Robust H-infinity methods towards the control and navigation of autonomous underwater vehicles

Major inherent properties make it difficult to control underwater vehicles. These factors include: the highly nonlinear, time-varying dynamic behavior of the underwater vehicle; uncertainties in hydrodynamic coefficients; the higher order and redundant structure when the manipulator is attached; disturbances by ocean currents; and changes in the centers of the gravity and buoyancy due to the manipulator motion which also disturbs the vehicle's main body. The first part of this research aims to develop a nonlinear H infinity Controller in order to counteract the effects of these disturbances. The research into the control of the underwater vehicle looks to the theories developed for the solution of the nonlinear state feedback H infinity problem which is described using a Hamilton-Jacoby inequality.; The success of future navigation of underwater vehicles will be the ability to accurately localize (determine the vehicles specific distance from some fixed point) itself within this underwater domain. The underwater world limits the types of sensors available, as compared to that of above the water surface. Electromagnetic energy propagates very little in water and, thus, instruments such as the Global Positioning System (GPS) have limited use in water. However, if truly autonomous systems are to be developed, good navigation sensory information is needed in order to achieve mission goals and provide safe operation.; The second component of the research proposes a robust filtering method for the underwater vehicle localization problem. Typical, localization algorithms employ the Extended Kalman Filter. But, conventional Kalman Filter methods suffer from the assumption of Gaussian noise statistics, which often lead to failures when they do not hold. Additionally, the linearization errors associated with the implementation of the standard EKF can also severely degrade the performance of the localization estimate. The proposed Robust EKF (REKF) addresses these limitations through the use of the bounded H infinity norm.; Much of the ocean has yet to be discovered. As a result, the use of a priori maps during underwater scientific missions are seldom available and other methods of localization must be used. Typical navigation for underwater missions relies upon fixed transponders that are surveyed into the underwater vehicles work area. As an alternative to localization based upon beacons, this research will look to the Simultaneous Localization and Mapping (SLAM) of the underwater vehicle as it moves through its mission area. Currently, the Stochastic Map is the most often used algorithm for SLAM. The Stochastic Map is essentially an augmented extended Kalman filter. However, as was stated previously, the EKF suffers through linearization and Gaussian assumptions. This research will address the linearization and Guassian assumption errors as they relate to the SLAM problem. The research proposes a new method - Robust Stochastic Mapping. It will use an augmented Robust EKF as a replacement for the standard EKF-SLAM algorithms.