Adroit Hand¶

This environments consists of a Shadow Dexterous Hand attached to a free arm. The system can have up to 30 actuated degrees of freedom. There are 4 possible environments that can be initialized depending on the task to be solved:

AdroitHandDoor-v1: The hand has to open a door with a latch.
AdroitHandHammer-v1: The hand has to hammer a nail inside a board.
AdroitHandPen-v1: The hand has to manipulate a pen until it achieves a desired goal position and rotation.
AdroitHandRelocate-v1: The hand has to pick up a ball and move it to a target location.

A sparse reward variant of each environment is also provided. These environments have a reward of 10.0 for achieving the target goal, and -0.1 otherwise. They can be initialized via:

AdroitHandDoorSparse-v1
AdroitHandHammerSparse-v1
AdroitHandPenSparse-v1
AdroitHandRelocateSparse-v1

Reference¶

These environments were first introduced in “Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations” by Aravind Rajeswaran, Vikash Kumar, Abhishek Gupta, Giulia Vezzani, John Schulman, Emanuel Todorov, and Sergey Levine. Which can be cited as follows:

@article{rajeswaran2017learning,
  title={Learning complex dexterous manipulation with deep reinforcement learning and demonstrations},
  author={Rajeswaran, Aravind and Kumar, Vikash and Gupta, Abhishek and Vezzani, Giulia and Schulman, John and Todorov, Emanuel and Levine, Sergey},
  journal={arXiv preprint arXiv:1709.10087},
  year={2017}
}