Foundations of Reinforcement Learning and Control: Connections and New Perspectives
Abstract
Reinforcement learning and control theory are two adjacent scientific fields that focus on optimizing the controller of unknown dynamical systems using feedback. Although both fields have common roots in dynamic programming, they have evolved with distinct methodologies, goals, and cultures. Despite decades of mutual influence, a significant gap persists between the two communities. This tutorial introduces adaptive control, actor-critic reinforcement algorithms and a new original way to combine these two paradigms for data-driven decision making on a classical locomotion control problem. Our aim is to provide keys to understand the core differences between both approaches and insights to help experts in each field better understand and engage with the tools and approaches of the other.
Funding: C. Vernade thanks the German Research Foundation for [Project 468806714 of the Emmy Noether Programme and Germany’s Excellence Strategy EXC 2064/1–Project 390727645] and acknowledges funding from the European Union [ConSequentIAL Grant 101165883]. C. Szepesvári acknowledges funding from the Canada CIFAR AI Chairs Program, Alberta Machine Intelligence Institute, and NSERC. C. Vernade and O. Eberhard thank the International Max Planck Research School for Intelligent Systems for support. M. White thanks the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canada Research Chair program, the Canada CIFAR AI Chair Program, and the Alberta Machine Intelligence Institute for research support. M. Muehlebach thanks the German Research Foundation for the support.
Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/educ.2026.0310.
Your Access Options
-
Login Options
Purchase Options
Save for laterOther Options
Token AccessClaim access using a tokenRestore guest accessApplies for purchases made as a guest
