Cooperative Collision Avoidance via proximal message passing

We propose a distributed controller to solve the Cooperative Collision Avoidance problem. We consider a network of vehicles, each with its own dynamic constraints and objective. The problem is to minimize the total network objective function subject to the vehicles' individual constraints and their shared collision avoidance constraints over a given time horizon. The proposed controller, a proximal message passing (PMP) algorithm, is iterative: At each iteration, every vehicle passes simple messages to its neighbors and then solves a convex program that minimizes its own objective function and a simple regularization term that only depends on the messages it received in the previous iteration. As a result, the method is completely decentralized and needs no global coordination other than synchronizing iterations. The problems that each vehicle solves can be done extremely efficiently and in parallel. We demonstrate the method on several examples using a model predictive control framework.