Model Predictive Fault Tolerant Control for Omni-directional Mobile Robots.

This paper describes the design of a Fault Tolerant Control scheme for an omni-directional mobile robot with four mecanum wheels. We consider actuation faults in which the wheels are not able to receive commands, but still can rotate freely due to the friction with the ground. A Non-linear Model Predictive Controller is employed, in order to appropriately exploit the inherited actuation redundancy of the omni-directional robot, and provide on the fly a unified accommodation solution for multiple combinations of actuation faults. A basic Fault Detection and Isolation scheme is responsible for identifying and isolate the faulty wheels. Accordingly, the control input vector and consequently the dynamic model of the vehicle are reformulated depending on the identified faults. Hence, the predictive control scheme is dynamically updated and calculates the optimal actuation solution, given the faults occurred. The proposed scheme, is able to guide the vehicle to any goal configuration within the workspace, while simultaneously satisfying state (e.g obstacle avoidance) and input (e.g motor limits) constraints. The efficacy of the proposed scheme is evaluated via a set of realistic simulation scenarios, where an omni-directional mobile robot executes a way point tracking mission with obstacle avoidance, inside a restricted workspace and in the presence of different combinations of actuation faults.