Navigation

Raymond's Projects

Welcome to my Projects Page!

[2006] RoboCup Rescue Robot League

Gee I'm getting behind in my updates!

Well we were at RoboCup again in 2006, fielding a team of robots including our updated robot, CASTER Scorpion, and a couple of Redback robots!

A particular feature of our new look CASTER robot was the big sensor arm on the back of the robot, which allows it to direct perception and look into holes, over boxes and the like. The sensorhead itself contains a variety of cutting edge sensors including a SwissRanger SR-2B range imager, FLIR ThermoVision A10 thermal camera and an Allied Vision Technologies high resolution camera. All up, we can produce 3D maps, detect victims via bodyheat and look at objects with better than 20/20 vision! Other sensors on the robot included two Hokuyo scanline laser rangefinders for 2D mapping and an omnidirectional camera for driving. A JVC InterLink notebook computer provided onboard computation and communications.

Our Redback robots were also highly capable and carried two cameras and a laser rangefinder capable of building 3D maps. Each also carried an OQO 01+ PC for communication and computation.

Our software was also further developed with a new, more complete user interface that combined mapping and driving tasks and allowed for the control of multiple robots. We published a paper at Systems, Man and Cybernetics 2006 on this interface. Our communications infrastructure also advanced with a combination of Player and ICE middlewares providing for realtime control as well as on-demand complex datatypes like merged high resolution images with depth maps.

With so much new stuff we were hoping for a good turnout - we were able to easily make the semi-finals but much to our disappointment, we suffered wire fatigue in an RS-232 line at the back of the arm which disabled CASTER Scorpion - right where it was in the way of our Redback robots! We managed to get to a victim with a Redback but unfortunately weren't able to get a close enough look to score points ... ah well ... always next time! :-)

Created: 2007-05-12 22:07 Last modified: 2007-05-12 22:12 (Sydney time)

[2004,now] Ongoing ResearchIn the grand tradition of those undertaking Ph.D. research, my current exact topic is, well, not quite so exact ...

In general terms, I'm working on ways in which robot systems may be endowed with the ability to learn to accomplish tasks in a real world, continuous, robot-centric environment. My application domain is that of the RoboCup Rescue Robot League and my goal is to be able to tell the robot to "move forward 10m and stop" and have it do that whilst traversing unstructured terrain, such as random stepfields. The trick however is to do so using behaviours that have not been hand coded - we intend to use reinforcement learning!

To make the task somewhat interesting, the robot we've chosen to apply this to is the Redback robot (see below and http://redback.web.cse.unsw.edu.au/ ) which has 4 degrees of freedom - the tracks on each side and the flippers at either end. Controlling this robot is a highly non-trivial exercise, even for a human, and can become quite difficult as it is often not possible to traverse a variety of obstacles without doing particular actions.

Clearly plain reinforcement learning (and even any real variant thereof) will have a very difficult time with such a task so to make life easier, I'm looking at ways of incorporating background knowledge in order to reduce the search space. Currently I'm looking at behavioural cloning and the use of fast, low fidelity environment simulators in order to define a subspace of the search space that the reinforcement learner in the real world can then have a shot at. We've already had some success in getting a cloned behaviour working for our CASTER robot that allows it to traverse some stepfields so it's a start!

Challenges that I'll probably need to address in the process of developing such a system include:

3D sensing from a lightweight advanced mobility platform
Robot-centric environment representation
Dimensionality reduction for the state space
Continuous domain reinforcement learning
Methods of incorporating background knowledge into reinforcement learning
Methods of moving learned experience in simulation to equivalent real world environments

Quite a set of challenges! Wish me luck! :-D

Created: 2005-12-12 14:01 Last modified: 2006-03-10 17:05 (Sydney time)

[2005] The Building of RedbackI was invited by NIST to the 2005 Rescue Robotics Camp, held at the Istituto Superiore Antincendi (Firefighters Training Institute) in Rome, Italy, to present a lecture and practical sessions on constructing a robot that we've dubbed the "Redback"!

The goal of this robot is to be an easy-to-build, low cost platform that requires no special parts, tools or skills. The Redback robot has its own website here!

