Cooperative and Geometric Learning for path planning of UAVs

Abstract

We propose a new learning algorithm, named Cooperative and Geometric Learning (CGL), to solve maneuverability, collision avoidance and information sharing problems in path planning for Unmanned Aerial Vehicles (UAVs). The contributions of CGL are threefold: 1) CGL exploits a specific reward matrix G, which leads to a simple and efficient algorithm for the path planning of multiple UAVs. 2) The optimal path in terms of path length and risk measure from a given point to the target point can be calculated. 3) In CGL, the reward matrix G is calculated in real-time and adaptively updated based on the geometric distance and risk information shared by other UAVs. Extensive experimental results validate the effectiveness and feasibility of CGL on the navigation of UAVs.