Lyapunov Robust Constrained-MDPs for Sim2Real Transfer Learning

    •  Russel, R.H., Benosman, M., van Baar, J., Corcodel, R., "Lyapunov Robust Constrained-MDPs for Sim2Real Transfer Learning" in Federated and Transfer Learning, August 2022.
      BibTeX TR2022-111 PDF
      • @incollection{Russel2022aug,
      • author = {Russel, Reazul Hasan and Benosman, Mouhacine and van Baar, Jeroen and Corcodel, Radu},
      • title = {Lyapunov Robust Constrained-MDPs for Sim2Real Transfer Learning},
      • booktitle = {Federated and Transfer Learning},
      • year = 2022,
      • month = aug,
      • url = {}
      • }
  • MERL Contacts:
  • Research Areas:

    Machine Learning, Optimization


Safety and robustness are two desired properties for any reinforcement learning algorithm. Constrained Markov Decision Processes (CMDPs) can handle additional safety constraints and Robust Markov Decision Processes (RMDPs) can perform well under model uncertainties. In this chapter, we propose to unify these two frameworks resulting in Robust Constrained MDPs (RCMDPs). The motivation is to develop a framework that can satisfy safety constraints while also simultaneously offer robustness to model uncertainties. We develop the RCMDP objective, derive gradient update formula to optimize this objective and then propose policy gradient based algorithms. We also independently propose Lyapunov-based reward shaping for RCMDPs, yielding better stability and convergence properties.