Connected Automated Vehicle Trajectory Planning at a Signalized Intersection Considering Tunable Predicted Trajectories of Human-Driven Vehicles

Given the anticipated prolonged coexistence of connected automated vehicles (CAVs) and human-driven vehicles (HDVs), it is imperative to study how to control CAVs in the mixed flow. In this paper, a dynamic trajectory planning method for CAVs at a signalized intersection is proposed. This method takes into account the stochasticity of HDVs and designs a tunable trajectory prediction method with higher prediction accuracy, thereby further reducing travel time and fuel consumption, improving intersection throughput and driving comfort. Building on the concept of α-trajectory proposed previously, we extend its application to improve the accuracy of the predicted trajectories of HDVs by treating α not as an exogenous, fixed value, but as an endogenous, tunable variable. The proposed method consists of three parts: information collection, tuning of α, and trajectory planning. Information collection involves gathering and transmitting data from the traffic system to CAVs using various devices. The tuning part is devoted to determining the proper tune time and the optimal values of α through specific criteria and a tuning model. The trajectory planning part then predicts the trajectories of HDVs more accurately based on the tuned α, incorporating a mixed integer programming model to optimize CAV accelerations. Numerical experiments in various scenarios show that the proposed method is well-adapted and outperforms the benchmark method with an exogenous and fixed value of α.