In addition to this year's RoboCup soccer competition, UNSW-CSE competed in the rescue robot league as part of team from the ARC Centre of Excellence for Autonomous Systems. In this competition, we build and program robots to traverse rough terrain to search for "victims" in a simulated building collapse. We won the award for best autonomous robot (most are remotely controlled, whereas ours can operate on their own). We also won the mobility challenge. This is intended to test the robot's ability to traverse very difficult terrain. We set a record for this.
These competitions always have their ups and downs and are definitely "character building". Our early runs went really well but then we violated a fundamental rule of competition strategy: don't make last minute, untested changes. On the second day we added a sensor that we thought couldn't possibly have unexpected side-effects but, of course, it did. Power to the rest of the robot was compromised with the result that we had a couple of disastrous runs. Needless to say, we were pretty depressed but, to the team's great credit, the problem was solved and we were able to recover and produce more great performances.
This year we introduced autonomous victim identification. This was done by Tim Wiley who used a combination of laser scanning to find likely victim locations, combined with thermal imaging to confirm the presence of the victim. This was very impressive work, especially considering that Tim is a CSE undergraduate. He might have taken a little too seriously our joking around that he wouldn't get a HD unless we won but I guess we have to give it to him now!
Adam Milstein's SLAM was, once again, excellent despite the robots having to navigate twisty mazes with slippery, uneven floors, murderous step fields, staircases and ramps. This year, 3D position tracking with a scanning laser was added, making it easier for the wheeled robot to recognise places to avoid, like the stairs and steep ramps.
We probably have the most complex software system of any of the teams, with the same software running three completely different robots in autonomous and tele-operated modes. It all hangs together thanks to Matt McGill's outstanding software engineering.
Raymond Sheh was our driver again. Although you wouldn't guess it from his normal behaviour, he is remarkably cool under pressure. In the mobility challenge, he drove the Negotiator robot to a record number of laps around the most difficult terrain in the arena, though he did burn out a motor doing it! Best of all, he has finally submitted his thesis!
Extra special thanks to Rudi Saleh for his heroic work in building and keeping all the hardware going and for his patience amidst the chaos. Where other teams had an army of engineering students building their robots, we had Rudi. He did an enormous amount of work with remarkable grace.
For the first time, this year we took UTS's PackBot to the competition. When you have to cover a large area, multiple robots, really help and the PackBot was used to find several extra victims. Jaime Valls Miro got the robot operational and Mohammad Norouzi kept it going at the competition.
Reza Farid has been working on 3D object recognition for rescue and collected a lot of data that we hope will help us add new capabilities in the future.
