Researches On A Novel Dish-like Underwater Robot And Attitude Predictive Control

In recent years, with the gradual shortage of land resources, ocean covered71%surface of the global has become the focus of every country. Underwater robot technology, as a multidisciplinary scientific research, has an indispensable role for ocean development and the consolidation of the territorial seas. That how can an underwater robot work independently, stably and efficiently in the complex environment of the ocean has been drawing much more attention in the whole industry. There is a lot of research on underwater robot, the structures and control methods of underwater robot are the core technology of spreading underwater robot applications. Study of the novel structure and control method for the underwater robot will play a great role on accelerating the practical application and engineering process of underwater robots.Underwater robot is a MIMO, strongly nonlinear, time varying and inertia system. GPC has been widely used in industry fields, but in order to use GPC in underwater robot, it needs a good predictive model as a precondition. This dissertation proposed a new type of underwater robot research has been done on the predictive attitude control method’s application to this robot’s navigation attitude control. Firstly, based on introducing and analyzing the structure and working mode of this robot, a hydrodynamic equation was obtained through mechanism modelling.The virtual model was simulated by the FLUENT, an authority hydrodynamic analysis software. As for the hydrodynamic performance of the submarine, considering the possible hydrodynamic influence caused by the inlet hole, the problem was eliminated through qualitative analysis on the opening structure by the correlation theory. Then the ramp flight performance of spheroid and dish-like body was simulated by the numerical method using RNG k-ε turbulence model. At the same time, PMM numerical experiments of spheroid and disk-like body were calculated using the technology of dynamic mesh, some key hydrodynamic coefficients were obtained through it. The feasibility and accuracy of numerical calculation was proved by the comparison between ellipsoid body hydrodynamic coefficients and the result of potential binning theory. It shows that this new type of underwater robot has good flexibility, stability and strong hydrodynamic performance characters.The profile of this robot was optimized based on the research above. Finally, the hydrodynamic equation was obtained based on the hydrodynamic coefficients calculated by the hydrodynamic motion simulation of the robot.After briefly analyzing the principle of SVR and the feasibility of applying the SVR to indentify the system model, a regression method, based on SVR, to identify the dish-like robot system control model through the MIMO was proposed, aiming to solve the problem of strong nonlinear, complexity of model parameters. SVR parameters optimization methods were analyzed and compared, a multi-windows grid search method was proposed to guarantee the accuracy and solve the problem of stagnation points within a reasonable amount of calculation. A predictive model, which can identify only and modify the system constantly, was introduced for adapting the complex and changeful working environment in time.The Logistic model used for describing population model was introduced to reject old data based on the weight calculated by it. New data was filtered by judging the linear correlation after its being mapped to higher space. And a new model would be gotten by this method after retrained. The lack of sparsity and the re-identification of large operation problems, which may appear in SVR’s application, were solved by this method. The predictive ability of the model obtained by online SVR was proved by comparing the identifying performance and its error and the control performance of online and offline SVR.A GPC algrithom for dish-like underwater robot was proposed to solve the lag problem caused by the strong inertia of the robot. Considering this MIMO nonlinear system, it’s necessary to simplify the model indentification and avoid large amount of matrix calculation during model prediction. In order to control easily, Multivariate function Taylor expansion was introduced to linearize the SVR form nonlinear system instantaneously and a difference equation of the system was obtained. So an online self-adaptive MIMO GPC algrithom for this dish-like underwater robot was proposed.Heading, roll and pitch were controlled by3controllers by this method. Fixed heading, fixed roll and PPM S motion experiment were carried on in the national key lab testing pool of underwater robot technology to test the maneuvering characters. There were3methods’ performance, NN-GPS,GPC with hydroynamic model and SVR-GPC. They were compared to verify the well performance of this nonlinear control method of SVR-GPC this dissertation proposed.