| This thesis presents an effective algorithm for multi-agent path planning utilizing an improved artificial potential field-based regression search (improved APF-based RS) method. It coordinates multiple agents in a wide variety of practical situations. We show that the algorithm can generate near-optimal trajectories while avoiding pathological issues in which an agent gets stuck due to local minima in the APFs or induced oscillatory behavior. We consider a wide variety of operating environments, which might be known, partially known, unknown, static, and/or dynamic. Additionally, this thesis introduces signaling mechanisms (e.g, internal semaphores) and other adjustments and perturbations so as to maximize agent cooperation and minimize collision risk. The performance of our path planning algorithm is tested and validated through extensive simulation. |
