基于Q-学习算法的交通控制与诱导协同模式的在线选择

采用Q-学习算法实现了交通控制与诱导协同模式的在线选择。首先,采用Q-学习算法训练多智能体,根据多智能体内部的推理得到不同交通状态下的最优协同模式,最终实现交通控制与交通诱导协同模式的在线选择与转换。仿真结果表明,本文提出的基于Q-学习算法的协同模式选择方法在一般交通拥挤状态下具有较好的协同控制效果,对比离线式模式选择方法更能适应交通状态的不断变化,从而达到有效避免严重交通拥堵、改善路网性能的目的。

On-line selection method of the traffic control and route guidance collaboration mode based on Q-learning algorithm

The on-line traffic control and route guidance collaboration mode selection was realized by the Q-learning algorithm. Using the multi-intelligence agents trained with the Q-learning algorithm, the optimal collaboration mode was obtained under different traffic conditions according to the inner inference of the multi-intelligence agent. So,the on-line selection and switching of the traffic control and route guidance collaboration mode was accomplished. The simulation results show that the proposed collaboration mode selection method based on the Q-learning is characterized by better collaboration control effect under the ordinary traffic congestion condition and more adaptive to constantly changing traffic condition than the traditional off-line mode selection method. The proposed method is helpful to avoiding the heavy traffic congestion and improving the traffic network performance.