Efficient Uncertainty Propagation in Model-Based Reinforcement Learning Unmanned Surface Vehicle Using Unscented Kalman Filter

This article tackles the computational burden of propagating uncertainties in model predictive controller-based policy probabilistic model-based reinforcement learning (MBRL) system for an unmanned surface vehicles (USV). We proposed filtered control using unscented Kalman filter (FPMPC-UKF) that introduces (UKF) a more efficient uncertainty propagation MBRL. A USV based on FPMPC-UKF is developed and evaluated by position-keeping target-reaching tasks real data-driven simulation. The experimental results demonstrate significant superiority method balancing performance burdens under different levels disturbances compared with related works USV, therefore indicate its potential challenging scenarios limited resources.