Model Predictive Control-Based Optimal Speed Control for Connected and Autonomous Vehicles in Signalized Road Networks

Abstract

Signalized road networks are among the most challenging driving environments in urban transportation systems, often causing driver delays, congestion, and safety risks. To address these challenges, we propose a predictive vehicle control strategy tailored for connected and autonomous vehicles (referred to as MPC-CAV) using vehicle-to-infrastructure (V2I) communication. MPC-CAV uses the model predictive control (MPC) framework that incorporates traffic light signal information. Furthermore, by considering the dilemma regions while taking into account the driver’s reaction and the vehicle’s braking ability, an efficient velocity trajectory is generated. Based on this generated velocity trajectory, the CAVs optimize the travel time crossing the intersection and smooth driving. Hence, the proposed strategy enhances travel efficiency and mitigates congestion, while ensuring strict compliance with speed limit constraints. To emphasize the superior performance of our proposed approach (i.e., MPC-CAV), we execute a detailed comparison with some existing methods, i.e., long-sighted CAVs, short-sighted CAVs, and Krauss methods, in various scenarios. All comparative simulations are executed and verified using a co-simulation platform between SUMO (i.e., high-fidelity traffic management software) and Python. The results indicate that the MPC-CAV achieves the shortest waiting time, reducing waiting time by over 50% to 70% compared to other vehicles. Additionally, MPC-CAV, short-sighted CAV, and Krauss methods strictly comply with road speed limit requirements, while the long-sighted CAV violates the speed limit at most green light stages. The proposed method is further emphasized by its ability to strictly obey the traffic light without red light violations and achieve precise deceleration to halt at the designated stop line.