Robot Reliability Estimation Using Interval Methods

Abstract

Reliability is a critical issue for many robotic applications, where failures can be disastrous. Traditional reliability analysis algorithms depend on the algebraic manipulation of given input data (representing subsystem and component reliability). However, this information is particularly difficult to obtain in robotics, where the data is typically uncertain at best. Thus there is inherent uncertainty in robot reliability estimates. In this paper, we discuss the applicability and usefulness of interval techniques to the robot reliability problem. Through a canonical example, we show how the use of a novel interval method can significantly improve the accuracy of robot reliability estimates over existing methods. The approach is show to be more efficient then using corresponding integer methods.