Predictive Control Based Position And Attitude Control Of Underwater Gliding Robotic Fish

Underwater gliding robotic fish is a new type of underwater robots. Combining the characteristics of robotic fish and underwater glider, it has advantages of small size, cost-efficiency, hybrid drive, flexible work and long duration. Gliding robotic fish has promising applications such as underwater cooperative detection, marine search and rescue, and so on.In this paper, according to the characteristics and application requirements of gliding robotic fish, after modelling the robot, model predictive control theory is employed to study attitude control and tracking control deeply, including the construction of the performance index function, the choice of numerical optimization algorithm and the design of the controller parameters, etc. The main results are summarized as follow.Firstly, the robotic fish is simplified into several dynamic and static mass points in appropriate location based on its composition and working principle. Then the whole body is regard as two rigid-body system composed of dynamic mass(slide block) and static mass(the rest part) for force analysis. Furthermore, the inertial frame, the body-fixed reference frame and the velocity reference frame are established and their transformation relations are derived. Subsequently, the kinematical equation is obtained with the direction cosine matrix differential equation and the transformation relations among the frames, and the kinetic equation is established through analyzing of the force and moment acting on the body. As a result, combining the kinematical equation with the kinetic equations gives the whole model of the robot, based on which the Simulink model is built to lay the foundation for following research.Secondly, considering the gliding robotic fish’s attitude control requirements in gliding motion and model features, the attitude controller based on nonlinear model predictive control method is designed to realize tail-enabled yaw angel control and pitch angel control with movable mass as input. In order to improve the control performance, the index function is modified to solve the saturation and jump of input. The role of each weighting coefficient is analyzed referring to the simulation result and the principle of parameters selection is presented. Moreover, for the purpose of enhancing the real-time performance of rolling optimization calculation, several numerical optimization algorithms are contrastively analyzed and some methods of reducing the computing time consumption are summarized. At last, the effectiveness of control design is demonstrated through extensive simulations.Finally, in consideration of the requirements of tracking and hovering working mode of underwater gliding robotic fish in practical application, tracking control based on nonlinear model predictive control is further studied. Target tracking controllers for three-input case(excess mass, slide block and tail angle) and two-input case(slide block and tail angle) are designed respectively. The former can realize not only target tracking but also hovering at the target point. The latter can only go through the target point but is with more energy efficient than the former. In view of the tracking ability limitation caused by the robot’s under-actuated characteristic for two-input case, the feasible region is analyzed based on the model. In addition, indirect target tracking controller which control the position through attitude control is further designed. Compared with the direct control, the indirect control has better control performance at the cost of increasing complexity. The correctness and effectiveness of the above analysis and design are verified via the simulation.