Trajectory Guidance for Connected Human-Driving Vehicles through the Interactions between Drivers and Roadside Units

Abstract

Utilizing massive real-time traffic information in the vehicle-to-everything (V2X) environment, road traffic systems can be enhanced by optimizing vehicle trajectory patterns. Because intelligent decisions can be made by roadside units (RSUs) with multiaccess edge computing (MEC) devices, this paper presents a trajectory guidance method for connected human-driving vehicles (CHVs) based on human–RSU interactions. Optimal guidance commands were determined based on a trajectory predictive control method, helping drivers operate the vehicles to follow the expected trajectories. We utilized the Gaussian mixture model to analyze the naturalistic driving data set collected by the project of the Next Generation Simulation (NGSIM) and determine the acceleration distributions of different guidance commands, including decelerate rapidly, decelerate slowly, keep velocity, accelerate slowly, and accelerate rapidly. The Monte Carlo sampling method was used to simulate different acceleration choices for command-based guidance information, considering human driver uncertainty. Sensitivity analysis was conducted to evaluate the performance of the proposed trajectory guidance method with different parameters. Experimental results showed that the average trajectory deviations at all positions are less than 5 m, indicating that guidance performance with reasonable guidance parameters is acceptable. Therefore, the proposed trajectory guidance method by human–RSU interaction can effectively support CHVs participating in V2X cooperation and has good practical application prospects.