A multi-strategy improved A* algorithm for intelligent vehicle path planning based on dynamic adaptive weights and kinematic constraints

In this study, a multi-strategy improved A* algorithm based on dynamic adaptive weights and kinematic constraints is proposed, which can solve the problems of high collision risk, path jitter, and the difficulty of balancing search efficiency and optimality in traditional A* algorithms for intelligent vehicle path planning. The proposed multi-strategy improvement framework consists of four core optimization stages. Firstly, obstacle safety expansion preprocessing is implemented to ensure the safety of real vehicle driving, and an illegal heading blocking and steering penalty mechanism based on vehicle kinematic constraints is introduced during node expansion to generate initial paths that comply with the underlying control logic at the source of node expansion. Secondly, a dynamic adaptive weighting mechanism based on local obstacle density and target distance is designed, and a heuristic function combining Manhattan and Euclidean distance weighting is constructed to balance search efficiency and path optimality. In addition, a greedy line-of-sight algorithm and a cubic B-spline curve post-processing method are utilized to prune redundant nodes and convert discrete paths into continuous, high-order smooth trajectories. Finally, combined with the limit of centripetal acceleration, the algorithm completes the coupled feedforward planning of trajectory curvature and velocity to meet the physical motion limitations of real vehicles. To ensure its applicability in complex and unstructured scenarios, the algorithm is further validated in a real road environment. The results show that the proposed algorithm achieves a zero-collision target, significantly reduces the number of invalid expanded nodes, effectively mitigates path tortuosity, and generates high-quality trajectories strictly complying with real-world vehicle dynamics and safety standards.