Abstract: To address the difficulty of simultaneously achieving trajectory smoothness and kinematic feasibility in single-layer planning methods for intelligent vehicle lane-changing scenarios, this paper proposes a hierarchical trajectory planning method based on fourth-order B-splines. The method adopts a hierarchical architecture of "geometric optimization, kinematic correction": first, a smooth reference trajectory is constructed using fourth-order B-spline curves, and the sequential least squares programming (SLSQP) algorithm is employed to optimize the control points, realizing spatial obstacle avoidance and geometric smoothness; subsequently, within the B-spline parameterization framework, multi-disk safety corridor constraints and a two-degree-of-freedom vehicle kinematic model are introduced to optimize the control sequence, ensuring that the trajectory satisfies vehicle kinematic constraints and comfort requirements. Building upon conventional hierarchical planning methods, the proposed approach converts the reference trajectory from the geometric layer into a bounded feasible region through a safety corridor generation mechanism, thereby endowing the motion layer with the ability to autonomously adjust the spatial trajectory within the corridor boundaries. Simulation results demonstrate that, compared with single-layer planning methods, the proposed method achieves superior trajectory smoothness and comfort performance in lane-changing obstacle-avoidance scenarios, while also improving computational efficiency by 61%, providing an effective technical solution for intelligent vehicle lane-changing and obstacle avoidance on structured roads.

Key words: intelligent vehicles; obstacle avoidance trajectory planning; hierarchical optimization; fourth-order B-spline curve; safety corridor

摘要： 针对智能车辆换道场景在单层规划方法中难以兼顾轨迹平滑性和运动学可行性的问题， 文章提出一种基于四阶 B 样条的分层轨迹规划方法。该方法以“几何寻优、运动学修正”为 分层架构：先利用四阶 B 样条曲线构建平滑参考轨迹，并通过序列最小二乘规划（SLSQP） 算法优化控制点，实现空间避障与几何平滑；再在 B 样条参数化框架下，引入多圆盘安全走 廊约束与车辆二自由度运动学模型来优化控制序列，确保轨迹满足车辆运动学约束与舒适性 要求。在传统分层规划方法基础上，文章基于几何层的参考轨迹，通过安全走廊生成机制将 其转化为有界可行域，使运动层在走廊边界内具备自主调整空间轨迹的能力。仿真结果表明， 与单层规划方法相比，该方法在换道避障场景中生成的轨迹平滑性和舒适性能表现更优。且 计算效率提升 61%，为结构化道路下智能车辆换道避障提供了有效的技术方案。

关键词: 智能车辆；避障轨迹规划；分层优化；四阶 B 样条曲线；安全走廊