REAL-TIME TRAJECTORY OPTIMIZATION FOR AUTONOMOUS VEHICLE LANE-CHANGING AND LANE-KEEPING USING MODEL PREDICTIVE PATH INTEGRAL

The advent of autonomous vehicles marks a revolutionizing transformation in transportation, with the potential to enhance safety and efficiency through advanced trajectory planning and optimization capabilities. Model predictive path integral (MPPI) algorithms play a significant role in achieving these benefits by providing real-time vehicle trajectory optimization, allowing for precise control in dynamic environments. Despite its advantages, existing research has not extensively explored the effects of different prediction horizon times on MPPI control performance. This paper addresses this gap by proposing a multi-input MPPI control method for the autonomous vehicle using a single-track vehicle dynamic model and investigate the influence of various prediction horizon times on trajectory optimization for lane-changing and lane-keeping maneuvers. The simulation results demonstrate that an appropriate prediction horizon time helps in managing the vehicle movements, leading to a stable and controllable transition from lane-changing to lane-keeping. These findings provide valuable insights for optimizing MPPI to enhance the safety and efficiency of autonomous driving.