TY - JOUR AU - Zhang,R. AU - Guo,H. AU - Sotelo,M.A. AU - Du,H. AU - Darius,A. AU - Li,Z. KW - artificial potential field KW - Autonomous vehicles KW - Frenet coordinate system KW - path planning KW - RRT algorithm T1 - New RRT-Based Method for Vehicle Path Planning in Curve Scenarios Considering Path Oscillations LA - eng PY - 2025/// SP - 18445 EP - 18459 T2 - IEEE Transactions on Vehicular Technology SN - 1939-9359 VL - 74 IS - 12 PB - Institute of Electrical and Electronics Engineers Inc. AB - This paper proposes a bidirectional RRT* algorithm integrated with dynamic artificial potential fields (APF) in the Frenet coordinate system to address obstacle avoidance challenges for autonomous vehicles in curved road scenarios. By decoupling longitudinal (s) and lateral (d) motion, the Frenet framework simplifies road constraints while reducing search space dimensionality by 70%. The APF mechanism dynamically balances obstacle repulsion, reference trajectory attraction, and goal-driven forces through an S-shaped repulsion function, suppressing path oscillations and improving computational efficiency. Theoretical analysis proves probabilistic completeness and reduced time complexity (O(m2 log m) vs. O(n2 log n)). Experimental validation demonstrates adherence to Ackermann steering constraints (κmax ≤ 0.25 m−1), achieving a 15.9% reduction in path length compared to RRT*, 17.4% faster computation than DWA, and 100% obstacle avoidance success. Real-world tests on U-turn and multi-curve roads validate trajectory stability with mean curvature 0.06 m−1 (45% improvement over RRT*) and sub-40 ms planning cycles, enabling real-time navigation in safety-critical environments. DO - 10.1109/TVT.2025.3585155 UR - https://portalcientifico.uah.es/documentos/6880a3cd3871d04c4eb49ec6 DP - Dialnet - Portal de la Investigación ER -