Convergence of the Value Function in Optimal Control Problems with\n Unknown Dynamics

2021/05/28 by Andrea Pesare, Pesare, Andrea, Michele Palladino +3 · 1 citation
Engineering · Computer Science · #Advanced Control Systems Optimization #Adaptive Dynamic Programming Control #Reinforcement Learning in Robotics

paper · pdf · doi:10.48550/arxiv.2105.13708

Abstract

We deal with the convergence of the value function of an approximate control\nproblem with uncertain dynamics to the value function of a nonlinear optimal\ncontrol problem. The assumptions on the dynamics and the costs are rather\ngeneral and we assume to represent uncertainty in the dynamics by a probability\ndistribution. The proposed framework aims to describe and motivate some\nmodel-based Reinforcement Learning algorithms where the model is probabilistic.\nWe also show some numerical experiments which confirm the theoretical results.\n

Cited by

• Average Optimal Control of Uncertain Control-Affine Systems

Related

• A model for system uncertainty in reinforcement learning
• A Convex Programming Approach to Data-Driven Risk-Averse Reinforcement Learning
• Optimal control with learning on the fly: a toy problem
• Stability Analysis of Optimal Adaptive Control using Value Iteration\n with Approximation Errors
• On the continuity and smoothness of the value function in reinforcement learning and optimal control