Direct Adaptive Control of Systems with Actuator Failures: State of the Art and Continuing Challenges

Abstract

In this paper, the problem of controlling systems with failures and faults is introduced, and an overview of recent work on direct adaptive control for compensation of uncertain actuator failures is presented. Actuator failures may be characterized by some unknown system inputs being stuck at some unknown (fixed or varying) values at unknown time instants, that cannot be influenced by the control signals. The key task of adaptive compensation is to design the control signals in such a manner that the remaining actuators can automatically and seamlessly take over for the failed ones, and achieve desired stability and asymptotic tracking. A certain degree of redundancy is necessary to accomplish failure compensation. The objective of adaptive control design is to effectively use the available actuation redundancy to handle failures without the knowledge of the failure patterns, parameters, and time of occurrence. This is a challenging problem because failures introduce large uncertainties in the dynamic structure of the system, in addition to parametric uncertainties and unknown disturbances. The paper addresses some theoretical issues in adaptive actuator failure compensation: actuator failure modeling, redundant actuation requirements, plant-model matching, error system dynamics, adaptation laws, and stability, tracking, and performance analysis. Adaptive control designs can be shown to effectively handle uncertain actuator failures without explicit failure detection. Some open technical challenges and research problems in this important research area are discussed.

A fault tolerant control allocation scheme with output integral sliding modes

“…Other FTC methods such as the Pseudo Inverse Method (e.g. Konstantopoulos and Antsaklis (1999)) use static output feedback to deal with actuator faults and the model-following approach of Tao and Joshi (2008) exploits an adaptive output feedback framework which does not require state estimation.…”

Vision-assisted adaptive target tracking of unmanned ground vehicles

“…For the UGV, the operational conditions or physical parameters of the vehicle will be unknown or changing. In addition, the vehicle may operate experience actuator fading problem .Much progress has been made in fault-tolerant control design with actuator failure since the early work on reconfigurable control and self-repairing flight control system in 1980 [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. Most of them concentrate on accommodating faults by using on-line fault-detection and identification/estimation to reconfigure control inputs.…”

A Direct Adaptive Control Approach in the Presence of Model Mismatch

“…We will focus on linear time-invariant systems (plant) described by: (1) where Ao, Sa E Rn " n , Bo, bb E Rn " 'n , x(t) E Rn , and u(t) E R'n . A Bo represent the nominal values of the system parameters, and ba, bb represent unknown parameter deviations, which are assumed to be constant.…”

“…where Direct adaptive control with actuator failures was extensively studied (e.g., [1], [2], wherein different adaptive schemes were developed, that included state feedback for state tracking (SFST), state feedback for output tracking (SFOT) and output feedback for output tracking (OFOT). While all these adaptive schemes require some assumptions in the form of plant-model matching conditions, the SFST scheme requires the most stringent matching conditions.…”

“…Direct model reference adaptive control (MRAC) has been known as an effective method for state or output tracking. In particular, the `state feedback-for-state-tracking (SFST)' adaptive architecture has the advantages of relatively simple implementation and effective state tracking in the presence of parameter uncertainties as well as actuator failures [1,2,3]. However, to ensure stability and asymptotic state tracking, the SFST control law requires rather stringent plant-model matching conditions to be satisfied.…”