Integrated linear quadratic model predictive control for multi-destination traffic network
Yu Han, Yufei Yuan, Serge Paul Hoogendoorn · Transportation Research Board 94th Annual MeetingTransportation Research Board · 2015
Linear quadratic model predictive control (LQMPC) has been considered as an efficient method for real-time traffic network routing control and signal control. It is more computationally efficient than other model predictive control strategies due to its linear characteristics. However, in the past research, LQMPC with multi-destination cases has not been considered. To fill this gap the authors propose an extension of LQMPC in this paper, which can guarantee realistic splitting rates at bifurcations based on so-called “link” constraints and “destination” constraints. Moreover, the authors also implemented integrated control concept instead of isolated control strategy. Two case studies are investigated to test the performance of the proposed controller. The results show that the controller can not only guarantee realistic splitting rates of the network but also achieve good performance in terms of total throughput and total delay. The computation time of the optimization process also demonstrates that the approach is efficient enough for real-time traffic network control.