Abstract

Traffic congestion has become a significant issue in urban road networks. There have been massive works about traffic signal optimization to improve the efficiency of traffic flow operation, and the so-called back-pressure control policy has proven to be excellent for oversaturated conditions. Most of the existing works with back-pressure are based on an adaptive phase sequence, and research with cyclic phase sequence is based on calculating the splits for different phases using the traffic flow data at the beginning of each cycle, which is unfair for the non-initial phases. In this paper, we propose a decentralized model predictive signal control method with fixed phase sequence using back-pressure policy. The main idea of the new method is to form a control loop using the model predictive control, enabling the system to obtain real-time feedback from the traffic network and dynamically adjusting signal timing plans at the beginning of each phase. As links within a certain area have various lengths, the same queue length can imply different traffic conditions, so a method to normalize queue lengths is proposed. The normalized queue length decreases drastically when the actual length approaches link capacity, thus avoiding spillover. The proposed method was tested in a virtual road network. Numerical results suggest that the new method improves performance under congested conditions in terms of throughput, Gini coefficient and comprehensive transportation efficiency.

摘要

交通拥堵已成为城市道路网络中的一个重要问题。已经有大量关于交通信号优化以提高交通流运行效率的工作，所谓的背压控制策略已被证明是非常适合过饱和条件的。现有的背压工作大多基于自适应相序，循环相序的研究是基于每个循环开始时的交通流数据计算不同阶段的分裂，这对非- 初始阶段。在本文中，我们提出了一种使用背压策略的固定相序分散模型预测信号控制方法。新方法的主要思想是利用模型预测控制形成控制回路，使系统能够从交通网络获得实时反馈，并在每个阶段开始时动态调整信号配时计划。由于特定区域内的链路长度不同，相同的队列长度可能意味着不同的流量状况，因此提出了一种归一化队列长度的方法。当实际长度接近链路容量时，归一化队列长度急剧减少，从而避免溢出。所提出的方法在虚拟道路网络中进行了测试。数值结果表明，新方法在吞吐量、基尼系数和综合运输效率方面提高了拥堵条件下的性能。相同的队列长度可能意味着不同的交通状况，因此提出了一种归一化队列长度的方法。当实际长度接近链路容量时，归一化队列长度急剧减少，从而避免溢出。所提出的方法在虚拟道路网络中进行了测试。数值结果表明，新方法在吞吐量、基尼系数和综合运输效率方面提高了拥堵条件下的性能。相同的队列长度可能意味着不同的交通状况，因此提出了一种归一化队列长度的方法。当实际长度接近链路容量时，归一化队列长度急剧减少，从而避免溢出。所提出的方法在虚拟道路网络中进行了测试。数值结果表明，新方法在吞吐量、基尼系数和综合运输效率方面提高了拥堵条件下的性能。