Large-Scale Drone Swarm Countermeasures Technology Based on MF-Q Algorithm
Authors:
- 29 March 2024
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 125-129
Abstract
During the confrontation process, the UA V swarm faces challenges such as a complex battlefield environment with high threats, high dynamics, high denial, strong confrontation, incomplete information, and dynamic targets, which requires the UA V swarm to have a high degree of intelligence and coordination. This paper proposes a large-scale UA V swarm confrontation technology based on the MF -Q algorithm, that is, the Q learning algorithm based on the mean field theory, to realize the autonomous decision-making of the UA V swarm. The MF -Q algorithm can effectively reduce the dimensions of the state space and action space of the UA V swarm, reduce computational complexity and communication overhead, and improve learning efficiency and convergence speed. This article verifies the effectiveness of the technology proposed in this article through simulation experiments. The results show that the technology proposed in this article can enable UA V swarms to achieve efficient collaboration in confrontation scenarios, improve the confrontation winning rate and survival rate, and provide methodological reference for cluster confrontation problems.Keywords
This publication has 5 references indexed in Scilit:
- Model-free mean-field reinforcement learning: Mean-field MDP and mean-field Q-learningThe Annals of Applied Probability, 2023
- Mean-field theory for scale-free random networksPhysica A: Statistical Mechanics and its Applications, 1999
- Markov games as a framework for multi-agent reinforcement learning1994
- Explorations of the mean field theory learning algorithmNeural Networks, 1989
- The mean-field theory of nuclear structure and dynamicsReviews of Modern Physics, 1982