Trajectory Planning for Automated Driving in Intersection Scenarios using Driver Models
This work addresses a critical problem for autonomous driving systems in complex urban environments, representing an incremental improvement in trajectory planning methods.
The paper tackles efficient trajectory planning for autonomous vehicles at urban intersections by proposing a novel framework that ensures social compliance and optimizes comfort under kinematic constraints, achieving fast behavior prediction and decision-making with demonstrated capabilities in various scenarios.
Efficient trajectory planning for urban intersections is currently one of the most challenging tasks for an Autonomous Vehicle (AV). Courteous behavior towards other traffic participants, the AV's comfort and its progression in the environment are the key aspects that determine the performance of trajectory planning algorithms. To capture these aspects, we propose a novel trajectory planning framework that ensures social compliance and simultaneously optimizes the AV's comfort subject to kinematic constraints. The framework combines a local continuous optimization approach and an efficient driver model to ensure fast behavior prediction, maneuver generation and decision making over long horizons. The proposed framework is evaluated in different scenarios to demonstrate its capabilities in terms of the resulting trajectories and runtime.