Reliability analysis of discrete robotic control systems

To automate the process of designing robotic control and sensing systems, techniques are developed for calculating the reliability that the system error stays inside a sequence of hyperellipsoids. The reliability is then used to select a reliable combination of control and sensing algorithms. An easily calculated lower bound on the liability for multidimensional systems is found. These techniques are then applied to a robotic visual positioning case study which has constraints on the total execution time and the positioning accuracy. Simulation results using PUMA 560 kinematic and dynamic parameters are presented. From 100 possible plans, three are deemed to be reliable in meeting both constraints.<<ETX>>