基于自适应精确罚函数的分布式资源分配算法

An adaptive exact-penalty-based distributed resource allocation algorithm

在多智能体系统中,分布式资源分配问题是近年来研究热点之一.分布式资源分配问题旨在通过智能体间信息交互实现资源最优配置.其中智能体局部约束给算法设计带来巨大挑战.首先,针对一阶多智能体系统,提出基于自适应精确罚函数的分布式资源分配算法,其中各智能体利用距离函数实现局部约束求解.此外,自适应设计思想旨在避免算法对全局先验知识获取.其次,利用跟踪技术实现二阶多智能体系统算法设计.并利用凸函数和非光滑分析法给出严谨的收敛性分析.最后,仿真结果验证了本文所设计优化算法对强凸分布式资源分配问题的有效性.

Recently, the distributed resource allocation problem is one of the important issues in multi-agent systems.The distributed resource allocation problem aims to realize the optimal allocation of resources through the information interaction between agents. The local constraints of each agent bring great challenges to the algorithm design. First, an adaptive exact-penalty-based distributed resource allocation algorithm is proposed for the first-order multi-agent system,in which the local constraint is reformed by the distance function. Besides, the priori computation or knowledge of the global cost function is avoided based on the adaptive control scheme. Second, the above proposed first-order algorithm is modified for the second-order multi-agent system based on the tracking control technology. Then, by virtual of the nonsmooth analysis and convex function theory, the rigorous convergence analysis is given. Finally, the proposed algorithms are claimed effectively by the simulation examples.