Game-Theoretic Trajectory Planning of Mobile Robots in Unstructured Intersection Scenarios

Abstract

The motion of autonomous mobile robot platforms in a shared operation area can lead to intersection scenarios that cannot be resolved by individual and uncoordinated motion planning. A coordinated solution approach is needed to solve these kinds of scenarios for which game theory provides a suitable framework. Therefore, this paper presents a game-theoretic trajectory planner based on a nonlinear receding horizon control scheme. To handle complex collision avoidance in unstructured environments with no reference paths, the proposed method extends an existing sensitivity enhanced iterated best response algorithm. Our reformulation of the sensitivity cost term within the optimal control problem of each vehicle enables a cooperative and time-efficient collision avoidance behavior. Another benefit of our proposed approach is that no central coordination unit is required, and only minimal communication between the vehicles is necessary. We simulatively compare our decentralized approach in different intersection scenarios involving 2 to 4 mobile robots to a centralized multi-robot trajectory planner. The simulation results show that our algorithm resolves the intersection scenarios in most cases with only a minimum average duration extension until each vehicle reaches its goal state compared to the centralized planner.