Trajectory Optimization Using Sequential Convex Programming with Collision Avoidance

In multi-robot systems, it is commonly used collaborative approaches to solve complex tasks faster and efficiently. In most of those approaches, the decisions are made centralized or require global information about the objective or the robots, limiting many real implementations. The present work is based on a decentralized solution for the Rendezvous problem, by using only local information about the robots and asymmetrical information about the meeting point. As the primary contribution, we propose a sequential convex programming approach to overcome the non-convexities when physical spaces are taken into account in the optimization problem, which provides the robots with the collision avoidance capability during their movement to the target point. Experiments are also performed to show the effectiveness of the proposed approach. © 2018, Brazilian Society for Automatics--SBA.