基于鞍点法的自适应分布式资源分配算法

研究一类带有不等式约束为凸函数的多智能体系统分布式资源分配问题.在资源分配问题中,各智能体拥有仅自身可知的局部成本函数和局部凸不等式约束.分布式资源分配旨在如何利用智能体间的信息交互设计一种分布式优化算法,完成定量资源分配的同时还保证最小化全局成本函数.针对该问题,基于卡罗需-库恩-塔克条件和比例积分控制思想,首先提出一种自适应分布式优化算法,其中凸不等式约束的对偶变量可实现自适应获取;然后,为了降低系统的通信资源消耗,设计一种动态事件触发控制策略以实现离散时间通信的分布式资源分配算法;最后,通过数值仿真验证所设计算法的有效性.

An adaptive distributed resource allocation algorithm via saddle point dynamics

This paper studies the distributed resource allocation problem with convex inequality constraints over the multi-agent systems. The local cost function and convex inequality constraints are known by themselves of each agent in the resource allocation problem. The aim of the distributed resource allocation problem is how to design a distributed optimization algorithm by using the information exchange between neighboring agents while minimizing the global cost functions. For this problem, based on the Karush-Kuhn-Tucker condition and proportional integral control idea, we firstly propose an adaptive distributed optimization algorithm, using which the dual variable of the inequality is obtained adaptively. Then, to reduce the communication resource consumption of the system, the discrete-time communication of the distributed resource allocation algorithm is realized by designing a dynamic event-triggered control scheme. Finally, the numerical simulation shows the effectiveness of the proposed algorithms.