基于多代理系统的算法集成模型及其博弈分析

Algorithm-integrated model based on multi-agent system & its game analysis

针对单个算法难以解决复杂优化的问题,提出了集成算法的黑板模型。各种算法被封装为代理(A-gent)参与优化问题的求解;对模型设计了优胜劣汰机制,使各Agent处于竞争环境中,迫使它们对模型的运算速度或优化能力做出贡献。为使模型保持较高运行效率,基于多方博弈观点,分析了不同算法Agent在竞争中需采取的策略,并探讨了几种非数值算法在模型中的个性化策略。以作业车间调度问题为例,验证了模型有效性。

A blackboard model of integrated algorithms was proposed to solve complicated optimization problems which couldn＇t be conquered within acceptable time by normal algorithms alone. In the model, various algorithms were capsulated as agents participating in solving optimization problems. A survival-of-the-fittest scheme was designed for the model, which forced agents either to improve the calculation speed or to enforce the optimization capability of the model. To keep the model operating efficiently, a game analysis aiming at agents＇ strategies was conducted, and personalized settings were designed for promoting some typical member algorithms＇ capabilities of the model. A testing example of job shop scheduling problem validated the effectiveness of the model.