Software Engineering research project option: COMP4720,
COMP4730 and COMP4540
Fourth year software engineering students can apply to do an
individual research-oriented project instead of the group project. You
need to get permission from the software engineering research project
convenor
[Chris Johnson in 2005];
normally, this form of project is restricted
to
students with a Distinction average.
Doing a research project enables you to demonstrate your own
inventive abilities and to work one to one with a supervisor on an
topic which is
usually related with his or her own current research. You will learn
how to undertake a research project, usually create some working
software which contributes to the research area, and report the results
by writing a 60-80 page thesis and a short software engineering process
report. (You can see some archive examples of previous theses--in 2003
and 2004 the thesis was included in the software engineering report).
This kind of project and report constitutes a good supporting evidence
of your research abilities and experience if you want to apply to do
postgraduate research, towards a PhD. But first you have to find a
project topic.
Topics are usually proposed by academics who are willing to
supervise your work. You don't have to be constrained by what you see
on the list below.
If you have something that you would like to work on as a project,
then feel free to discus it with the honours convener to see if it
could form the basis of an honours project and to identify a possible
supervisor. Some of the computer science honours projects can also be
adapted as research projects for software engineering - see link below.
The structure of BSEng research project option
Students can choose to
take a 15 unit or 18 unit research project. The choice of sizes allows
students to fit their study program, but only if the size of the
project topic allows this.
The 15 or 18
unit project is taken over 2 semesters by enrolling in COMP4720 (3
units) or COMP4730 (6 units) plus COMP4540 (12 units).
The project is taken by enrolling in course COMP4540 (usually in
semester 2), preceded by a smaller preparatory course with the same
supervisor (usually in semester 1).
The results for both
courses will be reported at the end of the main project comp4540. The
grade will be the same for both courses.
In principle, each research project will require
- report of literature review
- seminar presentation
- thesis report (40 pages-80 pages)
- project software engineering report (10 pages)
The delivery dates and format of these items will be specified
later and may vary to suit the project, after getting agreement from
the convenor
and supervisor.
You can see previous
years' theses as examples of the scope of work and reporting.
Please note that some of the projects below are very specific
in
nature. However, you may be able to discus minor changes to the
proposal with the supervisor. Other proposals are, in fact, too broad
in scope to describe an individual project. You should take
these descriptions as an indication of the areas of interest of the
supervisor. If you are interested in the area then you can discuss
potential specific projects with the supervisor.
In no particular order, here are the topics proposed for 2005:
THIS IS A PRELIMINARY LIST - version 6 14 Feb 2005
Other projects: The CS honours and eScience project pages show
projects and responsible
academics for other degree programs. Under some circumstances, it may
be
possible for BSEng students to participate or take a project similar to
these. You can propose one of these as your software
engineering research project, with necessary modifications imposed by
the SE research projects convenor, Chris Johnson; or you can propose
your own
project with the agreement of the convenor and an approved supervisor.
Other topics are listed for
From timetable to lecure
reminders: integrating a scheduler with the ANU timetable system
Supervisor: Chris
Johnson and/or Ramesh
Sankaranarayana
E-Mail:
Chris.Johnson@anu.edu.au
Background
There are many proprietory and open source calendars or personal diary
systems. Some of these provide group membership abilities so that a
formal class group, an informal study group, a club or society can post
events into the diaries of anyone who is authorised or interested. The
ANU administration provides personalised course timetable information,
on request, as a web page. It would enhance the value of the ANUY
timetable system if it could be fed into a general diary and reminder
system.
Software Engineering student research project
First: some analysis of potential users and their current and
near-future preferred means of accessing a calendar (for example,
through WWW onto desktop, laboratory computer, personal mobile laptop,
personal mobile phone or digital assistant; through mobile phone access
to HTML); second, select an appropriate open source calendar system
that can be extended to suit these needs, and can take information from
the ANU timetable.
A student working on this project would be involved through
More information
Location specific emergency
situation reporting LOCSESITREP
to be defined
Upscaled Fire Mapping System
Supervisor
Tom Worthington
Background
The Sentinel
Fire Mapping System provides a web based map of possible fires
across Australia. However, the system generates a new custom map for
each user and so it is performance limited to a few thousand
simultaneous users. Also the system is not compatible with mobile
devices used in the field by emergency personnel.
Software Engineering student research project
Design and build an accessible, high volume, mobile
compatible, front end for the Sentinel Fire Mapping Geographic
Information System.
Produce a new front end which:
- Caches the mapping data to provide for high volumes of use,
so that the system can be expanded to monitor the entire surface of the
earth and have millions of simultaneous users.
- provides an interface compatible with a range of devices
including mobile phones, PDAs and suitable for automatic translation
into other languages.
More information
0. public access to Sentinel service
http://www.sentinel.csiro.au/mapping/viewer.htm
1. "Dealing with Disaster – Using new Networking Technology
for Emergency Coordination", Tom Worthington, October 2003 <http://www.tomw.net.au/2003/enet.html>
2. "Dealing with Disaster - Using new Networking Technology for
Emergency Coordination - A personal view", Tom Worthington, May 2004 <http://www.tomw.net.au/2004/enetp.html>.
