Evaluation Of The ARTIS Sampling-based Path Planner Using an Obstacle Field Navigation Benchmark

Abstract

This work presents performance assessments of a sampling-based motion planner for path planning and guidance in urban terrain with sensing unforeseen obstacles in the loop. An existing benchmark suite is utilized for this purpose, in order to enable comparison with other approaches. Moreover this benchmark is used to determine how close generated paths match the provided baseline solutions to which other solutions can be compared with. The evaluation results indicate an efficient replanning of our online multi-query planning approach. Furthermore conclusions from variations in the experiments with respect to sensor field of view, sample density, replanning rate and motion safety aspects help to understand practical aspects of our roadmap-based planning and guidance method.