Abstract
In robotics research, it is often difficult to compare and evaluate techniques experimentally. This paper identifies these difficulties and provides solutions based on our work during the last four years in the field of sampling-based motion planning.