Global Emergency Warning System
Supervisor
Tom Worthington
Background
The recent fires in Australia and
tsunami in Asia have shown the need for rapid advice to the public on
disasters. While detection of tsunamis and bushfires are possible, the
problem of how to alert officials and the public is unsolved. Also
there is the problem of how to educate officials and the public in
detailed procedures ahead of the event, during and after.
Software Engineering student research
project
Design and build a prototype of a web and mobile phone system
providing emergency information across the world. One option is to use
the cell broadcast function of mobile telephones to provide fast (20
second) localized alerts in a geographic area, with the world wide web
to provide detailed information.
More information
- RE: Emergency Web Site Design, Tom Worthington, Dec 2004 <http://mailman.anu.edu.au/pipermail/link/2004-December/059819.html>
- The Cellular Emergency Alert Systems Association, 2004 <http://www.ceasa-int.org/>.
Emergency Situation Room in an
Access Grid Box
Supervisor
Tom Worthington
Background
Emergency services use expensive purpose-designed situation
rooms with specialized communications hardware for coordinating
emergency operations. There are not sufficient of these faculties to
handle all possible emergencies and those there are may be unavailable
in an emergency. Access grid nodes are high bandwidth, multiple data
projector communications endpoints designed for computatiopnal science
in particular. They provide similar facilities to the emergency rooms,
and could be pressed into service for this function.
Software Engineering student
research project
Adapt available collaboration tools for the access grid to
build an open source facility for emergency service personnel. Write
any needed additional software, a manual on how to adapt such a
facility ("EMERGENCY SERVICE CENTRE IN A BOX") and demonstrate it using
the ANU's Access Grid #1 (Room N101).
More information
- "Dealing with Disaster – Using new Networking Technology
for Emergency Coordination", Tom Worthington, October 2003 <http://www.tomw.net.au/2003/enet.html#The%20Grid:%20Future%20for%20Emergency%20Use_|outline>
- "Dealing with Disaster - Using new Networking Technology
for Emergency Coordination - A personal view", Tom Worthington, May
2004 <http://www.tomw.net.au/2004/enetp.html#grid>.
Optimize Government Archiving
Supervisor
Tom Worthington
Background
National Archives of Australia have produced a tool (XENA) to
convert documents in popular office formats to a standardized XML
format for long term archiving. As these documents will be stored for
hundreds of years, ensuring a compact and easy to read format is
important. The tool uses OpenOffice.Org as a conversion engine.
However, this produces verbose XML code, which is not compatible with
any other software.
Software Engineering student
research project
Critically compare similar long-term archiving projects and determine
the requirements for long term storage formats. Modify the output of
OpenOffice to produce a compatible, more standard, compact documents
for long term storage, creating a tool to convert OpenOffice word
processor documents. Create an XSLT transformation to turn these
documents into documents which can be displayed by a web browser
(possibly using XHTML Basic and XSLT).
More information
- Open Source at Digital Preservation Meeting in Canberra,
Tom Worthington, Link list, Sep 2004 <http://mailman.anu.edu.au/pipermail/link/2004-September/058557.html>
Super Tidy
Supervisor
Tom Worthington
Background
The Tidy program removes styling commands scattered through the code of
a web page and places them all in a style sheet at the top of the page.
However, Tidy only does this for one web page at a time, and the style
sheets generated are often poorly styled: they contain formatting
commands that are repeated within each page and between pages.
Software
Engineering student research project
Produce a tool to rationalize the use of formatting for a web site.
Produce a tool which will take the output of Tidy, collect all of the
styling commands for all the web pages of a web site and rationalize
them in the minimum number of style sheets. The tool must be
generalised to allow later contributions of additional optimisations.
Some of the technology of compiler code optimisations will need to be
considered and applied. Provide additional output to help people to
evaluate some of the quality of the website. Test the tool using the
ANU web site.
More information
- Clean up your Web pages with HTML TIDY, <http://www.w3.org/People/Raggett/tidy/>
- Website Design <http://www.tomw.net.au/2004/wd/index.html#udf>.
Topic Maps for Court Evidence
Supervisor
Tom Worthington
Background
Court evidence is presented and recorded mostly as text
documents. This evidence can run to tens of thousands of pages. This
makes summarization and comparison of the evidence a complex process.
XML topic maps provide a an abstract way to represent data, which can
then be presented texturally or visually.
Software
Engineering student research project
Investigate available topic mapping systems. Build a tool for
representing evidence to a court in the form of an XML topic map, with
text data entry and graphical visualisation of the data. Demonstrate
and validate using evidence from the ACT Coroner’s Bushfire Inquiry.
More information
- IT at the ACT Coroner’s Bushfire Inquiry, Tom Worthington,
Link list, Oct 2004, <http://mailman.anu.edu.au/pipermail/link/2004-October/058986.html>
Caveats (cwj)
is the evidence available in
a form that can be used
to demo without too much handwork?
