Adding diagnostics to intelligent robot systems

Robot systems are increasing in complexity. Trying to diagnose a robot that is non-functional or exhibiting suboptimal performance can be a major challenge. A framework for plug-n-play addition of diagnostics to modules in an object oriented software framework is presented. The methods for modeling of system modules, their transition to a Bayesian model and final implementation are described. The methodology is exemplified for a mobile manipulation system and experimental results are presented.