CAJ | 학술논문

在分布式信息查询系统中,使用多个Agent协作完成查询任务是一种有效的方法,所有Agent的总行程影响网络的通信流量,单个Agent的最大负载决定了壹询任务的完成时间.现有方法大多研究如何减少Agent的总行程,未考虑Agent的负载均衡问题.本文提出一种基于负栽均衡的多Agent迁移路线规划(Load balancedmulti-agent planning,LBMAP)算法首先寻找图中一条包含所有节点的TSP回路,然后使用动态规划算法将该回路分为多段,每个Agent访问其中一段,算法兼顾了两个优化目标:最小化Agent的总行程、最小化Agent的关键负载.仿真实验表明:随着Agent平均访问节点数的增大,LBMAP算法的性能趋近于理论最优值.
재분포식신식사순계통중,사용다개Agent협작완성사순임무시일충유효적방법,소유Agent적총행정영향망락적통신류량,단개Agent적최대부재결정료일순임무적완성시간.현유방법대다연구여하감소Agent적총행정,미고필Agent적부재균형문제.본문제출일충기우부재균형적다Agent천이로선규화(Load balancedmulti-agent planning,LBMAP)산법수선심조도중일조포함소유절점적TSP회로,연후사용동태규화산법장해회로분위다단,매개Agent방문기중일단,산법겸고료량개우화목표:최소화Agent적총행정、최소화Agent적관건부재.방진실험표명:수착Agent평균방문절점수적증대,LBMAP산법적성능추근우이론최우치.
In the agent-based distributed information systems, it is ordinary to use multiple agents to complete the task. The total trip impact on the network communications traffic, and the largest single workload determines the time to complete the task of inquiry. Existing methods focus primarily on reducing mobile Agent's total trip, while does not considering the load balancing problem. In this paper, a load balancing alogrithm for multi-agent itinerary planning is proposed. Firstly, the algorithm finds a TSP circuit, and then divide the circuit into segments, each Agent visits a segment separately. The algorithm considers two objectives when planning the Agent's travel path: minimizing the total trip and minimizing the critical workload. Simulation results show that LBMAP alogrithm produces near-optimal performance with increasing average number of visiting nodes.