Communicating Energy Use
Supervisor
Tom Worthington
Background
Modern buildings have dedicated computer based system
controlling and monitoring the operation of their energy consumption.
If this information could be provided to building tenants they could
change their behaviour, saving money and energy. Typically energy use
is relayed to the tenants by dedicated single board computers and LCD
displays. These devices are cheap to buy, but expensive to program to
provide useful information.
Software
Engineering student research project
Build an adaptable internet/web based front end to a building
building monitoring system (BMS). This alternative system will stream
the data from the BMS into an on-line database and display the data to
the building tenants using flexible web based visualization tools.
Implement the system at the Brindabella Business Park.
More information
- Brindabella Business Park http://www.canberraairport.com.au/5greenstar.htm
The Active
National University: applying
location awareness
Supervisor
Chris Johnson and possibly Tom Gedeon
Background
The proliferation of wireless access points and mobile access
devices across university campuses provides an opportunity to use
information on people's locations, provided by their wireless computing
devices, to support their working and recreational life on campus.
The Active Campus project
at University of California San Diego is an example of a prototype
system. In any such system there are technical issues about the
effectiveness of the
wireless network as a location sensor and information carrier,
issues of who provides and administers access, identification of
devices and people, the design of applications and their interfaces on
mobile devices, privacy and ethics in tracking
and informing on people, and the effectiveness of systems as people use
them and adapt to their use.
Wireless location on this scale has low precision and is highly
variable.
The ANU is deploying a network of 802.11b (WiFi) wireless access
points; it's time to turn the campus into the Active National University.
Software
Engineering student research project
Several projects in this area will potentially support each other:
- Support framework for
applications: Analyse, design and implement a robust,
extendable framework that demonstrates support for a small suite of
basic applications suited to the ANU campus. Mobile computing platforms
can be those anticipated in the near future, and the applications can
be selected or invented to support education, research, and social
activities. The system must incorporate a strong flexible emphasis on
privacy and security.
- Location Authority and
Service: Various structures and requiremetns for Location
Authorities for location based services have been suggested, in
particular by Shafer. A robust, long-life, easily maintained Location
Authority is essentially a database that maps the device identity of
wireless access points onto their geographic locations on campus, but
it also provides procedures and utilities for creating and maintaining
the actual data as access points are installed, moved, go down and
return to service. The LA should interoperate with Geographic
Information Systems such as those giving the coordinate systems
provided by available maps of ANU campus as a demonstration
example.Design and implement a robust, long-life, easily
maintained Location Authority. This is essentially a database that maps
the device identity of wireless access points onto their geographic
locations on campus, but it also provides procedures and utilities for
creating and maintaining the actual data as access points are
installed, moved, go down and return to service. The LA must
interoperate with coordinate systems provided by available maps of
campus.
- Novel applications:
Design, implement and evaluate the effectiveness and social issues of
novel location-aware applications on the Active National University
framework system. Human and social factors are as important here as
communications and information services.
More information
- Everyday Encounters with
Context-Aware Computing in a Campus Environment
Louise Barkhuus and Paul Dourish
in N. Davies
et al .(eds) UbiComp 2004,
LNCS 3205, pp 232-249, 2004
- Challenges in
Location-Aware Computing, C.A. Patterson, R.R. Muntz, and C.M.
Pancake, IEEE Pervasive Computing,
vol 2 no.
2, 2003, pp. 80-89
- Guiding, Tracking and perceptual computing
- Daily
Wireless Ubiquitous Computing
- ActiveCampus: Experiments in Community-Oriented
Ubiquitous Computing, William G. Griswold, Patricia Shanahan,
Steven W. Brown, Robert Boyer, Matt Ratto, R. Benjamin Shapiro, Tan
Minh Truong, IEEE Computer,
October 2004 pp. 73-81
- The Carrot Approach:
Encouraging Use of Location Systems, by Kieran Mansley, Alastair
R. Beresford, David Scott.
in N. Davies et al
.(eds) UbiComp 2004, LNCS
3205, pp 366-383, 2004.
- Location Authorities for
Ubiquitous
Computing, Steven A. N. Shafer,
in
Mike Hazas, James Scott, and
John Krumm (eds), UbiComp workshop
on Location-Aware Computing, 2003.
at http://www.ubicomp.org/ubicomp2003/workshops/locationaware/
- Intimate Computing?,
G. Bell, IEEE Internet Computing,
vol 8 no 6, Nov-Dec 2004, pp.91-93
Responsive Home management panel
Supervisor
Chris Johnson
Background
As sensors and actuators become cheaper and more people become
familiar with using digital technology, many Smart Home projects, kits
and equipment have become available- and obsolete. Current technology
for home automation is for geeks. Control panels, computer controls. It
has no personalisation or learning. It wouldn’t get my 80+ parent
anywhere near managing the technology, nor more than a tiny number of
my contemporaries. Many of the problems associated with smart home
technology are social (concerns about personal privacy and trust in the
system's maintaining that privacy) as much as they are technical. The
use of cheap video and still cameras and microphones in an office or
home can be used to capture memorable events and information, replacing
the video and still camera, and can be used to monitor and support aged
and infirm people living alone (detecting and reporting to a health
carer that that the occupant has fallen over, for example).