Created: 2005-12-12 13:45 Last modified: 2005-12-29 14:07 (Sydney time)

[2004, 2005] RoboCup Rescue Robot League 2005Well, we weren't the champions but 3rd place, for a 5 month old team, isn't bad!

Update: We're on TV! See the news page for more!

In 2005 I was once again part of RoboCup, this time in the Rescue Robot League! The aim of the competition is to test robot assisted search-and-rescue technologies. Teams field robotic systems that can enter a disaster arena about the size of a small house, find victims, determine their state of health and return with a map of victims and significant landmarks that is good enough to send human rescuers in with. Any human teleoperation is performed from a different room - in fact, during the competition, every morning the arena is moved around and messed up and team members aren't allowed to see the arena until after their run is complete!

I was the driver of our robot, CASTER, for all of the competitive runs, I was also one of the main hardware and software developers (in fact, most of the work on the robot and associated software was performed by myself and Dr. Waleed Kadous from The ARC Centre of Excellence for Autonomous Systems).

A number of publications resulted from our entry, please click on my Publications link in the menu for details!

Here is the short report on our team:
============================
CAS demonstrated its applied research into robot assisted search and rescue on the world stage through its involvement with the RoboCup Rescue Robot League competition. The Rescue project aims to promote and demonstrate relevant research in areas that include robotics for urban search and rescue, algorithms for sensing, mapping and victim identification, multi-robot co-ordination and autonomous behaviours and exploration.

Team CASualty, consisting of CAS members from UNSW and UTS, achieved
considerable success in its first attempt by coming 3rd overall in the 2005 competition, held in Osaka, Japan. The field included 26 entrants from many of the world's finest robotics research centres. The entry showcased some of the strengths in CAS in terms of hardware, mapping, sensing, algorithms and software.

Team CASualty fielded two robots for the competition. CASTER, a tracked,
advanced mobility robot, was equipped with state-of-the-art sensors and software for victim identification, 3D mapping and teleoperation. CASTER's ability to map, access and assess victims in very difficult terrain played a large part in the team's success. A wheeled, differentially driven robot, HOMER, was also presented. Though not deployed in the competition, demonstrated forward-looking techniques for autonomous navigation and map-building.

The team is currently preparing for the 2006 competition in Bremen,
Germany. It is envisaged that a co-operative team of autonomous and
semi-autonomous advanced mobility robots, equipped with mechanisms, sensors and software currently under development, will further demonstrate the strengths of CAS in this area.
============================

We have a (under construction!) website here!

Created: 2005-12-12 13:44 Last modified: 2006-01-24 02:27 (Sydney time)

[2003] Visual Tracking of a Small Domestic RobotAs a member of the 2003 RoboCup Four-Legged League soccer team, I was responsible for, amongst other things, feeding geometry-based localisation into the existing infrastructure, an Extended Kalman filter. Unfortunately, the information I was getting, such as "You are 30cm away from a line", really doesn't linearise well when that line could be anywhere on the field!

I decided to see if something else, such as a Monte-Carlo Localisation Filter, would do any better. So, when I went back to Curtin University to finish my degree (at the time, I was at the University of New South Wales as a cross-institutional student), I decided to implement the same sensor model but as an input to an MCL filter. The results were, I think, quite promising! The environment was an indoor, domestic office-style room with a regular pattern on the floor. This is really hard to linearise as any one observation of the pattern (crosses on the floor) could mean you were in one of 48 distinct positions (if you wanted to do this with a multimodal Kalman filter you'd need at least 48 distributions). We also had overhead cameras but they weren't terribly high resolution and there were many opportunities for occlusion, combined with other moving objects.

For more details, please see our paper! This was published at the 2004 RoboCup Symposium.

Created: 2005-12-12 13:44 Last modified: 2005-12-29 14:27 (Sydney time)

[2003] PentagoonPartway through 2003, a friend, David Wyatt, and I came across a design in Dr. Hans Moravec's book, "Mind Children", for a novel design for robot locomotion. A cross between wheels and legs, this design could run across a smooth surface as quickly and efficiently as any wheeled vehicle, yet it could climb stairs, including open stairs, as well as any equivalent legged vehicle!

