Guard: A safe reinforcement learning benchmark

The dataset used in the paper is a collection of robot locomotion tasks with various constraints.

BibTex: