Accelerating Visual-Policy Learning through Parallel Differentiable Simulation

In this work, we propose a computationally efficient algorithm for visual policy learning that leverages differentiable simulation and first-order analytical policy gradients.