A Study on the MUM-T Mission Using Multi Agent Reinforcement Learning

초록

The scope of missions that drones can perform has expanded, and the concept of MUM-T(Manned Unmanned Teaning) has emerged. Since the manned aircraft is responsible for the operation and control of many drones, it I essential to study the automatic path generation technique for multiple drones to reduce the pilot's mission load. This paper proposes the application of multi-agent reinforcement learning techniques as multiple drone automatic path generation techniques. Compared to single agent reinforcement learning, multi-agent reinforcement learning is advantageous to apply to environments where multiple agents with common goals, such as MUM-T environments. So in this paper, we conducted a study on the performance of a complex operation mission with a MA-POCA algorithm provided by Unity ML-Agent, and analyzed the probability of mission success through simulation.

키워드