eduzhai > Applied Sciences > Engineering >

Combining Model-Based and Model-Free Methods for Nonlinear Control A Provably Convergent Policy Gradient Approach

  • Save

... pages left unread,continue reading

Document pages: 29 pages

Abstract: Model-free learning-based control methods have seen great success recently.However, such methods typically suffer from poor sample complexity and limitedconvergence guarantees. This is in sharp contrast to classical model-basedcontrol, which has a rich theory but typically requires strong modelingassumptions. In this paper, we combine the two approaches to achieve the bestof both worlds. We consider a dynamical system with both linear and non-linearcomponents and develop a novel approach to use the linear model to define awarm start for a model-free, policy gradient method. We show this hybridapproach outperforms the model-based controller while avoiding the convergenceissues associated with model-free approaches via both numerical experiments andtheoretical analyses, in which we derive sufficient conditions on thenon-linear component such that our approach is guaranteed to converge to the(nearly) global optimal controller.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...