This repository contains a customized implementation of Mujoco's Arm26 model, that can be used with DeepMind's ACME framework to train a Reinforcement Learning agent. This is a more biologically realistic and more challenging effector to train than the typical reacher environment included in the original package.
This code is provided as-is, but feel free to log any issue if you find one. Any contribution is greatly appreciated.
The model relies on dm_control , as well as any dependencies dm_control may have, most notably a working Mujoco installation.
Import to your code via a normal import command import arm26. Make sure arm26.xml and arm26.py are in
the same folder.
You can create the enviroment by calling
import arm26
environment = arm26.load(task_name)
The syntax is the same as for dm_control.suite.load, with the first argument (domain_name) being omitted.
Below are examples showing that each of the six muscles work properly.
