复杂任务下的多智能体协同进化算法  

Cooperative evolution algorithm of multi-agent system under complex tasks

作　　者：刘家义 岳韶华[1,2] 王刚[1,2] 姚小强[1,2] 张杰[1,2] LIU Jiayi;YUE Shaohua;WANG Gang;YAO Xiaoqiang;ZHANG Jie(Air and Missile Defense College,Air Force Engineering University,Xi’an 710054,China;Graduate School,Air Force Engineering University,Xi’an 710054,China)

机构地区：[1]空军工程大学防空反导学院,陕西西安710054 [2]空军工程大学研究生院,陕西西安710054

出　　处：《系统工程与电子技术》2021年第4期991-1002,共12页Systems Engineering and Electronics

基　　金：国家自然科学基金青年科学基金(61703412);中国博士后科学基金(2016M602996);国家自然科学基金(61503407,61806219,61703426,61876189)资助课题。

摘　　要：针对多智能体系统在处理复杂任务时存在的低效率、高冗积、多智能体系统内协同模型算法存在交互冲突、资源损耗过高等问题,提出一种基于复杂任务的多智能体系统优化算法。在差分进化算法与局部优化算法的基础上对二者进行优化,结合强化学习的训练框架,构建训练网络,通过对学习步长进行修订,改变种群迭代优化准则,使得种群在计算力充足的情况下可以实现全局收益最大化,有效解决了指挥控制系统过程中的协同优化问题。In order to solve the problems of low efficiency,high redundancy,interaction conflict and high resource consumption of multi-agent system in dealing with complex tasks,this paper proposes an optimization algorithm of multi-agent system based on complex tasks.It is improved based on the differential evolution algorithm and the local optimization algorithm and combined with the training framework of reinforcement learning to construct the training network.By modifying the learning step and changing the iterative optimization criteria of the population,the population can maximize the global overall with the sufficient computing power,which effectively solves the collaborative optimization problem in the process of command and control system.

关 键 词：多智能体系统 协同算法 指挥控制系统 强化学习 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]