A Game Theoretic Approach to Distributed Planning for Serving Cooperative Tasks with Time Windows

Abstract

This thesis addresses the distributed planning of robot trajectories to efficiently serve tasks that have specific locations and time windows. Each task is completed when a sufficient amount of robots serve at the task location during the specified time window. A value is received by the participating robots upon successful completion of the task. These tasks arrive periodically over episodes. The objective of the robots is to plan their trajectories at the beginning of each episode to maximize the value of completed tasks. The path planning problem is mapped on to a game theoretical formulation and distributed learning is used to maximize the collective performance of the robots. Simulation and experimental results are provided for various cases to demonstrate the performance of the proposed approach.