B-spline-based trajectory optimization is widely used for robot navigation due to its computational efficiency and convex-hull property (ensures dynamic feasibility), especially as quadrotors, which have circular body shapes (enable efficient movement) and freedom to move each axis (enables convex-hull property utilization). However, using the B-spline curve for trajectory optimization is challenging for autonomous vehicles (AVs) because of their vehicle kinodynamics (rectangular body shapes and constraints to move each axis). In this study, we propose a novel trajectory optimization approach for AVs to circumvent this difficulty using an incremental path flattening (IPF), a disc type swept volume (SV) estimation method, and kinodynamic feasibility constraints. IPF is a new method that can find a collision-free path for AVs by flattening path and reducing SV using iteratively increasing curvature penalty around vehicle collision points. Additionally, we develop a disc type SV estimation method to reduce SV over-approximation and enable AVs to pass through a narrow corridor efficiently. Furthermore, a clamped B-spline curvature constraint, which simplifies a B-spline curvature constraint, is added to dynamical feasibility constraints (e.g., velocity and acceleration) for obtaining the kinodynamic feasibility constraints. Our experimental results demonstrate that our method outperforms state-of-the-art baselines in various simulated environments. We also conducted a real-world experiment using an AV, and our results validate the simulated tracking performance of the proposed approach.

翻译：基于B样条的轨迹优化因其计算效率与凸包特性（确保动力学可行性）被广泛用于机器人导航领域，尤其适用于具有圆形机体结构（便于高效运动）且各轴可独立自由运动（可充分利用凸包特性）的四旋翼飞行器。然而，由于自动驾驶车辆存在特殊运动学约束（矩形车身结构与各轴运动受限），将B样条曲线直接应用于该类车辆的轨迹优化面临显著挑战。本研究提出一种新型自动驾驶车辆轨迹优化方法，通过引入渐进路径展平技术、圆盘型扫掠体体积估计算法及运动学可行性约束来克服上述困难。其中，渐进路径展平技术通过迭代增加车辆碰撞点附近的曲率惩罚项，实现路径展平与扫掠体体积缩减，进而获得无碰撞路径。同时，我们构建了圆盘型扫掠体体积估计算法，通过降低扫掠体积的过度近似误差，使自动驾驶车辆能高效通过狭窄通道。此外，为构建运动学可行性约束条件，本研究在动力学可行性约束（如速度与加速度约束）基础上，引入夹持型B样条曲率约束（简化了B样条曲率约束形式）。实验结果表明，在多种仿真环境下，本方法均优于现有最优基线算法。我们进一步通过真实自动驾驶车辆实验验证了所提方法在仿真环境中展现的轨迹跟踪性能。