基于显式模型预测控制的大规模路网交通信号控制方法

A novel approach for traffic signal split based on explicit model predictive control

随着城市机动车保有量的不断增加,交通拥堵问题变得日益严重。作为一个有效的方法,模型预测控制(MPC)已经被应用于城市交通路网的控制中。然而由于需要在线进行优化,模型预测控制的计算负荷巨大,限制了其在对实时性要求比较高的场景中的应用。针对该问题,基于存储转发模型将显式模型预测控制(EMPC)引入大型交通路网的交通信号控制问题中。EMPC将多参数规划理论引入到线性时不变系统的约束二次优化控制问题的求解中,通过离线计算得到对应每个状态分区上的状态反馈最优显式控制律。实验结果显示,由于将在线优化过程转变为查表过程,EMPC可以显著提升交通信号控制的效率。

Urban traffic congestion becomes more and more severe and leads to the rapid growth of vehicles.As an effective approach,Model Predictive Control(MPC)is applied to control urban traffic networks.However,a key limitation of MPC is the potentially high online computational complexity,which limits its application in transportation.This paper introduces Explicit Model Predictive Control(EMPC)and designs a transportation signal controller based on the store-and-forward model for the traffic flow process in large scale urban traffic network.EMPC transfers the repeated online signal optimization computation to offline by multi-parametric Quadratic Programming(mp-QP),and provides an explicit optimal control law.The searching the lookup table online can obtain the explicit model control law for current states.The simulation experiments show that the EMPC approach is a breakthrough method to reduce the on-line computational complexity,and to increase the applicability of the EMPC in real-life traffic networks.