Improvements to Warm-Started Optimized Trajectory Planning for ASVs

Abstract: We present improvements to a recently developed method for trajectory planning for autonomous surface vehicles (ASVs) in terms of run time. The original method combines two types of planners: An A* implementation that quickly finds the global shortest piecewise linear path on a uniformly discretized map, and an optimal control-based trajectory planner which takes into account ASV dynamics. Firstly, we propose an improvement to the discretization of the map by switching to a Voronoi diagram rather than the uniform discretization, which offers a far more sparse search tree for the A* implementation. Secondly, modifications to the path refinement are made, as suggested in a paper by Bhattacharya and Gavrilova. The changes result in a reduction to the run time of the first part of the method of 85% for an example scenario while maintaining the same level of optimality.