AGX Dynamics and Machine Learning

These Python scripts illustrate how the AGX Dynamics python API can be integrated into a popular ML/RL environment.

Note that there are additional requirements for running these examples. In the data/python/RL/ you can find requirements.txt

Install these requirements with the command:

> pip install -r data/python/RL/requirements.txt

Simulate and reset

This tutorial demonstrates how to run an AGX simulation for a certain time, reset and restart the simulation. This is a typical scenario when running simulations together with Machine Learning (ML).

The Python script should be started with the python command. It can also be started with AGXViewer, but then it will only load the initial scene and run. To demonstrate the init, run, reset, run feature, you must start it with the python command.

View tutorial_simulate_and_reset.py source code

Multiprocessing tutorial

This script demonstrates how AGX and multiprocessing can be used together.

A Typical use case is running multiple instances of AGX simulations in a Machine Learning (ML) scenario.

This tutorial must be started with the python command.

View tutorial_multiprocessing.py source code

OpenAI Gym and AGX Dynamics

The directory data/python/RL contains a python script that can be used to run the existing training environments.

For more detailed information, see the Reinforcement chapter in the User Manual

Note: We do not support platforms with python versions < 3.7.

cartpole.py source code

pushing_robot.py source code
cartpole_rl pushing_robot_rl