Development of fixed-route autonomous driving for light vehicles with smart steering and energy saving

Abstract

With the rapid development of autonomous driving technology, its potential to improve traffic safety, efficiency, and energy saving is gradually emerging. As the core of the autonomous driving system, vehicle tracking and control has a direct impact on reducing the energy consumption of autonomous vehicles (AVs). The intelligent steering and control system algorithm based on global optimization proposed in this work can accurately track the trajectory generated by automatic driving and reduce energy consumption. A complete autonomous driving system includes mapping, positioning, sensing, routing, path planning, and trajectory tracking. Among these, tracking controls the driving speed and steering angle but also affects the vehicle's energy consumption. Thus, a stable and robust tracking algorithm is significant for the safety of AVs. This work first designs a two-layer model predictive control algorithm, combining vehicle kinematics and dynamics models. Compared with the traditional control algorithm, it includes vehicle motion characteristics and a speed generate module, eliminating the speed planning module and significantly saving computing resources. For autonomous driving design, sharing lanes with human-driven vehicles is the future, so a driving style close to human driving preferences to ensure safety is essential. In this work, the control quantity in line with the driving preference can be generated based on the optimization of human-driver trajectory to ensure driving safety. Meanwhile, the output efficiency of the motor is considered in the optimization process, and the motor always has a high-efficiency output during operation. It cannot only save energy but also improve the service life of the motor. In the urban environment, the primary consideration for autonomous driving is the lateral deviation of the vehicle. However, due to environmental changes and road aging, the friction coefficient of the road surface will change, which also leads to the insufficient force provided by the road surface in the process of high-speed driving, which quickly causes a rollover. Therefore, how to reduce the external disturbance introduced by the change in the road friction coefficient is very important. At the same time, the change in the vehicle's mass will also introduce internal disturbance, affecting the controller's performance. As the actual values of the vehicle and the external environment cannot be fully obtained, and the established dynamic model also has nonlinear and discretization, the tracking accuracy will be significantly reduced, especially in high-speed motion. Therefore, this work proposes to introduce the reinforcement learning controller of adaptive soft actor-critic (ASAC), which can accurately model the vehicle through training, to resist internal and external disturbances. At the same time, there has also been a good improvement in motor efficiency output. The main achievement of this work is to establish a robust automatic driving control system that can suppress external and internal disturbances. By optimizing human trajectory, generating control quantity in line with the driving preference of human driver, and improving the output efficiency of the motor while driving, tracking performance and energy saving to the maximum extent is guaranteed.