Data-driven Trajectory Planning Strategy for Connected Vehicles at Signalized Intersection

This paper presents a data-driven trajectory planning strategy for Connected and Automated Vehicles (CAVs), which can ensure probabilistic collision avoidance and improve the fuel economy along signalized corridors. First, a non-parametric regression (Gaussian Process Regression) is built based on the historical data to describe the uncertain relative distance between the preceding and host vehicle. Then the optimal control problem is formulated ans solved by Receding Horizon Control (RHC) framework, the probabilistic constraint is transformed into a deterministic constraint within a shorter control interval. At last, the results from the numerical simulation using the NGSIM (Next Generation SIMulation) data set show the proposed method's efficacy in improving the fuel economy, with a maximum fuel consumption reduction of 33.97% and a minimum reduction of 2.21%.