A Hierarchical Framework for Dynamic Local Trajectory Planning

To enhance the mobility efficiency, safety, and ensure the dynamic feasibility of mobile robots, a local trajectory planning algorithm is proposed within complex two-dimensional environments. The method employs a hierarchical framework for trajectory planning – one layer densely maintains data and computes detailed path and velocity ranges offline, while another layer sparsely maintains data and selects specific trajectories online. This framework leverages the insight that obtaining all feasible trajectories during offline computation is most effective, and local trajectories are acquired by weighing scores in online scenarios. We recommend avoiding online searches to reduce computational complexity. This approach doesn't prioritize searching for the lowest-cost path (often the shortest path), but rather maximizes the likelihood of reaching the goal when determining the next navigational step. The method employs a two-layer linear programming algorithm to ascertain the maximum feasible velocity along the path, thereby generating dynamically feasible local trajectories. Ultimately, the effectiveness of this approach is validated through simulation experiments.