus R50A which is classed as an Ultra Mobile PC. The specifications and picture of the device are available below.http://www.asus.com.au/products.aspx?l1=5&l2=25&l3=722 http://www.asus.com.au/999/images/products/2295/R50A_1.jpg The device is a suitable target platform for a computer vision application given its 2MP (1024 x 768) camera, 3G wireless connectivity and reasonable computing power. In addition it also has an in-built GPS to geo-locate captured video data. The device has a form factor suitable for being deployed on the dashboard of a vehicle and comes with appropriate mounting accessories. The device has been modified to run a standard Linux variant (Moblin) and all applicable device drivers have been configured. Students will have access to two or three of these devices for prototyping and testing their implementations. The goal is to design and implement a software architecture within this resource constrained hardware environment which manages the computer vision application. It is part of the task of the summer scholars to define how the overall system architecture will work, how the individual components are best organised and how the three project elements will work together to schedule, forward plan and parameterise the computer vision algorithms to maximise the useful work performed. Update: Check out the recent ComputerWorld media coverage of the NICTA project related to this summer scholar project: http://www.computerworld.com.au/article/313968/new_linux-based_technology_make_smarter_gps Requirements/PrerequisitesStudent Gain
Contact:E: lars.petersson@nicta.com.auW: http://nicta.com.au/director/research/programs/asst/people/lars_petersson.cfm Resource scavenging for a distributed Computer Vision applicationProject Code:R&CV_11Supervisor:Dr Lars PeterssonOutline:Project: Build a CPU cycle scavenging client/server system which contributes unused CPU cycles from donor machines to a central Linux clusterComputer vision algorithms are often computationally intensive and in our application the most computationally intensive part is the scanning of video for the purpose of detecting and identifying objects of interest. Fortunately, our scanning problem breaks up well into sub problems and can be executed in parallel and distributed over many computers. A common approach to cope with the scale of computing resources required in computer vision is to build dedicated compute clusters of x86 machines and distribute the problem over many dedicated CPU's, however this is not necessarily the most cost effective method. In a standard office environment, there are many high powered computers and servers which sit idle most of the time, even whilst users are using them. Few desktop, laptop or server computers are ever used close to their computing capacity. We would like to see these computers at 100% utilization by configuring the computers to contribute unused CPU cycles to a compute cluster whenever they are powered on. These office computers currently run a variety of operating systems, including Linux and Windows XP and are interconnected via a gigabit Ethernet network The project is concerned with designing, building and porting to a variety of operating systems an application which can contribute the spare CPU cycles of underutilised machines to a computing cluster running computer vision applications. The development and deployment of a distributed processing application means computer hardware is available both for primary business tasks like word processing and at the same time contributing spare CPU cycles as a compute node in a computer vision cluster. Having this software system means the we can realise a greater return on existing investments in hardware resources, reduce or eliminate the need to purchase additional computing resources as well as reducing energy overall consumption. As CPU cycles are typically being donated voluntarily by users, it is an important requirement that there is little or no inconvenience to the donor and that they have at least some degree of control over when and how their resources are utilised. Ensuring minimal impact on the donor will be an important success factor for the successful adoption and deployment of the application. Update: Check out the recent ComputerWorld media coverage of the NICTA project related to this summer scholar project: http://www.computerworld.com.au/article/313968/new_linux-based_technology_make_smarter_gps Requirements/PrerequisitesStudent GainYou will benefit from these projects as follows:
Contact:E: lars.petersson@nicta.com.auW: http://nicta.com.au/director/research/programs/asst/people/lars_petersson.cfm Useful GamesProject Code:R&CV_12Supervisor:Dr Lars PeterssonOutline:Games are becoming closer and closer to real life and some tasks we have to solve in games are similar to tasks we have to solve in real life. One surprising difference is that for many people solving the tasks in games is much more fun and less tiring than in real life. Consequently, it can be useful to transform tasks or work we do not like into a game and do our work by playing a game--or even better, let other players do the work for us.In this project we aim to develop a useful game for quality assurance of computer vision software. Our existing computer vision software is able to detect objects we see while driving a car. In order to test whether our software correctly identifies all objects, we have to manually browse through large amounts of video footage in several views. As part of this project we analyse how this task can be solved by a useful and fun game. Contact:E: lars.petersson@nicta.com.auW: http://nicta.com.au/director/research/programs/asst/people/lars_petersson.cfm
Monitoring Java programs in the real worldProject Code:SISE_01Supervisor:Dr Andreas BauerOutline:The goal of this project is to investigate means of monitoring the behaviour of Java programs while they execute. Monitoring is useful in many different scenarios, e.g., for testing systems or in a security critical context to avoid systems entering a bad state when they execute.On the site [1] an open-source tool can be downloaded which, given a formal specification of a monitorable property (which intuitively may translate to "it is always the case that a request signal is followed by an acknowledgement signal"), generates a so-called monitor for that property (more precisely, a state machine which acts as a monitor). The monitor can then be used to monitor the system for that particular property, that was prior entered by the user. However, there does not currently exist an automatic translation from the generated monitor to actual Java code, which is needed for making this method work in the real world. Thus, in this project, the student should investigate means for instrumenting Java programs such that they can be combined with the generated monitors. A new technology from SUN Microsystems in the form of [2] is available for that purpose, and will aid as a starting point for this investigation. Other promising approaches to integrate monitors with Java code may be based on aspect-oriented technologies, where the monitors are, but aspects of a program. On the other hand, [2] bears the advantage that not even access to the Java program's source code is required for monitoring it. The Pros and Cons of either solution, as well as other available options, should be determined in this project, and one particularly suitable solution demonstrated in terms of building a small, but fully functional proto-type. Background LiteratureW: LTL3 toolsW: BTrace Contact:E: andreas.bauer@nicta.com.auW: http://rsise.anu.edu.au/~baueran
Solar Thermal GroupProject Code:ST_01Supervisor:Dr John PyeDr Keith Lovegrove Outline:The Solar Thermal Group at ANU has recently completed construction of a new 500 m² dish concentrator, designed for mass-production, in cooperation with our commercial partner, Wizard Power, a Canberra-based business. There are a huge range of possible areas where a summer student interested in sustainable energy could contribute at this important stage, we'd encourage any interested student to get in contact with us and we can discuss a particular project plan. Both experimental and theoretical projects are possible.Contact:E: john.pye@anu.edu.auE: keith.lovegrove@anu.edu.au Gas turbine dynamic simulation using ASCENDProject Code:ST_02Supervisor:Dr John PyeOutline:The open-source mathematical modelling program, ASCEND, provides a framework for solving a wide range of engineering equations. This project would involve using ASCEND to develop a dynamic model for the small gas turbine currently proposed for installation at the focus of the new ANU SG4 dish concentrator. We would seek to calculate the annual output achievable from this dish/engine, using as sophisticated a model as time permits, and to perform a range of design optimisations based on the system outputs. The model system is likely to incorporate an 'economiser' heat exchanger.The student would gain understanding of a detailed multi-physics engineering problem with high current relevance and impact, and develop useful skills in process modelling and design. Contact:E: john.pye@anu.edu.au
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Please direct all enquiries for this website to:
JavaScript must be enabled to display this email address.
Page authorised by: Head of Department, InfoEng |
|
The Australian National University CRICOS Provider Number 00120C — ABN: 522 34063906 |