Assuming that the devices are installed, users will only come
to trust that the camera is not Big Brother if they are given
intuitive, easily used, reliable controls over what devices are active
and where their data flows to, within and outside the home.
Software
Engineering student research project
Analyse the literature, design and implement prototype touch
screen or cursor-managed control panels for assumed home installations
of devices and dataflows projected into the near future, providing very
high quality usability and engendering high degree of trust by older
users in particular. A graphical programming metaphor is suggested, but
simplified language vocbularies should also be investigated.
More information
- numerous
home
automation products include visual configuration control e.g. Premise
Home
Control Software (www.smarthome.com/manuals/1410_brochure.pdf)
– many
touchscreen home network device controllers
(www.smarthome.com/lcdpanel.html).
- Humble et al, in UbiComp2003, Playing with the
bits: user configuration of ubiquitous domestic environments
- Truong, in UbiComp2004, CAMP: a magnetic
poetry interface for end-user programming of capture applications for
the home
Ubiquitous Computing toolkit
supervisor: Chris Johnson
David Parsons (Massey University, Albany campus in Auckland,
NZ) suggests that ubiquitous
computing needs a framework for software development, like Java
Enterprise Struts and Tiles. The framework would be formulated as a
complentary kit of patterns and implementation classes, starting with
patterns such as
-
agent/filter
-
publish/subscribe
-
transactional
-
messaging
-
weak connectivity
An alternative approach to a toolkit for ubiquitous computing is the
Context Widget Toolkit by Anand Dey, originally of Georgia Tech.
The phenomena of distributed persistent
objects, location awareness, persistence, data synchronisation at a
range of strength (weak and strong coupling) also need to be considered.
Software
Engineering student research project
Investigate ubiquitous computing architectures from software
reports and
projects; determine what is meant by an enterprise framework in this
context; critically compare various approaches; design a ubiquitous
computing support framework, and describe and implement a subset of the
classes in the kit, demonstrating them in lightweight applications...
and then, after lunch... :-) A subset of these objectives will be
enough.
More information
- David Parsons, Java Architectures for Mobilised Enterprise
Systems, in HICSS38 Jan 2005,
http://csdl.computer.org/comp/proceedings/hicss/2005/2268/09/22680298c.pdf
- Daniel Salber and Anind K. Dey and Gregory D. Abowd, The
Context Toolkit: Aiding the Development of Context-Enabled
Applications, in Proceedings
of CHI 99, and as technical report of Graphics, Visualization and
Usability Center, Georgia Institute of Technology GIT-GVU-98-33, http://www.cc.gatech.edu/gvu/reports/1998/abstracts/98-33.html
- Seng W. Loke, Evi Syukur, Peter Stanski, Adding
Context-Aware Behaviour to Almost Anything: the Case of Context-Aware
Device Ecologies in 2nd
International Conference on Mobile Systems, Applications and Services
MobiSys 2004 workshop on context awareness http://www.sigmobile.org/mobisys/2004/context_awareness/papers/mobisys-ca.pdf
- Tao Gu, Hung Keng Pung, and Da Qing Zhang, Towards an
OSGi-Based Infrastructure for Context-Aware Applications, IEEE Pervasive Computing, vol 3,
no.
4, Oct-Dec 2004, pp. 66-74
- K. Hendricksen, J, Indulska, and A. Rakotonirainy,
Infrastructure for Pervasive Computing: Challenges, Proc. Informatik 01: Workshop on Pervasive
Computing, Univ. Vienna, 2001, pp. 214-222 http://www.dstc.edu.au/m3/papers/Informatik2001.pdf
Compressed file system for Mac OS X
Supervisor
TBA (Eric McCreath?) proposed by Tim Wilson-Brown
Background
Some operating systems such as Windows XP have optional
compressed file systems, but Apple Mac does not. Analyse available and
unimplemented designs for compressed file systems, consider their value
against the steadily falling cost of uncompressed storage.
Software
Engineering student research project
Design and implement a compressed file system for the Apple
Mac OS X operating system.
Neural networks
theory and applications
Supervisor
Tom Gedeon
tom@cs.anu.edu.au
Background
A number of
projects are possible in this area. I am interested in a number of
topics, such as
- extracting rules from neural networks
- information
retrieval using neural networks
- data mining and feature selection
- cascade neural network structures
- hierarchical neural network
structures
- neural network applications.
I have published papers
in all of these areas with former students so there is plenty of
earlier work to build on. No previous experience with neural networks
is necessary. Most projects will use the very popular backpropagation
neural network training algorithm. If you are interested in neural
networks and would like to see if you might want to do a project in
this area, please e-mail tom@cs.anu.edu.au
or come and see me.
