The Dilemma Diagnoser Approach and its Application to the Fault-Tolerant Control of Planetary Exploration Rovers

Abstract

Most of the fault-tolerant control strategies found in the literature assume conclusive diagnosis, i.e. the current fault mode of the plant is well-known. Although, speculative diagnosis can be a more realistic approach while noise corrupted measurements and plant disturbances hamper the construction of a precise diagnosis statement. Speculative diagnosis consists in providing a set with the most probable fault modes in the control system. The idea of a set with probable fault modes is not new, but reconfiguration is not an easy task in this context. Prompt reconfiguration under lack of a conclusive statement is not properly approached in the current literature. The risks involved in such decision are evaluated and used in the modeling of the Dilemma Diagnoser. It is a decision maker coupling speculative diagnosis statements and control reconfiguration to achieve a safe decision in a particular sense, i.e. keep the control system stable and close to the desired reference. The problem is modeled as a bi-matrix game, the two players are the Diagnoser (choses among several available fault modes in a speculative set) and the Switcher (choses between reconfigure instantaneously or wait until the next diagnosis sample to make a decision). It can be solved by game theory using Mixed-Strategy Nash Equilibrium. A specific control strategy is also developed to recover a multi-wheeled rover from steering motor failures. This strategy is integrated with the Dilemma Diagnoser and applied to the case of the ExoMars Rover. Note that all methods presented here are applicable to all kinds of multi-wheeled rovers and are capable to cover all amplitudes and combinations of steering failures as long as sufficient driving power is available. The fault-tolerant controller is tested in the Planetary Exploration Laboratory (PEL) of the German Aerospace Center (DLR-Oberpfaffenhofen); the controller is embedded in the ExoMars B2 Breadboard Model. The results are satisfactory and allow the vehicle to follow a predefined path formed by waypoints whether faults are present or not. Tests were conducted to ensure robustness of the fault-tolerant control system while driving with satisfactory performance either on Kalk Sand (high sinkage) or Lava Sand (moderate sinkage). Our proposed techniques are capable to lead the faulty rover to the desired path smoothly and progressively decreasing both attitude and displacement errors. The main contributions of this work are: the introduction of the Dilemma Diagnoser, the proposition of an alternative control strategy for steering motor failures, and experimental validation of fault-tolerant controller.