The Pc metric: A safety performance measure for evaluation of collision avoidance algorithms

摘要

Several techniques have been proposed for navigation and collision avoidance algorithms in the field of mobile robotics. But the research on establishing standardized performance measures for evaluating such algorithms are gaining attention very recently. In this work, a novel performance metric is proposed to comprehensively measure and evaluate the performance of robots in terms of safety, during navigation and collision avoidance. The metric is based on the concept of probability of collision and is applicable to environments involving humans which are considered to be highly dynamic and uncertain. The definition and the estimation of the proposed metric are elaborated initially. The metric is then validated through an experiment. Finally, the application of the metric is demonstrated using a virtual human environment simulation. The results show the effectiveness of the metric in quantitatively evaluating and comparing the safety performance of the robot with respect to humans.