Tuning perception and motion planning parameters for Moveit! framework

Abstract

This paper performs a benchmark of main parameters of perception and planning available in MoveIt! motion planning framework in order to identify parameters the most affecting the overall performance of the system. The initial benchmark is performed on a virtual simulation of UR3 robot workspace with a single obstacle. The performance is measured by means of successful runs, path planning and execution durations. The results of the benchmark are processed and, based on the results, three parameters are chosen to be optimized using Particle Swarm Optimization. The optimization of the parameters is performed for the same motion planning problem as presented in the first benchmark. In order to test the performance of the system with optimized parameters, four more benchmarks are performed using the simulated and real robot workspace. The results of the benchmarks indicate improvements in most of the measured indicators.