Both the design and control of a robot play equally important roles in its task performance.
However, while optimal control is well studied in the machine learning and robotics
community, less attention is placed on finding the optimal robot design. This is mainly
because co-optimizing design and control in robotics is characterized as a challenging
problem, and more importantly, a comprehensive evaluation benchmark for co-optimization
does not exist. In this paper, we propose Evolution Gym, the first large-scale benchmark for …

In this section we describe Evolution Gym's simulator in detail. We describe the simulator's
representation of objects, the dynamics of the underlying simulator, our implementation of
contact forces, and other techniques we use to improve the quality of Evolution Gym. Finally,
we present an analysis of how our simulation scales with the number of voxels, followed by
some of the hyperparamters in our implementation.