A fault-diagnostic system for unmanned underwater vehicles has been designe
d and tested in real operating conditions. Actuator faults have been consid
ered, relying on approximate models of the vehicles' dynamics. Fault detect
ion and diagnosis is accomplished by evaluating any significant change in t
he behaviour of the vehicle. This task is performed by a bank of estimators
: a filter is implemented for each actuator fault type, including the no-fa
ult case. The estimators used are extended Kalman filters (EKF), due to the
presence of nonlinearities in the dynamic models. Experimental results are
reported, to demonstrate the effectiveness of the proposed approach. (C) 1
999 Published by Elsevier Science Lm. All rights reserved.