Development and testing of a reliability performance index for modular robotic systems

One issue pertaining to modular robotic systems is the integration of modules into a fully functional robot system. Two criteria for modules integration not previously investigated are the reliability and accuracy of the system. This paper presents the results of the development of a framework for a "criterion" embodying these two modular robot characteristics. Using a probabilistic representation of manipulator kinematics and a reliability block diagram model of the manipulator system, a reliability performance index (RPI) representing the probability of no hardware or software failure and the manipulator achieving a specified position and orientation is developed. The RPI is tested with a case study consisting of a three degree-of-freedom planar manipulator assembled from a choice of six joint modules of varying reliability and precision and a choice of six link module combinations of varying lengths and machining tolerances. A straight-line, square trajectory is specified and the RPI is calculated for each combination of joint modules and links, a total of 1296 different combinations. An analysis of variance (ANOVA) is performed on the results using the different joint locations and link options as factors and the different joint modules and link options as factor levels. The different factors are tested for significance and the Tukey Studentized Range Test is performed to determine significance of the different joint modules and link options. Using this statistical testing, a 70% reduction in the module design space is achieved using the RPI. Optimization using other appropriate manipulator criteria can then be performed to generate the final configuration. Additional extensive case studies are needed to fully develop the RPI to a stage necessary for implementation into a computer-aided design system for modular robot configuration design. The RPI may also be useful in the quantification of the overall system reliability and performance of any system based upon measured error, such as control systems.<>