A non-linear optimal strategy for simultaneous planning and tracking of autonomous vehicles

Abstract

Motion planning and control are critical for autonomous vehicles, enabling trajectory generation and control to meet defined objectives and constraints. However, traditional methods often rely on pre-defined optimised trajectories, which are computationally expensive to generate, and hierarchical schemes, which introduce delays. While many methods exist, the use of a perpendicular line-based reference for trajectory generation remains underexplored. This study introduces a novel controller under a simultaneous planning and tracking framework named the Perpendicular Line-Ahead Planner Controller (PLAPC). Using the Non-linear Model Predictive Control (NMPC) technique, the PLAPC simultaneously generates trajectories and performs control in real-time, without requiring pre-computed trajectories. The controller tracks predefined velocity references while adjusting control inputs to meet various objectives such as minimising distance to the reference line ahead, optimising comfort and improving energy efficiency by incorporating Linear-Exponential (LINEX) loss function terms. This approach integrates simultaneous trajectory planning, velocity tracking, and control into a unified framework.

The PLAPC was evaluated through high-fidelity simulations using MATLAB Simulink, CasADi symbolic tools, and IPG CarMaker software. Initial validation involved a comparison with a traditional path-tracking controller (PTC) and was further tested across diverse path shapes, distances, and velocity profiles. The results demonstrated that the PLAPC dynamically generates trajectories in real time and achieves superior velocity tracking, even with a front-wheel steering (FWS) vehicle topology as a baseline. The framework's versatility and computational efficiency highlight its adaptability and practicality for autonomous vehicle systems, significantly contributing to the advancement of autonomous vehicle motion planning and control strategies.