Robust Model Predictive Path Integral Control: Analysis and Performance Guarantees

In this paper we propose a novel decision making architecture for Robust Model Predictive Path Integral control (RMPPI) and investigate its performance guarantees applicability to off-road navigation. Key building blocks of the proposed are an augmented state space representation system consisting nominal actual dynamics, placeholder different types tracking controllers, safety logic propagation, importance sampling scheme that takes into account capabilities underlying control. Using these ingredients, derive bound on free energy growth dynamical which is function task constraint satisfaction level, controller, error stochastic optimization used within RMPPI. To validate growth, perform experiments in simulation using two namely iterative Linear Quadratic Gaussian Contraction-Metric based We further demonstrate RMPPI real hardware GT AutoRally vehicle. Our outperforms MPPI Tube-MPPI by alleviating issues aforementioned model predictive controllers related either lack robustness or excessive conservatism. provides best worlds terms agility disturbances.