README.md

Predicate tasks.

This package contains tasks associated with "Behavior Priors for Efficient Reiforcement Learning" (https://arxiv.org/abs/2010.14274), "Exploiting Hierarchy for Learning and Transfer in KL-Regularized RL" (https://arxiv.org/abs/2010.14274) and "Information asymmetry in KL-regularized RL" (https://arxiv.org/abs/1905.01240). This is research code, and has dependencies on more stable code that is available as part of dm_control, in particular upon components in dm_control.locomotion and dm_control.manipulation.

To get access to preconfigured python environments for the tasks, see the task_examples.py file. To use the MuJoCo interactive viewer (from dm_control) to load the environments, see explore.py.

Installation instructions

1. Download MuJoCo Pro and extract the zip archive as ~/.mujoco/mujoco200_$PLATFORM where $PLATFORM is one of linux, macos, or win64.

2. Ensure that a valid MuJoCo license key file is located at ~/.mujoco/mjkey.txt.

3. Clone the deepmind-research repository:

      git clone https://github.com/deepmind/deepmind-research.git
      cd deepmind-research

4. Create and activate a Python virtual environment:

      python3 -m virtualenv box_arrangement
      source box_arrangement/bin/activate

5. Install the package:

      pip install ./box_arrangement

Quickstart

To instantiate and step through the go to one of K targets task:

from box_arrangement import task_examples
import numpy as np

# Build an example environment.
env = task_examples.go_to_k_targets()

# Get the `action_spec` describing the control inputs.
action_spec = env.action_spec()

# Step through the environment for one episode with random actions.
time_step = env.reset()
while not time_step.last():
  action = np.random.uniform(action_spec.minimum, action_spec.maximum,
                             size=action_spec.shape)
  time_step = env.step(action)
  print("reward = {}, discount = {}, observations = {}.".format(
      time_step.reward, time_step.discount, time_step.observation))

The above code snippet can also be used for other tasks by replacing go_to_k_targets with one of (move_box, move_box_or_gtt and move_box_and_gtt).

Visualization

dm_control.viewer can be used to visualize and interact with the environment. We provide the explore.py script specifically for this. If you followed our installation instructions above, this can be launched for the go to one of K targets task via:

python3 -m box_arrangement.explore --task='go_to_target'

Citation

If you use the code or data in this package, please cite:

@misc{tirumala2020behavior,
      title={Behavior Priors for Efficient Reinforcement Learning},
      author={Dhruva Tirumala and Alexandre Galashov and Hyeonwoo Noh and Leonard Hasenclever and Razvan Pascanu and Jonathan Schwarz and Guillaume Desjardins and Wojciech Marian Czarnecki and Arun Ahuja and Yee Whye Teh and Nicolas Heess},
      year={2020},
      eprint={2010.14274},
      archivePrefix={arXiv},
      primaryClass={cs.AI}
}