Performance of a switching controller for pursuit-evasion scenarios with noisy measurements

This work presents an application of a switched control law for the pursuing agent in the Homicidal Chauffeur game. Past work by the authors has focused on the development of a fast decision making algorithm for pursuit-evasion scenarios based on partitioning the state space with regional objective functions chosen from a library of available strategies. The algorithm determines where, in terms of the state space, that each objective should be applied. In doing so, what is ordinarily a minimax problem becomes a simpler optimal control problem. One advantage of this strategy is that it is possible to rapidly update a control law based on measurements of another agent's dynamic capabilities acquired during play. In this work, we detail the construction of a switched controller for the Homicidal Chauffeur game. The pursuer initially has no knowledge of the evader's dynamics and must use speed measurements to update its control law during the game. Results outlined in this paper show a 25.6% increase in capture time compared to the case with perfect information with no decrease in the number of states that ultimately lead to capture.