This thesis considers the design and evaluation of autopilots for Remotely Operated Underwater Vehicles (ROVs), unmanned submarines used in offshore oil, salvage and military applications. A very comprehensive hydrodynamic model of a ROV produced by the National Maritime Institute, Feltham, Middlesex, is subjected to an extensive verification study. It is concluded that conventional hydrodynamic modelling techniques are very expensive and uncertain and hence any ROV autopilot must be, in some manner, adaptive; that is, independent of 'a priori' knowledge of the vehicle. The theory, implementation and simulated performance of three different adaptive autopilots is presented, based on the NMI model. Two of these systems use multivariable recursive system identification techniques to estimate the performance of the vehicle on-line. These methods are also discussed as an alternative route to ROV models. A summary of the thesis is given along with recommendations for areas which require further study. An appendix is included which describes a series of tank trials at Admiralty Research Establishment (Haslar); one of the goals of these tests was to validate this simulation study.

The dynamic characteristics of autonomous underwater vehicles (AUVs) present a control problem that classical methods cannot often accommodate easily. Fundamentally, AUV dynamics are highly non-linear, and the relative similarity between the linear and angular velocities about each degree of freedom means that control schemes employed within other flight vehicles are not always applicable. In such instances, intelligent control strategies offer a more sophisticated approach to the design of the control algorithm. Neurofuzzy control is one such technique, which fuses the beneficial properties of neural networks and fuzzy logic in a hybrid control architecture. Such an approach is highly suited to development of an autopilot for an AUV. Specifically, the adaptive network-based fuzzy inference system (ANFIS) is discussed in Chapter 4 as an effective new approach for neurally tuning course-changing fuzzy autopilots. However, the limitation of this technique is that it cannot be used for developing multivariable fuzzy structures. Consequently, the co-active ANFIS (CANFIS) architecture is developed and employed as a novel multi variable AUV autopilot within Chapter 5, whereby simultaneous control of the AUV yaw and roll channels is achieved. Moreover, this structure is flexible in that it is extended in Chapter 6 to perform on-line control of the AUV leading to a novel autopilot design that can accommodate changing vehicle pay loads and environmental disturbances. Whilst the typical ANFIS and CANFIS structures prove effective for AUV control system design, the well known properties of radial basis function networks (RBFN) offer a more flexible controller architecture. Chapter 7 presents a new approach to fuzzy modelling and employs both ANFIS and CANFIS structures with non-linear consequent functions of composite Gaussian form. This merger of CANFIS and a RBFN lends itself naturally to tuning with an extended form of the hybrid learning rule, and provides a very effective approach to intelligent controller development.

No description available.

No description available.

The design of automatic systems for steering a ship presents difficult challenges because of their dynamic properties which vary considerably within the range of sailing conditions. Automatic steering of ships has its origin at the beginning of the century and was prompted by the introduction of the gyrocompass. Until the earlier 70s almost all autopilots for a ship were based on the proportional-derivative-integral (PID) controller. The main disadvantage with PID controllers is that the optimal parameters setting can be achieved only for a particular sailing condition. This shortcoming was and is still dealt with in the framework of adaptive theory where the controller parameters are adjusted in the attempt to seek the optimum of a pre-set performance function. Despite such a potential advantage, at present adaptive control theory is limited to linear plants and requires a certain amount of a-priori information for a successful application. This thesis is concerned with the applicability of intelligent control techniques to the problem of designing course-keeping and course-changing autopilots for ships. For this reason the framework of intelligent control theory is introduced and a pragmatic definition of intelligent controllers is stated. The learning and adaptive features of neural networks and fuzzy logic systems are exploited and used to solve advantageously the control design problem. Adaptive networks are used as a unifying structure where different kinds of neural networks and fuzzy logic paradigms can be described. In this framework, comparisons between neural networks and fuzzy logic systems are made and results from one field can be easily extended to the other. Although the use of such systems for the design of autopilots is in its early stage, the majority of the contributions which have appeared in literature have focused on the use of feedforward networks trained with the back-propagation algorithm. The main contributions of this thesis are the critical analysis of the feedforward network controller trained with the back-propagation algorithm, the proposition of an alternative controller architecture based on the use of radial basis function networks and to give conditions under which the stability analysis of the intelligent controllers so designed can be evaluated.

Fixed-wing unmanned aerial vehicles (UAVs) with the ability to hover combine the speed and endurance of traditional fixed-wing fight with the stable hovering and vertical takeoff and landing (VTOL) capabilities of helicopters and quadrotors. This combination of abilities can provide strategic advantages for UAV operators, especially when operating in urban environments where the airspace may be crowded with obstacles. Traditionally, fixed-wing UAVs with hovering capabilities had to be custom designed for specific payloads and missions, often requiring custom autopilots and unconventional airframe configurations. With recent government spending cuts, UAV operators like the military and law enforcement agencies have been urging UAV developers to make their aircraft cheaper, more versatile, and easier to repair. This thesis discusses the use of the commercially available ArduPilot open source autopilot, to autonomously transition a fixed-wing UAV to and from hover flight. Software modifications were made to the ArduPilot firmware to add hover flight modes using both Proportional, Integral, Derivative (PID) Control and Model Reference Adaptive Control (MRAC) with the goal of making the controllers robust enough so that anyone in the ArduPilot community could use their own ArduPilot board and their own fixed-wing airframe (as long as it has enough power to maintain stable hover) to achieve autonomous hover after some simple gain tuning. Three new hover flight modes were developed and tested first in simulation and then in flight using an E-Flight Carbon Z Yak 54 RC aircraft model, which was equipped with an ArduPilot 2.5 autopilot board. Results from both the simulations and flight test experiments where the airplane transitions both to and from autonomous hover flight are presented.