Fully probabilistic control for uncertain nonlinear stochastic systems

This paper develops a novel probabilistic framework for stochastic nonlinear and uncertain control problems. The proposed exploits the Kullback–Leibler divergence to measure between distribution of closed-loop behavior dynamical system predefined ideal distribution. To facilitate derivation analytic solution randomized controllers systems, transformation methods are applied such that dynamics controlled becomes affine in state input. Additionally, knowledge uncertainty is taken into consideration controller. derived controller shown be obtained from generalized state-dependent Riccati takes state- control-dependent functional system. demonstrated on an inverted pendulum cart problem, results obtained.