Trajectory Optimization with Dynamic Drivable Corridor-Based Collision Avoidance

Trajectory planning for autonomous vehicles is essential for ensuring driving safety, passenger comfort, and operational efficiency. Collision avoidance constraints introduce significant computational complexity due to their inherent non-convex and nonlinear characteristics. Previous research has proposed the drivable corridor (DC) method, which transforms complex collision avoidance constraints into linear inequalities by constructing time-varying rectangular corridors within the spatiotemporal domains, thereby enhancing optimization efficiency. However, the DC construction process involves repetitive collision detection, leading to an increased computational burden. To address this limitation, this study proposes a novel approach that integrates grid-based obstacle representation with dynamic grid merging to accelerate collision detection and dynamically constructs the DC by adaptively adjusting the expansion strategies according to available spatial dimensions. The feasibility and effectiveness of the proposed method are validated through simulation-based evaluations conducted over 100 representative scenarios characterized by diverse and unstructured environmental configurations. The simulation results indicate that, with appropriately selected grid resolutions, the proposed approach achieves up to a 60% reduction in trajectory planning time compared to conventional DC-based planners while maintaining robust performance in complex environments.