Fuzzy Logic theory and applications
Supervisor
Tom Gedeon
tom@cs.anu.edu.au
Background
A number of
projects are possible in this area. I am interested in a number of
topics, such as
automated construction of fuzzy rule bases from data
- hierarchical fuzzy systems
- fuzzy interpolation
- information
retrieval using fuzzy logic
- universal approximators
- fuzzy
logic applications.
I have published papers in all of
these areas
with former students so there is plenty of earlier work to build on.
No previous experience with fuzzy logic is necessary. If you are
interested in fuzzy logic and would like to see if you might want to
do a project in this area, please e-mail me tom@cs.anu.edu.au
or come and see me.
Note - this topic may be
more suited to Computer Science but Software Engineers are not
excluded.
Face Recognition
Supervisor
Tom Gedeon
tom@cs.anu.edu.au
Background
A number of
projects are possible in this area, such as
- image processing for face recognition (some experience with
computer graphics would
be useful)
- HCI research
tool to collect facial images
- biologically plausible architectures
for face recognition
- building and modifying computer models of real
faces.
No previous work on face
recognition is necessary. If you are interested in face recognition or
some similar
topic and would like to see if you might want to do a project in this
area, please e-mail me tom@cs.anu.edu.au
or come and see me.
Modeling the Human Brain
Supervisor
L. Andrew Coward/Tom Gedeon
Email andrew.coward@anu.edu.au
tom.gedeon@anu.edu.au
Background
Understanding of how neurons are organized to support higher
cognition in human beings is an important research area. This project
would be part of a research program aimed at modeling memory and
attention in terms of the behaviour of neurons within a structured
neural architecture.
student research project
The cortex column (www.cs.toronto.edu/~miguel/papers/ps/ecs02.pdf)
is a prominent feature of the mammal (including human) brain. This
project would create a computer model of a cortex column within an
architectural framework called the recommendation architecture (http://cs.anu.edu.au/~Andrew.Coward/scotland.pdf).
The model would include a model for a neuron with inputs resembling
axon action potentials which could learn cognitively useful responses
to combinations of those inputs under the guidance of other neurons in
the column.
A student working on this project would be involved through
- Participation in a research project on modeling human
cognitive behaviour in terms of neurons
- Object oriented crossplatform programming (using the
Smalltalk language).
More information
Please visit the web page
http://cs.anu.edu.au/~Andrew.Coward
Cluster based transcoding for
bandwidth scaling and video pre-processing
Supervisor
Rhys Hawkins and Markus
Buchhorn
This project is an extension of a Software Engineering research project
from last year by Justin Bedo called Video
Preprocessing Cluster. In this project, an MPI based system was
created which could rescale H.261 video streams in real-time with low
latency. The next obvious step is to extend this to cope with higher
bandwidth video streams. The proposal is to extend this work to be able
to process DV (digital video) from most consumer digital cam-corders
and HDTV (mpeg2).
More information
- Justin Bedo's thesis SER Thesis 2004
parallel programming tool MPI http://www-unix.mcs.anl.gov/mpi/
DV http://www.adamwilt.com/DV-tech.html
HDTV http://www.standards.com.au/catalogue/script/Details.asp?DocN=AS998467444104
Probabilistic Temporal Planning
Supervisors
Doug Aberdeen, Olivier Buffet, Sylvie Thiébaux [NICTA/RSISE]
Planning is the process of finding which sequence of decisions
to take so as to achieve a given goal. We are currently working on a
large planning domain aiming at scheduling tasks with probabilistic
outcomes [1]. To our knowledge, the tool which has been developed is
the first one handling time, resources and probabilities at the same
time.
Various questions can be addressed in this framework.
Improving the planning algorithms is a major issue, may that be in the
framework of Markov Decision Processes or more classical planning.
Other issues as plan visualization or human-machine interaction are of
interest, as the plan should be easy to understand and the user may
need to give feedback to the software (by setting new constraints for
planning).
There are several sub-projects available. We highly encourage
interested students to come and talk to us to discuss specific
possibilities. Software used in this project may end up being use
by the Defence Science Technology Organisation.
More information
- D. Aberdeen and S. Thiébaux and L. Zhang,
Decision-Theoretic Military Operations Planning, in Proceedings of the Fourteenth
International Conference on Automated Planning and Scheduling
(ICAPS'04), Whistler, Canada, June 2004.
Sport Signal Analysis
Supervisors
Olivier Buffet, Doug Aberdeen [NICTA/RSISE]
Kinetic Performance Technology is developing an on-line
performance analysis service (GymAware) which takes data collected by
sensors in a strength training gym and produces detailed reports to
assist coaches in the preparation and training of athletes. The system
is currently under development and relies on a robust analysis of the
signal: identifying phases in the athlete's moves (i.e. segmenting the
signal) should be made with great accuracy.
The first system developped for this segmentation was a
rule-based approach. The main aim of this project is to replace this by
an algorithm which should learn how to make this segmentation through
examples. This should be achievable through graphical models such as
Hidden Markov Models [1] or Conditional Random Fields [2]. The project
aims at determining more precisely which machine learning tools can be
used, making a theoretical and practical study in the framework of
GymAware.
