T-REG

Make a T-Rex walk! Please see this document for a general overview of the project.

This Repo contains two subdirectories:

Repo 1: T-REG-RL

Playground for using reinforcement learning (via the unity-ml-agents package) to make the T-REX walk

T-REG-RL Get Started

• Install Unity Hub (includes signing up to Unity personal)
• Shallow clone the ml-agents repo git clone --depth=1 --branch release_20 https://github.com/Unity-Technologies/ml-agents.git
• Create and activate a conda environment (e.g., python 3.10) using the command: conda create -n treg python=3.10.8
• Follow this instructions to complete installation. See the instructions summary below:

pip install torch -f https://download.pytorch.org/whl/torch_stable.html
pip install -e ./ml-agents-envs
pip install -e ./ml-agents
pip install protobuf==3.20

• Download or clone this repo (e.g., git clone git@github.com:alan-turing-institute/T-REG.git)

To run a sample simulation, follow the instructions below

• Navigate to the T-REG repo and run the simulation. Use the command below:

cd T-REG
mlagents-learn config/ppo/TReg.yaml

Repo 2: T-REG-HMI

Have users control the T-Rex limbs