有限阶段马尔可夫决策的可变限速控制模型

Control model of variable speed limit based on finite horizon Markov decision-making

分析了高速公路主线可变限速控制的作用,研究了现有的限速方法,将高速公路主线可变限速控制过程看作是离散时间的马尔可夫决策过程,提出基于强化学习与有限阶段马尔可夫决策的可变限速控制模型,通过与交通环境的交互学习进行模型的动态调整。采用有限阶段向后递归迭代的算法对模型进行求解,运用Paramics仿真软件对长吉高速公路全程进行仿真。仿真结果表明:在平均限速值低于设计时速6.25%的情况下,平均流量不仅没有降低反而增加了3.20%。可见,该模型可以有效提高交通流量,改善高速公路主线的交通状况。

The function of variable speed limit（VSL） on expressway mainline was analyzed,and the existing methods of speed limit were studied.The control process of VSL was taken as Markov decision-making process of discrete time.A model of VSL based on reinforcement learning and finite horizon Markov decision-making was proposed.The model was dynamically adjusted through interacting with traffic environment,and solved by using finite horizon backward recursive iterative algorithm.The traffic environment of Chang-Ji Expressway was simulated by using Paramics.Analysis result shows that average traffic volume doesn＇t reduce,but increases by 3.20% when average limit speed decreases by 6.25% compared with design speed.So the model is feasible to increase traffic volume and improve traffic condition on expressway mainline effectively.1 tab,4 figs,14 refs.