More information
- L.R. Rabiner, "A Tutorial on Hidden Markov Models and
Selected Applications in Speech Recognition", IEEE ASSP Magazine, pp
257--286, February 1989, 77 (2).
- J. Lafferty, A. McCallum, and F. Pereira. Conditional
random fields: probabilistic modeling for segmenting and labeling
sequence data. In 18th International Conference on Machine Learning
ICML, 2001.
Automatic Hierarchy Discovery
Supervisor
Olivier Buffet [NICTA/RSISE]
Reinforcement Learning consists in learning a policy (how to
act in various situations) in order to maximise a payoff (which
describes a goal to achieve). Considering probabilistic environments,
we work in the framework of Markov Decision Processes (MDP) [1], where
an action may lead to different outcomes. As state spaces may be very
large, it is of major interest to consider a given problem as a
hierarchy of problems ("divide and conquer" approach). Unfortunately,
if various works have led to efficient algorithms exploiting such a
hierarchical structure, automatically discovering a hierarchy remains a
very difficult task.
A recent work [2] has proposed a first approach to automatic
hierarchy discovery, based on analysing which observed variables seem
to be the most dependent on the decisions made. Yet, this algorithm
relies on an appropriate choice of variables. The main concern of this
project is therefore to find how to get rid of this choice of
variables, the system being able to create "artificial variables" if
they seem more appropriate than the original ones.
More information
- L. Kaelbling, M. Littman and A. Moore, "Reinforcement
Learning: A Survey", Journal of Artificial Intelligence Research, 4, pp
237--285, 1996.
- B. Hengst, "Discovering Hierarchy in Reinforcement Learning
with HEXQ", in Proceedings of the Nineteenth International Conference
on Machine Learning (ICML'02), pp 243--250, 2002
Text-based adventure game for
teaching Digital Communications
Supervisor
Haley Jones [Dept of Engineering]
Background
Before the days of fabulous graphics, there were many computer
adventure games based on either just text or text and very simple
graphics. Recently, Dr Paul Francis of The Department of Physics
and RSAA (Mt Stromlo) at the ANU has developed a text-based
adventure game to assist in teaching astrophysics to first year
students: http://www.mso.anu.edu.au/~pfrancis/Stromlogame/
This game was developed using TADS (Text Adventure Development System). http://www.tads.org/
TADS is a freeware programming system that can help you create
high-quality interactive fiction. It includes an object-oriented
programming language with many built-in objects and classes,
appropriate for adventure games. In particular, many of the string and
language parsing functions most commonly used by such games are a part
of the standard TADS. There is also, of course, the flexibility to add
your own objects and classes.
This project would involve a student undertaking to develop
such a game, using TADS, to help students learning about
telecommunications systems. There is plenty of scope. The student would
need to, in collaboration with me, come up with a set of
specifications, as well as a storyline (which could be based on a
well-known story, e.g., Harry Potter). They would need to determine
what types of things are and are not feasible to teach in such an
environment. It is hoped that the program would eventually be used in
one of our 3rd year telecommunications engineering courses.
It is unlikely that this project could be completed in one year, so the
student would need to also come up with a set of logical stages for the
project and plan to undertake a subset of these as their particular
project.
A good knowledge of object-oriented programming and a good
imagination would be required for this project. An interest in
analog and/or digital modulation schemes would help but is not
essential.
Web-based kiosk for FEIT
Supervisor
Haley Jones [Dept of Engineering]
Background
The Faculty has been desiring a computer kiosk to be placed at
one or more of the entrances to the Faculty - perhaps one each for the
CSIT, Ian Ross and Engineering (and RSISE?) buildings. The
student would need to come up with a set of specifications for what
would be most useful on such a kiosk, and it would probably be best of
all 3 were the same. This would be web-based but structured differently
to the existing faculty web-pages. For example, someone using the kiosk
would need much less detailed information than someone surfing the web.
They may wish to know where someone's office is, who is doing research
in what area or where a particular lab is.
There is an existing content management system for the
Faciluty web pages. Making use of a common system and minimising
repeated editing and storage of common information would be desirable.
Ease of maintenance by non-programming staff is highly desirable.
A good knowledge of web-based programming would be useful for this
project.
Data linkage user interface
Supervisor
Peter Christen
peter.christen@anu.edu.au
Background
The ANU Data Mining group has been developing an open source
data linkage system called Febrl (Freely extensible biomedical record
linkage), which includes modules for data cleaning and standardisation
(of names, addresses, dates of birth), deduplication, data linkage (or
matching), and geocoding. See the project home page for more details
(including papers, presentations, code to download): http://datamining.anu.edu.au/linkage.html
The system is written in Python and currently is customised
and controlled by simple Python based scripts.
Software
Engineering student research project
Develop a user friendly graphical user interface, either Web
based or using one of the many GUIs available for Python.
