Frenet Corridor Planner: An Optimal Local Path Planning Framework for Autonomous Driving

Abstract: Motivated by the requirements for effectiveness and efficiency, path-speed decomposition-based trajectory planning methods have widely been adopted for autonomous driving applications. While a global route can be pre-computed offline, real-time generation of adaptive local paths remains crucial. Therefore, we present the Frenet Corridor Planner (FCP), an optimization-based local path planning strategy for autonomous driving that ensures smooth and safe navigation around obstacles. Modeling the vehicles as safety-augmented bounding boxes and pedestrians as convex hulls in the Frenet space, our approach defines a drivable corridor by determining the appropriate deviation side for static obstacles. Thereafter, a modified space-domain bicycle kinematics model enables path optimization for smoothness, boundary clearance, and dynamic obstacle risk minimization. The optimized path is then passed to a speed planner to generate the final trajectory. We validate FCP through extensive simulations and real-world hardware experiments, demonstrating its efficiency and effectiveness.

Overview of the Frenet Corridor Planner for Autonomous Driving

The paper introduces the Frenet Corridor Planner (FCP), an optimal local path planning framework specifically designed for autonomous driving applications. The authors recognize the necessity for real-time adaptive local path generation that aligns with pre-computed global routes to ensure effective and efficient autonomous navigation. They emphasize the importance of online path planning capable of addressing static and dynamic obstacles in real-time scenarios.

Methodology

The FCP is formulated as an optimization-based local path planning strategy grounded in path-speed decomposition. This framework consists of two principal components: the path and speed planners, which operate in tandem to yield computational efficiency. The core algorithm utilizes the Frenet coordinate system. By transforming vehicle and obstacle states into the Frenet space, FCP decomposes the planning problem into two dimensions: longitudinal ($s$-axis) and lateral ($d$-axis) displacements relative to a reference path.

Key Components

1. Modeling of Environment:

   • Vehicles are represented as safety-augmented bounding boxes, while pedestrian clusters are treated as convex hulls in Frenet space. This representation aids in defining a drivable corridor where static obstacles are strategically bypassed by determining deviation sides algorithmically.

2. Optimization Framework:

   • A modified space-domain bicycle kinematic model governs the path optimization, optimizing for path smoothness, boundary clearance, and minimizing collision risk with dynamic obstacles. These pathways are derived by solving an optimization problem with constraints reflecting boundary limits and kinematic feasibility.

3. Simulation and Experimentation:

   • The paper details the validation of FCP via both simulations and real-world hardware tests, underscoring its robustness and efficacy. The FCP outperforms traditional methods such as A$^\star$ and RRT$^\star$ in computational efficiency and path smoothness, as evidenced by quantitative metrics such as algorithm runtime, angular acceleration, and clearance from obstacles.

Implications and Future Work

The research underscores the significance of reliable local path planning in the broader context of autonomous vehicle control architectures. Practically, the FCP provides a robust solution for navigating dense and dynamic urban environments. Theoretically, this work shows potential pathways for refining kinematic models and obstacle representation techniques in the Frenet frame.

Future research may explore integrating more advanced machine learning-based decision-making models within the framework to further enhance its adaptability and robustness against uncertainties. Moreover, expanding FCP to accommodate an even broader range of dynamic scenarios and terrains could widen its applicability. Additionally, the investigation into reducing the non-convexity of the optimization problem to further enhance computational efficiency remains a promising avenue for exploration.