论文部分内容阅读
以无人机(Unmanned Aerial Vehicle,UAV)和无人车(Unmanned Ground Vehicle,UGV)的异构协作任务为背景,提出了一种地-空异构多智能体协作覆盖模型,通过UAV和UGV的异构特性互补,扩展和改进了异构多智能体动态覆盖问题。在覆盖过程中,UAV利用速度与观测范围的优势对UGV的行动进行指导;同时考虑智能体的局部观测性与不确定性,以分布式局部可观测马尔科夫(Decentralized Partially Observable Markov Decision Processes,DEC-POMDPs)为模型搭建覆盖场景,并利用多智能体强化学习算法完成对环境的覆盖。最后通过仿真实验验证了覆盖模型的有效性。
Based on the heterogeneous collaborative mission of Unmanned Aerial Vehicle (UAV) and Unmanned Ground Vehicle (UGV), a collaborative overlay model of ground-air heterogeneous multi-agent is proposed. Through UAV and UGV Complementarity of heterogeneous features, and extend and improve the dynamic coverage of heterogeneous multi-agent. In the process of coverage, UAV makes use of the advantages of velocity and observation range to guide UGV’s action. At the same time, considering the local observability and uncertainty of the agent, Decentralized Partially Observable Markov Decision Processes DEC-POMDPs) to build a coverage model for the model, and use the multi-agent reinforcement learning algorithm to complete the environment coverage. Finally, the validity of the coverage model is verified by simulation experiments.