Interval Methods for Improved Robot Reliability Estimation

In this paper, we present and discuss a new interval-based method of reliability estimation using fault trees for the case of uncertain and time-varying input reliability data. The approach is based on the generation of output distributions (probability estimates with appropriate ranges of uncertainty) which preserve the effects of uncertainty in the input (component or subsystem-level) data. The input data is represented using appropriate interval-based structures, and formal interval analysis is used in the propagation of the data, via fault trees. We show that the method avoids the key problem of loss of uncertainty inherent in some previously suggested approaches for the timevarying case. We further show that the method is more computationally efficient than methods proposed previously to solve the above problem. The method is illustrated using an example of reliability estimation for a robot manipulator system.