Robert Dyro

Robert Dyro


Robert is a second-year graduate student in the Aeronautics and Astronautics Department. Prior to coming to Stanford, he attended University of California, Los Angeles where he received his B.S. in Aerospace Engineering with a minor in Philosophy. His current research focuses on optimal control algorithms that explicitly leverage knowledge of uncertainty in learning based systems. He’s interested in how second order optimization can speed up learning and provide safety guarantees in autonomous systems.


ASL Publications

  1. R. Dyro, J. Harrison, A. Sharma, and M. Pavone, “Particle MPC for Uncertain and Learning-Based Control,” 2021. (Submitted)

    Abstract: Autonomous decision-making in novel or changing environments requires quantification and consideration of uncertainties in the system or environment dynamics that impact downstream control performance. Thus, as robotic systems move from highly structured environments to open worlds, incorporating uncertainty in learning or estimation into the control pipeline is essential for robust and efficient performance. In this paper we present a nonlinear particle model predictive control (PMPC) approach to control under uncertainty. This approach, due to the particle representation of uncertainty, is capable of handling arbitrary uncertainty specifications. We implement our nonlinear PMPC scheme with a sequential convex programming non-convex optimization scheme, and we discuss practical implementation of such a framework. We investigate our approach for two robotic systems across three problem settings: time-varying, partially observed dynamics; sensing uncertainty; and model-based reinforcement learning, and show that our approach improves performance over baselines in all settings.

    @inproceedings{DyroHarrisonEtAl2021,
      author = {Dyro, R. and Harrison, J. and Sharma, A. and Pavone, M.},
      title = {Particle MPC for Uncertain and Learning-Based Control},
      year = {2021},
      note = {Submitted},
      keywords = {sub},
      owner = {jh2},
      timestamp = {2021-03-23}
    }