Hierarchical hybrid path planning method based on improved sampling and potential field fusion

This paper addresses the trade-off challenge among safety margin, real-time performance, and trajectory trackability in autonomous vehicle path planning under complex environments. We propose a hierarchical hybrid path planning framework that combines global backbone planning (RRT-ADV/PRM-ADV/ACO) with local following (APF/DWA). The framework introduces goal biasing, dynamic step length, and curvature/angle constraints to balance global reachability and execution smoothness. Systematic experiments demonstrate that: (1) ablation study quantifies individual contributions of improvements, achieving 10.7%, 74%, and 318% enhancements in path length, smoothness, and safety clearance respectively; (2) multi-scenario experiments validate 100% success rate across narrow passages, U-shaped traps, and complex multi-obstacle environments; (3) parameter sensitivity analysis provides optimal configuration recommendations (p_goal=0.3, Δ=2.0m, θ_max=30°). The Hybrid-RRT achieves 4.7% and 20.2% improvements in path length and smoothness compared to single methods, providing a reproducible framework for engineering applications.