Research On Identification And Control Of An Open-frame ROV

The dissertation is on the research of the identification and control for an open-frame remotely operated underwater vehicle named "GDROV". The vehicle is designed for the purpose of the inspection of cracks, crevice and other potentialproblem . of dikes. It is a underwater vehicle equipped with many advanced inspecting sensors. The two main inspecting sensors equipped are profiler sonar and imaging sonar. The movement of the GDROV has to be stable for the sake of the regular performance of these two sonar. Therefore, The research on how to acquire stable movement of the vehicle is crucial to the real application of thevehicle. " The' computer simulation environment is very important to the design and regulation of the underwater vehicle motion control system when the filed-test is not feasible for some reasons. To the "GDROV", it is very difficult to precisely acquire its hydrodynamic coefficients through theory computation or PMM experiments for the reason of its open-frame structure and design characteristics.So, The research on how to identify the coefficients of the vehicle with the help of collected experiment data through system identification method has to be dealt carefully.Therefore, the dissertation can be divided into two main parts:1. the research on the identification of hydrodynamic coefficients and the construction of simulative environment of the vehicle.2. the research on the motion control of the vehicle.For the first part, an identification equation of the vehicle is first constructed on the integrated consideration of the equipped sensors, the physical constraint of identification tank and the structure and motion characteristics of the vehicle.Then, a least-square technique is utilized to estimate the hydrodynamic coefficients of the vehicle. Last, simulation environment of the vehicle is constructed and simulation experiments are carried out to test the effectiveness of the LS identification technique.For the second part, Firstly, a sliding mode fuzzy control method is presented integrating sliding mode control and fuzzy control. The control method can greatly abate the chattering of the VSC controller and increase the control performance. Simulation and filed-test result verified the effectiveness of the control method. Secondly, in order to increase the robustness of the neural network controller trained by error back-propagation method, a novel training method is devised. The training method is constructed through combining guidance law of sliding mode control theory with EBP training method. A neural network controller is constructed and trained by the new training method. Simulation results show that the neural network controller trained exhibit good robustness faced with noise and unmodeled dynamics. Lastly, An approach to design membership function and fuzzy rule base of FLC simultaneously with GA is presented .Simulation results verified that the performance of the GA-optimized FLC was better than the FLC tuned manly.