Increasing Feasibility with Dynamic Priority Assignment in Distributed Trajectory Planning for Road Vehicles

Abstract

<p>Nonconvex optimal control problems for large-scale networked control systems (NCSs) can be distributed to accelerate computation time. One distribution strategy is priority-based non-cooperative distributed model predictive control. The priorities assigned to agents determine if a feasible solution for all agents exists. This problem has been investigated in the domain of robotics for many years, and has recently been picked up in the domain of road vehicles. We develop a dynamic priority assignment algorithm for road vehicles based on an approach from the domain of robotics. In our algorithm, each vehicle determines its priority in a distributed fashion. The priority of a vehicle increases with the number of potential collisions on its planned path to increase the success of the distributed control problem. We further propose an approach for priority-based non-cooperative distributed model predictive control with dynamic priorities and prove recursive feasibility for the NCS. This paper presents the first evaluation in experiment of the priority assignment algorithm. We compare the algorithm’s performance to dynamic random priorities and to static priorities.</p>