Safe and efficient trajectory optimization for autonomous vehicles using b-spline with incremental path flattening.

🧾 Introduction

• B-spline-based trajectory optimization is effective in robotics due to its computational efficiency and convex-hull property, especially for quadrotors.
• However, applying B-splines to autonomous vehicles (AVs) is challenging due to rectangular body shapes and kinodynamic constraints.
• This study proposes a novel approach that overcomes these challenges for AV trajectory optimization.

⚙️ Method

• The proposed method integrates three components: Incremental Path Flattening (IPF), a disc-type swept volume (SV) estimation, and kinodynamic feasibility constraints.
• IPF flattens paths by increasing curvature penalties around collision areas, enabling collision-free path shaping.
• A disc-type SV model reduces over-approximation and allows efficient navigation in narrow spaces. Clamped B-spline curvature constraints are introduced to ensure kinodynamic feasibility, including limits on velocity and acceleration.

✅ Result

• In simulation, the proposed method outperformed state-of-the-art baselines in path safety and efficiency.
• Real-world experiments with an AV confirmed that the method’s tracking performance aligned with simulation results.
• The approach proved effective in handling both collision avoidance and dynamic feasibility in constrained driving environments.