Trajectory Planning and Tracking Control of Autonomous Vehicles Based on Improved Artificial Potential Field

At present, trajectory planning and tracking control methods for autonomous vehicles (AVs) are complicated and mainly suitable for specific scenarios. In this paper, based on vehicle dynamics model, a novel approach of cleverly introducing the velocity-related safe distance into the artificial potential field function is proposed. The improved traffic environment potential field is then combined with model predictive control (MPC), and lateral stability constraints such as lateral load transfer rate are added to the MPC when solving for the control quantity. We name this method AutoPField-MPC, and it offers the following advantages: First, it allows AVs to automatically adjust the distance from other obstacles in more complex scenarios, while also dynamically influencing steering and braking time. Second, it reduces extensive parameter adjustment work when considering the safe distance. Third, the improved design of the potential field function in this paper makes the method more concise and universal. The simulation results from CarSim and Simulink demonstrate that the method presented in this paper can simultaneously achieve multiple active safety objectives, such as speed maintenance, self-adaptive distance adjustment, lane keeping, lateral stability, emergency braking, and emergency steering, with good real-time performance. As a result, it can effectively address the security issues of vehicles and pedestrians when AVs avoid obstacles in various complex scenarios, such as wet roads, multi-obstacle vehicles, and emergency obstacle avoidance.

I. Introduction

With the rapid development of autonomous driving technology, people are placing greater emphasis on active safety technologies for vehicles, which involve taking proactive measures to mitigate safety risks and prevent traffic accidents [1], [2]. The risks associated with driving vehicles are diverse [3], [4], [5], [6], [7], [8], [9], [10], [11], [12]: First of all, high-speed vehicles are prone to rear-end, side-slips, rollovers and lane departures, especially in the case of rain, snow or accidents; Secondly, in a complex traffic environment with numerous obstacle vehicles and instances of illegal driving by vehicles in front, inadequate analysis of the dynamic traffic environment may lead to collisions with other vehicles; Third, collisions may occur when pedestrians cross the road illegally and vehicles are unable to avoid them in time, which may result in serious injuries or even fatalities. Currently, autonomous driving technology can be classified into rule-based approaches, data-driven methods, or a combination of both.