Wu Han, Shang Huiliang
Research Center of Smart Networks and Systems School of Information Science and Engineering, Fudan University, Shanghai 200433, PR China.
Research Center of Smart Networks and Systems School of Information Science and Engineering, Fudan University, Shanghai 200433, PR China; Academy for Engineering and Technology, Fudan University, Shanghai 200433, PR China.
ISA Trans. 2020 Jul;102:208-220. doi: 10.1016/j.isatra.2020.03.004. Epub 2020 Mar 7.
This paper proposes a novel distributed multi-agent dynamic task allocation method based on the potential game. Consider that the workload of each task may vary in a dynamic environment, and the communication range of each agent constrains the selectable action set. Each agent makes the decision independently based on the local information. Firstly, a potential game-theoretic framework is designed. Any Nash equilibrium is guaranteed at least 50% of suboptimality, and the best Nash equilibrium is the optimal solution. Furthermore, a time variant constrained binary log-linear learning algorithm is provided and the global convergence is proved under certain conditions. Finally, numerical results show that the proposed algorithm performs well in terms of global searching ability, and verify the effectiveness of the distributed dynamic task allocation approach.
本文提出了一种基于势博弈的新型分布式多智能体动态任务分配方法。考虑到在动态环境中每个任务的工作量可能会有所不同，并且每个智能体的通信范围限制了可选择的动作集。每个智能体基于本地信息独立做出决策。首先，设计了一个势博弈理论框架。保证任何纳什均衡至少有50%的次优性，并且最佳纳什均衡是最优解。此外，还提供了一种时变约束二元对数线性学习算法，并在一定条件下证明了其全局收敛性。最后，数值结果表明所提出的算法在全局搜索能力方面表现良好，并验证了分布式动态任务分配方法的有效性。
