engleski

Guidance of Formations of Autonomous Underwater Vehicles by Virtual Potentials Method

This dissertation presents a method for the decentralized coordinated guidance of autonomous underwater vehicles (AUVs) based on virtual potentials. The virtual potentials (VP) motion planning framework possesses the qualities of cross-layer uniting in a single formulation several imporant navigation goals: (i ) stable and effective cruise towards a target, (ii ) avoidance of collisions, (iii) formation-keeping. Two realistic models of vehicle dynamics were used – a cruise-type vehicle and one independently actuated in heave. A layer linking the VP framework with the low-level controllers was developed, as was a navigation filter of the sigma-point Unscented Kalman filter variety. Furthermore, realistic process and measurement noises were simulated, which include non-stationary effects and effects beyond first order statistics, as well as bias, flicker and intermittent sensor failure. A non- consensual local knowledge of the environment, in terms of obstacles, was also simulated. A special discussion was dedicated to the broadcast communications protocols in a formation. The VP framework was also used as a tool to generate realistic pseudo-stationary sea currents in 3D. Two novel criteria for performance assessment were proposed and used in testing. The VP framework was shown to exhibit robust, stable, resource- efficient formation-keeping and navigation towards prescribed goals.

unmanned marine vehicles; autonomous underwater vehicles; decentralized control; coordinated control; navigation; guidance and control of marine vehicles; virtual potentials; rotor vector fields; local minima avoidance; collision avoidance; formation cruising

nije evidentirano