Imperial College Robotics Competition 2012
Andrew Davison, reader in Robot Vision at the Department of Computing, Imperial College London, organizes every year as part of the Robotics Course a competition amongst all students where groups compete to build and program the robot which can most effectively achieve a certain challenge against the clock.
2012's challenge was to build a robot able to cross in the minimum amount of time an area filled with randomly located obstacles without bumping into any of them.
Our team of four (Antonio Azevedo, Clemens Lutz, Francesco Giovannini and myself) developed a solution based on an occupancy grid mapping technique that was adapted for high speed computation. Our robot relied on accurate odometry for self-localisation, and used a medium-distance sonar along with proximity active light sensors to detect obstacles. The data from the sensors was first gathered into a coarse-grained occupancy grid and a fast-planning algorithm was then used to find the optimal path to the destination. A thiner-grain grid was finally computed only along the route taken by the robot, hence decreasing computational needs. All grids were able to update near real-time without noticeably slowing down the obstacle detection, thus allowing to increase drastically the robot's speed.
Only a third of all teams were able to complete the whole course. Our robot heavily forestalled its competitors at each of the two runs:
Run #1: our robot finished first in 28 seconds; other finishers did it in 66 seconds in average.
Run #2: our robot finished first in 21 seconds; other finishers did it in 49 seconds in average.
Click the photograph to download a video of the robot
Antonio Azevedo, Nicolas Paglieri, Clemens Lutz & Francesco Giovannini