Neurodynamics-based Model Predictive Control Of Autonomous Underwater Vehicles

In the21century, the exploitation and utilization of the ocean is increasing greatly. As the main working platform in the complex ocean environment, autonomous underwater vehicle (AUV) has obtained enormous attention all over the world. Motion control is one of the key technology of AUV’s research. Since AUV is a typical MIMO system with the property of strong nonlinearity, close coupling and constrained uncertainty, it is quite difficult to achieve the precise control of AUV’s movement. In this paper, in view of all these characteristics of AUV system, we use the advanced control technology and superior online optimization algorithm to realize the accurate control of AUV.By virtue of coordinate transformation and Newton-Euler equation of rigid body, AUV’s kinematics model and dynamics model are established according to the kinematics and the fluid mechanics. By some reasonable simplification, the two models can be reduced to a four degree-of-freedom kinematic model and a dynamic model of vertical plane. Taking these two simplified models as control object, we then design the control algorithm for the system. Model predictive control (MPC) is an advanced control technology. It has many significant advantages, such as low requirement for model precision, the ability to implement the optimization dynamically, high robustness and be able to deal with the constraint conditions et al. All these superiorities make it a very suitable method for the control of underwater vehicle. The core of model predictive control is the rolling optimization process in the limited time horizon, this requires the relevant optimization algorithm has the ability to perform calculation with high accuracy in real-time. On account of the ability to deal with large scale computing in parallel and the fast convergence feature, Neural network becomes an effective means of solving the online optimization problem.In summary, in this paper, we transform the optimization of the objective function in MPC into a quadratic programming problem with constraints. Then a projection neural network is constructed to obtain the control increment through real time optimization. By introducing the neural dynamics optimization method into the MPC of AUV, we perform the stabilization and trajectory tracking control successfully. The simulation results show effectiveness of the proposed method for precise stabilization and trajectory tracking control of AUV.