SmilingRobo Blog

Training ExoMy Rover in Isaac Gym

By SmilingRobo Team on Sep 13, 2024
SmilingRobo

ExoMy [MarsRover] in Isaac Gym

The setup specifically trains ExoMy to navigate in an unstructured Mars environment.

Demo video

Access it on SmilingRobo Simulations here

Overview

We are thrilled to announce the availability of the ExoMy Mars Rover project, trained in NVIDIA Isaac Gym to navigate an unstructured Mars-like environment. ExoMy is a fully 3D-printed rover, developed by the European Space Agency (ESA), and is inspired by the larger ExoMars rover designed for planetary exploration.

In this project, the focus is on training ExoMy using reinforcement learning (RL) in Isaac Gym to autonomously traverse the challenging terrain of Mars. The project makes full use of Isaac Gym’s powerful simulation capabilities, and it serves as an outstanding example of how advanced simulation technologies can be leveraged for real-world robotics challenges.

About ExoMy Rover

ExoMy is an open-source, 3D-printed rover, modeled after the ExoMars mission rover. It’s designed to be a low-cost alternative that anyone can build, with both the hardware and software freely available to the public. Extensive instructions are provided for assembling the rover, making it an ideal educational platform for hobbyists, researchers, and space exploration enthusiasts.

The rover’s design and functionality make it perfect for simulating exploration in the harsh conditions of Mars. This project builds on that by integrating ExoMy into the Isaac Gym environment, allowing it to learn to navigate autonomously through unstructured terrain.

ExoMy Open Source project here

Simulation Setup in Isaac Gym

In the simulation, ExoMy is tasked with traversing uneven, rocky landscapes that resemble the surface of Mars. The Isaac Gym environment provides a highly detailed simulation that captures the physics and dynamics necessary for a realistic training experience.

Using reinforcement learning techniques, the rover is trained to overcome obstacles and optimize its movement across challenging terrain. This not only improves the rover’s efficiency but also prepares it for potential real-world use in extraterrestrial exploration.

Key Features

  • ExoMy Hardware and Software: Fully open source, with detailed building and assembly instructions. Users can replicate this project by building their own ExoMy rover.

  • Isaac Gym Integration: Isaac Gym, NVIDIA’s high-performance simulator, allows for massive parallel training of robots, making it the perfect platform for training ExoMy in a Martian environment.

  • Unstructured Mars Environment: The simulation replicates Mars’ unstructured terrain, featuring rocky surfaces, slopes, and other hazards that challenge ExoMy’s movement capabilities.

Running the Simulation

To run this simulation and train ExoMy, you need to have NVIDIA Isaac Gym installed. The project includes all the necessary files to get started, and the rover has been pre-configured to handle a variety of terrains. The reinforcement learning algorithms used ensure that ExoMy continuously improves as it encounters new challenges in the simulation.

Once the rover has been trained in Isaac Gym, users can apply these learned behaviors to the real ExoMy rover. This is where the power of simulation meets physical hardware, offering a seamless transition from virtual environments to real-world applications.

Demo of simulation

Access it on SmilingRobo Simulations here

Copyright © 2024 SmilingRobo