基于Agent与分解协调的综合生产计划研究  

Study on Decomposition-Coordination and Agent Based Aggregate Production Planning

作　　者：张人千[1] 李艳艳[1] 

机构地区：[1]北京航空航天大学经济管理学院,北京100083

出　　处：《数学的实践与认识》2007年第3期40-50,共11页Mathematics in Practice and Theory

基　　金：国家自然科学基金(70501002)

摘　　要：以作业单元为局部决策Agent,车间管理者为全局协调Agent,引入生产节点间的内部结算价格,基于多Agent系统,建立了综合生产计划的分布式决策模型.通过将局部Agent决策目标的总和与全局Agent决策目标进行对比,证明了所引入的内部结算价格就是全局Agent目标函数关于物流平衡约束的Lagrange乘子.基于Lagrange分解协调原理,设计了局部作业单元Agent和全局协调Agent的迭代协调算法.该迭代算法以上次计算的中间结果作为对其它作业单元生产需求的估计,从而能将各个生产单元Agent的决策模型分离,实现了分布建模与求解.在算例研究中使用启发式规则来确定Lagrange乘子迭代的步长系数,保证了较好的收敛性,证明模型和算法是有效的.This paper presents a distributed decision model for aggregate production planning. The model is based on MAS system, which regards activity unit as local agent and workshop manager as global coordinator agent. Internal price is introduced and is used by global agent to coordinate local agents. Comparing the object of global agent and the sum of objective functions of local agents, it is verified that internal price just is Lagrange multipliers of the objective function of global agent for logistic balance constraints. So, a decomposition-coordination iteration algorithm is produced as the negotiation rule between local agents and global agent, In the iteration procedure, the results of last iteration are used as approximation of production activity units＇ demand. So decision models of local activity agents are independent and can be solved individually. For better constringency in numeric example, a heuristic criterion is used to compute the iterative step of Lagrange multiplier. And, the numeric study proves that the decision model and algorithm is effective.

关 键 词：生产计划 AGENT 拉格朗日 分解协调 

分 类 号：F224[经济管理—国民经济]