Retrieving and presenting
captured experience
Supervisor:
Prof. Mick Cardew-Hall, Mr
Jeremy
Smith (Dept of Engineering, FEIT)
Background
An experience capture software system, called Simpress, has
been developed by the Department of Engineering to acquire and manage
information and experience arising from the automotive design and
production process. This has been implemented and running with a
partner in the automotive industry for over 2 years, primarily aimed at
capturing experience that is otherwise lost by the organisation.
Software
Engineering student research project
This software engineering project focuses on the next stage,
the retrieval and presentation of the experience collected to engineers
and designers during the design and assessment of new models and
parts. An umbrella framework and architecture needs to be
developed to extract relevant information stored in Simpress based upon
the user's current design activities, and present the information in a
intuitive and visual manner. This project would combine:
- interrogation and searching of a design and manufacturing
database, using methods such as case-based reasoning (CBR),
blackboards, text mining, CMaps or topic maps
- interaction and interoperability with CAD and design
packages, using APIs
- presentation and visualisation of results and part
information to the user, via web browsers and plugins
Aims/Outcomes:
- architecture framework for design and deployment of a
system combining commercially available CAD packages, the Simpress
system and additional components required for analysis or presentation
of results
- prototype implementation for one common commercial CAD
package and the existing Simpress system
More information
- Z. Li, 'Review of Product Information Retrieval:
Representation and Indexing', 2004
ASME Computers and Information in Engineering Conference, Sept
28-Oct 2, 2004, Salt Lake City, Utah, USA
- J. A. Penoyer, G. Burnett, D. J. Fawcett & S.-Y. Liou,
'Knowledge based product life cycle systems: principles of integration
of KBE and C3P', Computer-Aided
Design, Vol 32 (2000), pp311-320
- R. Pilani, K. Narasimhan, S. K. Maiti, U. P. Singh & P.
P. Date, 'A Hybrid Intelligent Systems Approach for Die Design in Sheet
Metal Forming', Int Journal of
Advanced Manufacturing Technology, Vol 16 (2000), pp370-375
- E. Doege & F. Boinski, 'Computer-aided design and
calculation of deep drawn components', Journal of Materials Processing Technology,
Vol 46, (1994), pp321-331
Hierarchical Garbage Collection
Simulator
supervisor John Zigman
http://djvm.anu.edu.au/projects//hgc.php
Benchmark: Towards High-performance
and Fault-tolerant Distributed Java Implementations
supervisor John Zigman
http://djvm.anu.edu.au/projects//benchmark.php
Cost Simulator: Towards High-performance
and Fault-tolerant Distributed Java Implementations
supervisor John Zigman
http://djvm.anu.edu.au/projects//cost.php
Optimizing for design patterns
Supervisor: Dr
Steve Blackburn and
Dr Alonso Marquez
Email:
Steve.Blackburn@anu.edu.au and
alonsomarquezes@yahoo.es
Background
The software engineering benefits of design patterns are
widely
understood. Consequently their use has become ubiquitous in industry.
This widespread uptake is in spite of many patterns incurring
significant performance penalties. This phenomena is only just emerging
and so far has not received attention from the programming languages
community.
Previously we have shown that bytecode transformations at
class
loading
time can be used to make significant performance optimizations for Java
programs. Our approach is powerful and unorthodox, sidestepping
heavyweight pointer analyses typically used by the mainstream compiler
community. This project will leverage this approach to optimize for
some heavily used design patterns.
We anticipate that this work could lead to publication.
Sane Garbage
Email:
Steve.Blackburn@anu.edu.au and
Daniel.Frampton@anu.edu.au
Background
Garbage collection is key to modern languages like Java and
C#. Our
group plays a leading role in the development of garbage collection
algorithms and we maintain the leading infrastructure for memory
management research, MMTk. Garbage collectors are notoriously difficult
to debug, as they subvert most assumptions about program behavior (by
manipulating the object graph behind the back of the application).
There are very few useful tools for debugging garbage
collectors---typically we must resort to primitive methods such as the
careful use of print statements!
This project will develop a heap sanity checker. The heap
sanity
checker will be completely unobtrusive, only observing, not altering
the state of the system. The sanity checker will always be present in
the system, though inactive unless invoked at the command line. The
checker would report on dangling pointers, uncollected garbage, etc.
The tool would also be extensible, allowing the addition of
algorithm-specific checks (such as reference count checks for a
reference counting collector). This tool would be invaluable to garbage
collection research.
Java to C++ Translation
Email:
Steve.Blackburn@anu.edu.au and
Robin.Garner@crsrehab.gov.au
Background
MMTk is a memory management toolkit that is an important tool
for
memory management research. MMTk is written in a special dialect of
Java which avoids dynamic memory allocation and the use of Java
libraries which might perform such allocation, and adds to Java unboxed
types for memory primitives such as Address, Word, and ObjectReference.
The result is a powerful, flexible and high performing toolkit which
allows researchers to rapidly experiment with garbage collection ideas.
