Direct Model Reference Adaptive Control with Actuator Failures and Sensor Bias

Abstract

ACTUATOR and sensor faults have been implicated in several aircraft loss-of-control accidents and incidents. Direct model reference adaptive control (MRAC) methods have been suggested as a promising approach for maintaining stability and controllability in the presence of uncertainties and actuator failures without requiring explicit fault detection, identification, and controller reconfiguration (e.g., [1,2]). (An extensive amount of literature is available on adaptive control in the presence of actuator failures, but could not be listed in this brief Note due to space limitations; however, a detailed reference list is available in [2]). In addition to actuator faults, sensor faults may also compromise safety [3–5]. A common type of sensor fault is unknown sensor bias, which can develop during operation in one or more sensors, such as rate gyros, accelerometers, altimeter, etc. If used directly in an MRAC law, such offsets in sensor measurements can have detrimental effects on closed-loop stability, which can no longer be theoretically guaranteed. Accommodation of sensor faults in an MRAC setting has been addressed (e.g., [6–8]). However, adaptive control of systemswith simultaneous actuator and sensor faults has not been adequately addressed. Toward that goal, MRAC control laws using state feedback for state tracking were developed in [9] for the case with simultaneous sensor bias and actuator failures, and signal boundedness as well as bounded or asymptotic tracking were obtained. This Note elaborates on the results of [9] and presents numerical examples to illustrate the methods.