An Efficient Underwater Navigation Method Using MPC with Unknown Kinematics and Non-Linear Disturbances

Many Autonomous Underwater Vehicles (AUVs) need to cope with hazardous underwater medium using a limited computational capacity while facing unknown kinematics and disturbances. However, most algorithms proposed for navigation in such conditions fail to fulfil all conditions at the same time. In this work, we propose an optimal control method, based on a receding horizon approach, namely MPC (Model Predictive Control). Our model also estimates the kinematics of the medium and its disturbances, using efficient tools that rely on the use of linear algebra and first-order optimization methods. We also test our ideas using an extensive set of simulations, which show that the proposed ideas are very competitive in terms of cost and computational efficiency in cases of total and partial observability.