Overview

RoboHive, a modular framework for research in the field of Robot Learning and Embodied AI. RoboHive ecosystem encompasses a range of pre-existing and novel environments, including dexterous manipulation with the Shadow Hand, whole arm manipulation tasks with Franka and Fetch robots, and various quadruped loco- motion tasks. In comparison to previous works, RoboHive offers a streamlined and unified task interface, utilizes the latest simulation bindings, features tasks with rich visual diversity, and supports common hardware drivers for real world developments. The unified interface of RoboHive offers researchers a convenient and accessible platform to study a multitude of learning paradigms such as imitation, reinforcement, multi-task, and hierarchical learning. Furthermore, RoboHive includes expert demonstrations and baseline results for most environments, providing a standard for benchmarking and comparisons.

Features:

• Most extensive and diverse task collection

• Fully customizable visually rich task designed  for behavior generalization.

• Rewards agnostic task success metrics 

• Support for diverse algorithmic families + pre-trained baselines

• Sim and hardware agnostic robot class for easy transition between sim <> real

• Teleoperation support. Human+Expert dataset