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

Zhao, Y; Wang, J; Li, P; Wang, H · 2025 · PubMed Central

DOI: 10.1038/s41598-025-28018-1

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the limitations of existing vehicle trajectory planning algorithms in complex, multi-agent road scenarios, particularly those involving significant curvature. Traditional methods relying on a singular Frenet coordinate system often fail to accurately handle variations in road curvature and vehicle posture, leading to a reduced solution space and an inability to approximate optimal paths. Furthermore, conventional decoupled spatiotemporal planning often results in conservative or aggressive maneuvers that compromise safety and efficiency. To resolve these issues, the authors propose a curvature-adaptive trajectory planning framework that integrates spatiotemporal constraints with adaptive spatial discretization. The proposed method employs a three-stage approach. First, it utilizes an adaptive curvature-based discretization algorithm to segment complex curved roads into quasi-straight segments, establishing local Frenet coordinate systems for each segment based on the vehicle's center. This "straightens" the road geometry, allowing for smoother trajectory planning. Second, the framework constructs a dynamic feasible region in a three-dimensional Station-Lateral-Time (S-L-T) space, modeling obstacles as convex polyhedral constraints. Third, it generates trajectories using piecewise Bézier curves optimized via quadratic programming. This optimization balances multiple objectives, including path smoothness, efficiency, and safety, while enforcing C2 continuity constraints at segment connections to ensure global smoothness. The algorithm operates within a local solution space derived from decision-making layers, avoiding computational complexity explosions associated with high-dimensional state spaces. Experimental results demonstrate the efficacy of the proposed framework across 135 complex road scenarios. The algorithm achieved a 100% planning success rate, outperforming benchmark methods. Specifically, the framework showed a 6.15% improvement in planning robustness and a 9.41% increase in trajectory smoothness compared to traditional single-Frenet-frame approaches. Additionally, the divide-and-conquer strategy enabled parallel computation of subproblems, controlling planning time within 0.21 seconds, which meets real-time performance requirements. The method also reduced the trajectory envelopment radius, enhancing safety margins. The significance of this research lies in its ability to provide a robust, real-time solution for autonomous driving trajectory planning in challenging environments. By effectively decoupling the constraints of complex road curves and vehicle posture, the algorithm improves both the safety and naturalness of driving behaviors. The integration of heuristic human-like decision-making with rigorous mathematical optimization offers a novel approach to balancing computational efficiency with planning accuracy, addressing critical gaps in current intelligent driving systems regarding generalization and robustness in dynamic, curved-road scenarios.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• steering pattern