A lexicographic hierarchical planning framework using motion primitives for autonomous vehicles

Trajectory planning in autonomous driving is critical to both vehicle safety and driving efficiency. Based on rule hierarchy, general approaches such as optimization-based planning and learning-based planning have been widely explored, yet some of them often suffer from poor interpretability, high computational complexity or insufficient real-time performance. To address these limitations, this paper proposes a lexicographic hierarchical planning framework that combines discrete motion primitive library with hierarchical constraint satisfaction. First, a motion primitive library is used to generate multiple candidate trajectories. Next, common highway driving rules, including collision avoidance, minimizing crossings of solid or dashed lane, enforcing trajectory feasibility and satisfying speed limits, are applied hierarchically to prune the candidate trajectory space. In addition, the framework can be enhanced by integrating with the Internet of Things (IoT) infrastructure, thereby enabling vehicles to utilize external data (such as environmental conditions, traffic conditions and facility signals) for more reasonable trajectory planning. Simulation results demonstrate that the proposed method guarantees safety with absolute priority and achieves real-time trajectory planning at roughly 20 Hz in highway and intersection scenarios. The project implementation is publicly available at: ( https://github.com/xiahr1989/HighwayEnvSimu ).