Multi quadrotors coverage optimization using reinforcement learning with negotiation

Abstract

This paper proposes an optimization scheme to maximize the area coverage of multiple quadrotor unmanned aerial vehicles that are deployed to monitor an operational area/space. Each quadrotor initially performs a single agent reinforcement learning to determine target points with optimal coverage area. Whenever each quadrotor encounters the others within a predetermined negotiation region that is defined by an inter-agent distance threshold, it will activate a multiagent reinforcement learning with action negotiation algorithm and coordinate its movement policies to maximize the total coverage area and avoids inter-agent coverage overlaps. Results of simulation evaluations are shown to illustrate the performance of the proposed learning-based coverage optimization method.