Adaptive dynamic traffic control of urban networks : a macroscopic fundamental diagram approach

Abstract

Urbanization has induced dramatic growth in car usage in metropolises around the world, which results in growing traffic congestion, accidents and pollution. Efficient utilization of existing infrastructures via appropriate traffic control schemes is crucial to handling the fast-growing travel demand. Conventional traffic control methods concentrate on link-level strategies. Oversaturated traffic conditions with queues spilling back to upstream links and the huge spatial dimension would introduce significant challenges to the local traffic signal control strategies at the link level. Hence, under heavily saturated traffic conditions, traffic control strategies capturing network-level congestion should be devised to alleviate network congestion. The network-level congestion can be significantly alleviated by identifying some critical intersections and regulating them effectively. This finding gives rise to the concept of perimeter control by leveraging the recent advances in the macroscopic fundamental diagrams (MFDs). The MFD intuitively describes a low-scatter rela­tionship between the network vehicle accumulation and production, providing an analytically simple and computationally efficient framework for aggregate model­ing of urban traffic network dynamics. Therefore, this dissertation proposes an MFD-based optimal control framework for traffic networks. Perimeter control, which aims to manipulate the transfer flow at the boundaries of the region, is a promising solution to address the spatial dimension challenge in dealing with network-scale traffic congestion. Existing MFD-based data-driven and feedback perimeter control strategies do not consider the heterogeneity of real-time data measurements. Besides, traditional reinforcement learning (RL) methods for traffic control usually converge slowly for lacking data efficiency. Moreover, conventional optimal perimeter control schemes require exact knowledge of the system dynamics and thus they would be fragile to endogenous uncertainties. To handle these challenges, Study 1 proposes an integral reinforcement learning (IRL) based approach to learning the macroscopic traffic dynamics for adaptive optimal perimeter control. A continuous-time control is developed with discrete gain updates to adapt to the discrete-time sensor data. Different from the conventional RL approaches, the reinforcement interval of the proposed IRL method can be varying with respect to the real-time resolution of data measurements. To reduce the sampling complexity and use the available data more efficiently, the experience replay (ER) technique is introduced to the IRL algorithm. The proposed method relaxes the requirement on model calibration in a model-free manner that enables robustness against modeling uncertainty and enhances the real-time performance via a data-driven RL algorithm. Numerical examples and simulation experiments are presented to verify the effectiveness and efficiency of the proposed method. Considering the time-varying nature of the travel demand pattern and the equilib­rium of the accumulation state, Study 2 extends the set-point perimeter control (SPC) problem investigated in Study 1 to an optimal tracking perimeter control problem. Unlike the SPC schemes that stabilize the traffic dynamics to the desired equilibrium point, the proposed tracking perimeter control (TPC) scheme will regulate the traffic dynamics to a desired trajectory in a differential framework. Study 2 proposes an adaptive dynamic programming (ADP) approach to solving the optimal TPC problem. The convergence of the ADP based algorithms and the stability of the controlled traffic dynamics are proven via the Lyapunov theory. Numerical experiments are performed to demonstrate the effectiveness of the proposed ADP-based TPC. Com­pared with the SPC scheme, the proposed TPC scheme achieves both improvements in reducing total travel delay and increasing cumulative trip completion in our case studies. Coupling perimeter control and regional route guidance (PCRG) is a promising strategy to decrease congestion heterogeneity and reduce delays in large-scale MFD-based urban networks. For MFD-based PCRG, one needs to distinguish between the dynamics of the plant that represents reality and is used as the simulation tool, and the model that contains easier-to-measure states than the plant and is used for devising controllers, i.e., the model-plant mismatch should be considered. Traditional model-based methods require an accurate representation of the plant dynamics as the prediction model. On the other hand, existing data-driven methods do not consider the model-plant mismatch and the limited access to plant-generated data. Therefore, Study 3 develops an iterative adaptive dynamic programming (IADP) based method to address the limited data source induced by the model-plant mismatch. An actor-critic neural network structure is developed to circumvent the requirement of complete information on plant dynamics. Performance comparisons with other PCRG schemes under various scenarios are carried out. The numerical results indicate that the IADP controller trained with a limited data source can achieve comparable performance in minimizing the total travel delay with the benchmark model predictive control (MPC) approach using perfect measurements from the plant. In cases of higher input errors, IADP achieves a better performance than MPC. Most existing studies on optimal traffic control of MFD-based networks do not consider the effect of expressways passing through urban regions. Ring expressways are built in many megacities (e.g., Beijing) with on-and off-ramps to connect the city’s periphery areas where ramp metering is usually desired to protect the freeways from over congestion. Few studies have explored the cooperation of perimeter control, route guidance and ramp metering strategies in improving the whole network mobility. Study 4 proposes a cooperative adaptive dynamic programming (CADP) approach to solve the cooperative control problem for a mixed urban-expressway network. The network is composed of a multi-region urban network modeled by the MFD and a ring expressway going through the periphery regions modeled by the asymmetric cell transmission model. Different from the traditional decentralized ADP (D-ADP) method, the proposed CADP approach trains the agents of perimeter control, route guidance, and ramp metering to fully cooperate in improving the whole network performance. Numerical studies demonstrate that the CADP can significantly reduce the total travel delay compared with the model-based decentralized strategies and the D-ADP strategy. In addition, the city center is well protected from over-congestion by applying the CADP approach. In conclusion, this thesis contributes to the literature on network-level optimal perimeter control and regional route guidance, and to traffic management of mixed urban-expressway networks.