Pathologies and Challenges of Using Differentiable Simulators in Policy Optimization for Contact-Rich ManipulationShare on Twitter Facebook LinkedIn Previous Next