Energy-Efficient Multi-agent Cooperative Search Control Based on Deep Reinforcement Learning on Uneven Terrains

Abstract

Anti-flocking algorithm for multi-agent to search a given area of interest(AOI) has been relatively mature. Multi-agent is basically used to search for uneven terrain, but the most existing anti-flocking algorithms are designed for flat terrain, so agents often use the shortest path to move between navigation targets. Using the shortest path to move in uneven terrain will consume more energy. At present, agents basically use portable energy to provide power, so we should try to reduce energy consumption. This brief proposes an energy-efficient multi-agent cooperative search control based on deep reinforcement learning on uneven terrains. The proposed algorithm selects the navigation target points through deep reinforcement learning, and encourages the agent to move between the navigation target points along the contour line. Simulation results show that the proposed control protocol is a promising energy-efficient solution for multi-agent operating on uneven terrains.