Candidate: YiFeng Wang
Title: A Pursuit Evasion Game Approach to Obstacle Avoidance
Date: August 13, 2021
Time: 11:00 AM
Place: Remote
Supervisor(s): Smith, Stephen L. - Nielsen, Christopher
Abstract:
We propose an event triggered obstacle avoidance control scheme that guarantees the safety of a mobile robot in a material transport task at a warehouse. This collision avoidance strategy is based on a pursuer static obstacle (P.S.O.) game which is a modification of the reversed homicidal chauffeur game. In the PSO game, the evader represents the mobile robot and is modelled by a Dubins vehicle while the pursuer represents the human working at the warehouse and is modelled as an omnidirectional agent.
Additionally, we consider the presence of a static obstacle in the work space. Based on our analysis, the evader has two options to avoid both the pursuer and the obstacle. The first option is to turn hard in the opposite direction of both the pursuer and the obstacle. The second option is to go in between the obstacle and the pursuer. The work of the reversed homicidal chauffeur game shows that the first evasive strategy guarantees the mobile robot's safety. However, a clear advantage of implementing the second evasive strategy is that the mobile robot can delay its evasive maneuver and hence the mobile robot can follow its predetermined path for a longer period of time. Therefore, the mobile robot actively checks the feasibility of executing the second evasive strategy rather than executing the first evasive strategy right away.
Based on a numerical study of our collision avoidance strategy, we argue that both the task duration and the path error due to the evasive action are reduced when compared to simply turning hard in the opposite direction of both the pursuer and the obstacle. Our event triggered obstacle avoidance controller is also compared to strategies based on ISO 13482 and the solution of the velocity obstacle set which are commonly used in the current robotic industry. Based on this comparison, we argue that there is a clear trade-off between the reduction in duration and path error and the guarantee of safety during a material transport task.