Autonomous vehicles – path planning and trajectory optimization
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
NHCE
Abstract
Autonomous vehicles (AVs) represent a transformative technology in intelligent transportation systems, relying heavily on the interplay of path planning and trajectory optimization to navigate complex and dynamic environments safely, smoothly, and efficiently. Path planning determines a feasible route from a starting point to a destination while accounting for obstacles, traffic regulations, road constraints, and environmental dynamics. Trajectory optimization refines this planned route by defining precise vehicle motions, such as speed, steering angle, and acceleration, to ensure safe, smooth, and energy-efficient execution that respects the vehicle's kinematic and dynamic constraints. These processes are interdependent, with path planning providing a high-level route and trajectory optimization ensuring that the route can be executed in real time. Autonomous vehicles rely on a range of sensors, including LiDAR, cameras, radar, GPS, and inertial measurement units (IMUs), to perceive and understand their surroundings. The data gathered is continuously processed to generate an accurate environment map, which informs both path and trajectory planning. Algorithms such as A* and Dijkstra’s are commonly employed for global path planning to compute optimal routes on high-level maps. Locally, machine learning techniques, such as behavioral cloning, reinforcement learning, and imitation learning, are used to adaptively refine trajectories and improve decision-making. Behavioral cloning allows AVs to emulate human-like driving by learning from human driving data, enabling natural and intuitive navigation even in challenging scenarios such as urban intersections or minimal-map environments.