22 July 2015 – An NUS Computing team supervised by Professor David Hsu and Associate Professor Lee Wee Sun excelled in a robotics challenge at the IEEE’s International Conference on Robotics and Automation (ICRA 2015), which was held in Seattle, Washington, USA, late May this year.
The team, whose members were second year PhD student Luo Yuanfu, and post-doctoral fellows Bai Haoyu and Ye Nan, won first place in the Humanitarian Robotics and Automation Technology Challenge, which was designed ‘to promote the use of robotics and automation technologies for the reliable and robust detection and classification of landmines’. The teams were evaluated on their robot’s ability (or inability) to accurately detect landmines, incidence of robot collisions, area swept and coverage time, and the number of triggered ‘explosions’.
According to Haoyu, who spoke on behalf of the team, “One of the challenging aspects is that we did not get many chances to test our algorithm on the real robot. When we first tested our algorithm on the real robot, the result was much worse than what we got in the simulation test. We [were] very upset [when] we knew the result of the ground test.”
Still the team prevailed with their solution. “We formulated the problem as a mathematical framework called Partially Observable Markov decision process (POMDP). The POMDP framework enables us to quantitatively model the uncertainties in the environment and robot motion. Our advanced algorithms can solve the POMDP model and instruct the robot to act optimally under uncertainty,” explained Haoyu.
ICRA is the oldest and largest robotics conference, with over 3000 attendees and 1000 papers presented annually.