Benchmarking framework for robotic arc welding motion planning

Abstract Producing a small series of large complex parts using robotic arc welding is challenging due to the time it takes to program the robot. Currently, offline programming software assists the operator in programming hundreds of welds. This is often still an iterative process because the welding robot or torch may be in collision during part of the weld motion. Adding motion planning software in this workflow could reduce the time required to generate a robot program. However, the capabilities of motion planning algorithms in this context are not clear. We present a generic framework to evaluate and compare motion planning algorithms for robot arc welding. This framework allows integration of existing open source and commercial planners on different levels of abstraction depending on their capabilities. Different planners can be combined and benchmarked as a group if a single planner is not sufficient to solve all aspects of the problem. The framework is applied to two realistic welding cases to show its potential.