A Multi-Agent Reinforcement Learning-Based Positioning Control Algorithm for Line-of-Sight Communication Relaying in Micro-Robot Swarms

The operational environment in modern battlefields is rapidly evolving, with increasing missions in urban indoor spaces. Swarm micro-robots are widely used for reconnaissance to reduce risks to human personnel. To support effective operations, advanced swarm control technologies are required for navigation and cooperation in complex environments. This paper proposes a Multi-Agent Reinforcement Learning(MARL)-based positioning control algorithm that dynamically maintains Line-of-Sight(LoS) connectivity between agents. Unlike conventional methods assuming static or omnidirectional communication, relay robots reposition themselves in real time to optimize network topology and maintain stable information flow. The reward function incorporates LoS validation and distance-based signal attenuation, encouraging visibility-preserving formations. Simulation results demonstrate improved communication reliability and spatial coordination in uncertain environments.