Research on vehicle trajectory planning algorithm integrating spatiotemporal constraints and adaptive curvature

The trajectory planning method based on the Frenet coordinate transformation effectively reduces computational complexity and enhances planning accuracy. However, in complex road scenarios involving multiple intelligent agents, the singular Frenet coordinate transformation is incapable of handling the variations in road curvature and vehicle posture among multiple target vehicles. This limitation leads to a reduced planning solution space, making it difficult to approximate the optimal path. To address this challenge, this study proposes a curvature-adaptive trajectory planning framework combined with spatiotemporal integration. This method involves three key innovations: (1) A road segmentation algorithm that discretizes complex curved roads into a set of quasi-straight segments through adaptive spatial discretization; (2) A multi-coordinate system adaptive transformation based on the predicted vehicle center in the temporal domain, which integrates temporal constraints with spatial Bézier curve optimization in a hierarchical planning architecture; (3) A multi-objective cost function is designed to comprehensively evaluate heuristic features such as path smoothness, efficiency, and safety, and candidate paths are optimized and ranked through weight configuration. Experimental results demonstrate that, compared with the benchmark method, the proposed framework achieves a 100% planning success rate in 135 complex road scenarios, with a 6.15% improvement in planning robustness and a 9.41% increase in trajectory smoothness. These findings confirm that the proposed framework has certain advantages in terms of smoothness and safety in complex road scenarios.