Distributed UUV cluster task allocation algorithm based on deep reinforcement learning

oleh: HAO Guanjie, YAO Yao, CHANG Peng, ZHANG Xiaoshuang

Format: Article
Diterbitkan: Editorial Office of Command Control and Simulation 2023-06-01

Deskripsi

Task allocation is one of the basic and key issues in the research of agent clusters. UUV clusters are limited by underwater detection and communication capabilities in task allocation. UUV individuals can only obtain partial information around them, and conventional global algorithms cannot be applied well. A task allocation algorithm based on deep reinforcement learning and distributed UUV cluster organizational structure is proposed. The algorithm first realizes the partial task allocation of each UUV individual, and then the information between adjacent individuals is consistent and coordinated, thus realizing the optimal task allocation of UUV cluster. Through the simulation experiment, the algorithm in this paper converges faster than the genetic algorithm, has less traffic and higher task allocation efficiency than the contract network algorithm, and the distributed architecture does not rely on the "command center". The UUV cluster system has higher robustness and higher task allocation reliability.