MMTk was implemented within Jikes RVM, a Java virtual machine written
in Java. Since then MMTk has been ported to another virtual machine
written in C. Rather than port the source code from Java to C, it is
desirable to somehow produce a C library usable by the C program. In
the first instance, this was done using the gnu compiler for java, gcj.
We would like a more general solution. This project will write a Java
to C++ translator for MMTk. The limited subset of Java used by MMTk
makes the project feasible. We envisage the automatic translator both
generating C++ files and headers with inline functions in them,
allowing key elements of the interface to be inlined into the host
virtual machine, greatly improving performance.
Object recognition and
localisation of known objects by inverting ray-tracing
Supervisor
Eric McCreath
Background
Often model driven object recognition systems will extract
features such as lines within a scene. These are then mapped into some
invariant space and are matched against a set of known objects.[1] The
quality of the recognition and localisation is dependent on the quality
of the features extracted.
One approach that does not require feature extraction is a
direct comparison between a rendered image and the camera image. This
will involve searching over a space of the object's orientation and
localisation. Clearly this has considerable computational
overhead.
Project
The aim of this project is to survey the current single camera
object recognition and localisation approaches. Also the aim is to
study the viability of an approach that simply inverts a rendering
system. This project will also involve considering novel
approaches that could be used to reduce the search space.
An student that takes on this project would be required to
attend a weekly research meeting. Also the aim is to produce a paper
from the results of this research.
More information
[1] - Weiss and Ray, "Model-Based Recognition of 3D Objects
from One View" Proc. of European
Conf. on Vision, 1998
Network-level Optimization of
Communication on Cluster Computers
Supervisor
Peter Strazdins
http://cs.anu.edu.au/~Peter.Strazdins/postgrad/NICOptm.html
Extended Threads Emulation in an
SMP Computer Simulator
Supervisor
Peter Strazdins
http://cs.anu.edu.au/~Peter.Strazdins/postgrad/ExtLWPSulima.html
The Effect of Architectural
Variations on the NAS Parallel Benchmarks
Supervisor
Peter Strazdins
http://cs.anu.edu.au/~Peter.Strazdins/postgrad/NPBArchEvals.html
Supervisors:
Roland Goecke
Email:
Roland.Goecke@nicta.com.au
Tom Gedeon
Email:
Tom.Gedeon@anu.edu.au
Background
As computer systems become more and more part of our daily
life, the
issues of HCI and user-adaptive systems are highly important. In the
past, the user had to adapt to the hardware, but the trend nowadays is
clearly towards more human-like interaction through user-sensing
systems. Such interaction is inherently multi-modal and it is that
integrated multi-modality that leads to robustness in real-world
situations. One new area of research is affective computing, i.e. the
ability of computer systems to sense and adapt to the affective state
('mood', 'emotion') of a person. An essential capability is to track
and recognise facial expressions.
Project
The aim of this project is to perform some experiments on
facial
expression tracking and recognition. This project suits a student
who is interested in human-computer interaction with an emphasis on
computer vision / image processing.
The project involves an evaluation of currently used methods
for
facial feature tracking, the implementation of the most promising ones,
an evaluation of features representative of facial expressions, and
experimental work on automatically recognising some facial expressions.
The facial feature tracking involves the implementation of solutions
to the basic problems of finding a face in an image by skin colour
modelling and tracking the face by an appropriate face model (e.g.
active appearance models). Furthermore, methods need to be analysed
and developed for the tracking of facial features like the mouth or the
eyebrows. Based on the locations and movements of these facial
features, models need to be build that are representative of certain
facial expressions, so that these can be recognised.
Supervisors:
Roland Goecke
Email: Roland.Goecke@nicta.com.au
Tom Gedeon
Email:
Tom.Gedeon@anu.edu.au
Background
This project is also in the area of human-computer
interaction.
Automatic speech recognition (ASR) software has reached a state, where
it can be employed successfully in environments that don't have a lot
of acoustic background noise, like an office for example. However,
there performance is still low in noisy environments, for example in a
car or outdoors. One way of overcoming these problems is through the
addition of video from the person talking. Visual speech recognition is
successfully used by people who are able to lip-read and it has been
shown to improve ASR performace as well.
Project
This project is suitable for a student who is interested in
HCI
research, with an emphasis on signal processing in the audio and
video modality.
The aim of this project is to build an audio-video (AV) ASR
system and
to perform experiments on different ways of combining the audio and
video data to see which gives the best results. ASR systems often use
statistical models like Hidden Markov Models (HMM) and this technique
shall also be used here. The HMMs are built based on training data and
then they can be used in the recognition experiments. A convenient way
of building HMMs is through the Hidden Markov Toolkit (HTK). The
Audio-Video Australian English Speech data corpus AVOZES will be used
to provide both training and experimental data. The project will also
investigate different visual features to evaluate the usefulness of
geometric versus image-based features. Geometric features are based
on facial feature points (e.g. lip corners) and their movements while
speaking. Image-based features derive speech-related information from
the pixel values of, for example, the mouth region. Both approaches
have their pros and cons, and their respective performance will be
compared.
