| In this dissertation performance of adaptive and optimal control methods in dealing with control of parametric unknown and uncertain systems are evaluated. Dynamic equations of autonomous underwater vehicle (AUV) have enough degree of complexity to be a challenging choice. Equations are nonlinear function of variables and parameters are highly affected by operating conditions.; While there are computational advantages in employing discrete controller and parameter estimator, behaviour of a continuous system is better represented by numerical solution of differential equations than its discrete model. Hybrid adaptive control formed based on the estimates of system discrete model parameters is a method that emerged from this idea. This approach, however, has its own problems. Poles, zeros and gain of discrete time model of a system are functions of sampling frequency. A minimum phase continuous system may be discretized to a non-minimum phase sampled data model by just increasing the sampling rate. Range of sampling period producing a stable invertible discrete model of a minimum phase AUV dynamics is a key question which has to be answered before any step in the design of a controller is to be undertaken. This gets more complicated when the unknown system is assumed open loop unstable. Moreover, increase in sampling frequency leads to a substantial drop in high frequency gain. This forms a numerically ill conditioned problem. The proposed adaptive normalization shows perfect ability in overcoming numerical problems emerging from very low or high gain systems. Direct and indirect methods of adaptive control augmented by adaptive normalization manage the closed loop stability and transient performance of AUV in a wide range of sampling rate.; Practically, the source of deliverable thrust is limited. In a manoeuvre, the system may require some extra thrust to overcome overall drag imposed on the system. If enough force can not be supplied, cross coupling among variables, i.e. speed and velocity of yaw (pitch), causes failure in accomplishing defined tasks. Off-line and/or on-line approaches toward seeking a bound for rudder control are imperative. On-line approach based on generalized predictive adaptive control of nonlinear system is to deliver an optimal bound for rudder control, covering the lack of ability of the system in producing required thrust.; Since a controller is usually designed based on a generally known linear part of a system, or its estimates, ignoring higher order dynamic terms, in circumstances, may have crippling effects on the control process. These terms are commonly known as unmodeled dynamics. Adaptive control methods require them to be bounded to known limits. Moreover, due to unanticipated faults, a system may be affected by change in the parameters and/or addition of extra unmodeled terms. To provide an effective solution, special provisions in the control system are introduced as fault accommodation. The core of the design is a nonlinear function predictor based on the CMWCI. The objective of the special neural network based mechanism is to predict the effects of unmodeled fault which then constitutes the feedforward part of control law.; Autonomous underwater vehicle can not undertake any manoeuvre in a restricted environment, unless appropriate provisions are embedded in the control system. Optimal guidance algorithm is to manage system behaviour in dealing with such a case. (Abstract shortened by UMI.)... |