The key to this high degree of mobility lay in the robot having 5 vertically telescopic legs, arranged at the points of a regular pentagon, at the base of which was a wheel. Travelling on a flat surface, each wheel simply rotates as normal (and in the original design, attributed to Hitachi, was steerable). To climb stairs and other obstacles, each wheel was raised in turn, the robot rolled forward then the leg put down on top of whatever obstacle was encountered.

We decided to build this robot and, over the course of a couple of months, designed and manufactured the 5 leg modules, each consisting of two servos, sense switches, telescoping mechanism and microcontroller board, and programmed these boards and a central microcontroller to automatically climb stairs! The control mechanism was elegantly simple - each leg was raised if its forward bump sensor hit something except of another leg was already raised, then the robot was moved forward, then the leg put down until its vertical contact sensor was activated. A leg that fell off an obstacle would be put down until its vertical contact sensor was reactivated.

We were very fortunate when we demonstrated the robot - we encountered a case of the inverse demo effect! We couldn't get it to work in practice but in front of the professor, it worked beautifully! It climbed and descended stairs we had built for it 3 times and climbed over a multimeter without missing a beat! Pitty we couldn't make it do the same when we wanted to video it ... ah well :-)

We looked through publications, patent records and the like and couldn't find any reference to the Hitachi robot, we even asked Dr. Moravec if he had any more details about the robot but he couldn't find anything. Then in 2004, I was in contact with Dr. Shigeo Hirose who had more recently also developed a similar 8 wheeled robot, dubbed "ZeroCarrier"! According to him, the Hitachi robot was by a Dr. Ichikawa but no reference could be found to it either.

My old photo gallery of it is still up here!

Created: 2005-12-12 13:43 Last modified: 2005-12-29 14:04 (Sydney time)

[2002, 2003] RoboCup Four-Legged League 2003

We are the champions!

In the Italian city of Padua, the rUNSWift team from the University of New South Wales regained their title of the world champion RoboCup Four-Legged League soccer team!

For those of you unfamiliar with the competition, RoboCup consists of a number of individual competitions that aim to test robot technologies in a competitive environment. The Four-Legged League involves teams of Sony Aibo ERS-210A (in 2003) robots that play soccer against each other! These robots are not remote controlled - the (human) teams must program them to behave appropriately, using its sensors (primarily its onboard camera) to observe the world and act accordingly. Quite a challenge!

My primary tasks as a member of this team were:

Develop and implement solutions for two technical challenges

Playing with a Black and White Ball (shoot a black-and-white ball into the goal)
Localisation challenge (accurately move to several locations on the field without the aid of localisation beacons)

Develop and implement a solution for the main game code for localising the robot from the field markings

Particularly helpful for moving the ball up the sides of the field quickly without getting stuck!

My technical challenges came 5th and 1st in the world respectively! I'd also like to think that my contribution to the main game code also helped us in our 4-3 win against the University of Pennsylvania in the grand final!

Created: 2005-12-12 13:42 Last modified: 2005-12-29 14:06 (Sydney time)

[2001] WARDEN Video Surveillance System

A video analysis, indexing and retrieval system for connecting to a surveillance camera multiplexer. Designed to run on a mid-range PC (in 2001 - a 500MHz Athlon), Warden provided the following features:

Demultiplexing of 16 camera streams
Motion graphing for easier searching
Motion thresholding to save space
Rotating buffers so periods leading up to "interesting" events are guaranteed to be saved
2 months of recording on the internal 80GB HDD
Java front-end for quickly finding periods of interest in the video stream

Despite being written in 2001, the system was still in use as recently as 2004 (and might still be - I've not been back since!). During that time it has helped to resolve several security incidents without needing an officer to wade through hours of tape.

Created: 2005-12-12 13:41 Last modified: 2005-12-13 10:32 (Sydney time)

Powered by MODx 0.9.5 (rev 2106).

Source: database Created: 2005-12-12 12:32 Last modified: 2006-12-27 22:05 (Sydney time).