Improved RRT-based Trajectory Planning for Connected Autonomous Vehicles on Urban Roads

In response to the challenges of high computational costs, prolonged time consumption, and difficulty in incorporating traffic rules of standard trajectory planning algorithms. This paper proposes an improved rapidly-exploring-random-trees (RRT)-based trajectory planning method for connected autonomous vehicles (CAVs) in urban roads through utilizing real-time traffic information from vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. The proposed method does not require precise modeling and possesses fast planning speed and good applicability like RRT. It also removes over curves of the paths generated and satisfies traffic rules by introducing a variable step-size adjustment strategy, a heuristic search method, and a cost function incorporating driver behavior constraints. The generated trajectory is further optimized using redundant node removal and spline interpolation. Then, through lots of simulation experiments under various driving scenarios, the proposed method reduces the average planning time by 39.36% and shortens the average trajectory length by 12.59% compared to the classical RRT algorithm. Finally, the proposed algorithm's effectiveness in planning under dynamic obstacles and actual vehicle conditions was also validated.