A hierarchical trajectory planning framework for autonomous driving

In this paper, we introduce a layered trajectory planning method for autonomous vehicles. The trajectory planning problem is decomposed into three parts: path planning, speed planning, and path-speed iteration. In the path planning section, we combine the advantages of search-based and optimization based methods. Firstly, we use different search-based methods for safe navigation and then smooth the searched path to get a high-quality reference path. Secondly, the path planning problem under the Frenet frame is formulated as a quadratic program (QP), which can be solved efficiently. Then a speed profile is generated based on the planned path. Furthermore, we propose an iterative strategy for path planning and speed planning to ensure the feasibility of the planned trajectory. Our algorithm is implemented in C++, and the path planning part is open-sourced on our GitHub page 1.1Please visit https://github.com/LiJiangnanBit/path-optimizer to get the source code.