An open-space planner.

🧾 Introduction

• The Open Space Planner (OSP) module was developed for autonomous path planning in unstructured environments such as parking lots or areas without lane markings.
• It is based on the method proposed in “Autonomous Driving Trajectory Optimization With Dual-Loop Iterative Anchoring Path Smoothing and Piecewise-Jerk Speed Optimization”.
• The OSP was tested in both the CARLA simulator and real-world vehicle experiments.

⚙️ Method

• Hybrid A* was used to generate an initial feasible path considering vehicle kinematics.
• The path was then optimized using Sequential Quadratic Programming (SQP) to minimize jerk and acceleration, resulting in a smooth and dynamically feasible path.
• To ensure probabilistic completeness, a kinodynamic informed RRT* was implemented as a backup planner.

✅ Result

• The OSP module successfully generated collision-free, smooth, and dynamically feasible path in both simulation and real-world settings.
• The hybrid planning structure allowed robust handling of complex environments, with RRT* ensuring fallback coverage when optimization failed.
• The resulting path demonstrated low jerk and acceleration profiles, enhancing ride comfort and vehicle control performance.