PODS: Policy Optimization via Differentiable Simulation

         Miguel Zamora 1 Momchil Peychev 1 Sehoon Ha 2 Martin Vechev 1 Stelian Coros 1

             Abstract                 potentially unsafe. Fortunately, recent years have seen excit-
    Current reinforcement learning (RL) methods use       ing progress in simulation technologies that create realistic
    simulation models as simple black-box oracles.       virtual training grounds, and sim-2-real efforts (Tan et al.,
    In this pap ...