Research On System Architecture And Obstacle Avoidance Control Technology Of Autonomous Underwater Vehicle

The technology of autonomous underwater vehicle (AUV) has achieved an impressive development during the last decade and is playing an important role in ocean research and national defense. To fulfill the demands of further ocean exploration, the AUV is required to accomplish a variety of complex tasks and more intelligence is needed to ensure success of tasks. As an intelligent system, the AUV should be able to adapt its behaviors to the changing environments and make its own decision to accomplish the tasks assigned to it. Architecture of AUV is one of the basis technologies for AUV research and is of great practical value. The objective of this thesis is to establish a universal architecture suitable for different types of AUVs required to carry out different kinds of tasks.In this thesis, the problem of architecture and dynamic obstacle perception, obstacle avoidance planning and control, path planning and motion control is researched in detail and an AUV developed in our laboratory is used as the test bed.Hybrid architecture of AUV was proposed in this thesis and modular based design concept was applied in the implementation of the architecture. Functionality of each modular and relationships between different modules were specified before implementation. Sage-Husa adaptive kalman filtering algorithm was applied in dynamic obstacle perception, precision of perception and real-time requirements were full filled without demanding two much computation resources of the system. A new obstacle avoidance planning method which seeks for optimum desired speed in velocity vector coordinate system is proposed. A set of desired obstacle avoidance speed can be got by integrated analysis on the motion ability, speed, yaw angle of underwater vehicle as well as speed and direction of dynamic obstacles. The final desired speed can be deduced according to the rule of minimal yaw angle. Adaptive finite time PD controller based on fuzzy parameter is designed to control the underwater vehicle to avoid obstacles. The new controller can adjust the sensitivity coefficients adaptively according to collision time. It can work seamlessly with obstacle avoidance planning module so that an integrated design of planning and control for obstacle avoidance is realized.PSO algorithm is applied to the problem of global path planning in complex ocean environments. The problem of global path planning with densely distributed obstacles is tackled by the combination of main line binary tree method and the PSO algorithm, thus reducing computation resources required. Neural network parallel self-learning controller is designed to control cruising motion of the underwater vehicle. Feasibility of the controller was verified by sea-trial experiment.Experiment results shows that the proposed architecture is fit for different task demands and the functional modules designed are reliable and of practical value. Results proposed in this thesis are helpful for further research on architecture of underwater vehicles and related techniques.