Abstract: First, the paper briefly introduces the architecture of traffic signal control system. Then the paper analyses the advantage of multi-agent technique. On the basis of analyzing the relative literature, the paper puts out multi-agent traffic control system structure based on intersection mechanism between agents. Then, the paper applies the hybrid genetic algorithm as the computing model of intersection agent, makes delay as PI value adopts punish function method to optimize the split. From the optimization results of Matlab application, it can be concluded that hybrid GA has better optimization result than hill-climb method.

Key words: traffic signal control, hybrid genetic algorithm, agent, multi-agent system

摘要： 首先简述了交通控制系统的控制结构，然后对多Agent技术在控制系统中的应用优势进行了分析，在对已有相关文献进行分析的基础上，提出了以路口Agent为单位的多Agent交通控制系统结构和路口Agent的内部结构，并对Agent间的协调机制进行了讨论.以混合遗传算法作为路口Agent的计算模型，以延误作为PI值，采用惩罚函数对绿信比进行了优化，通过matlab编程进行实际数据验证，结果表明混合遗传算法的优化结果优于已有的爬山法.

关键词: 交通信号控制, 混合遗传算法, 智能体, 多智能体系统