Autonomous Underwater Vehicle Motion Planning via Sampling Based Model Predictive Control

Abstract

In order to improve the ability of independent planning for AUV (Autonomous Underwater Vehicle), a new method of motion planning based on SBMPC (Sampling Based Model Predictive Control) is proposed, which is combined with model predictive control theory. Input sampling is directly made in control variable space, and sampling data is substituted into the predictive model of AUV motion. Then surge velocity and yaw angular rate in next sampling time are obtained through calculations. If predictive states are evaluated according to the performance index previously defined, optimal prediction of AUV states in next sampling can be used to realize motion planning optimization. Effects of three sampling methods (viz. uniform sampling, Halton sampling and CVT sampling) on motion planning performance are also compared in simulations. Statistical analysis demonstrates that CVT sampling points has the most uniform coverage in two-dimensional plane when amount of sampling points is the same for three methods. Simulation results show that it is effective and feasible to plan a route for AUV by using CVT sampling and rolling optimization of MPC (Model Predictive Control).