Kyu3224/quadruped-locomotion-policy
By Kyu3224 · ↓ 0 · ♥ 1 · Updated 2025-08-15
What's in this repo
This page lists the files in Kyu3224/quadruped-locomotion-policy, including flat.onnx, flat_robust_100hz.onnx, flat_pos.onnx and more.
- 21 ONNX models (.onnx files)
- 44 files, including 21 .onnx, 20 .pt, 1 .md
- Last updated 2025-08-15
ONNX files
| File | Size | Download |
|---|---|---|
| onnx/anymal_c/flat.onnx | — | Download |
| onnx/anymal_c/flat_robust_100hz.onnx | — | Download |
| onnx/hound1/flat.onnx | — | Download |
| onnx/hound1/flat_pos.onnx | — | Download |
| onnx/hound1/flat_robust.onnx | — | Download |
| onnx/hound1/flat_robust_100hz.onnx | — | Download |
| onnx/hound1/rough.onnx | — | Download |
| onnx/hound2/flat.onnx | — | Download |
| onnx/hound2/flat_robust.onnx | — | Download |
| onnx/hound2/flat_robust_100hz.onnx | — | Download |
Showing 10 of 21 .onnx files — see the full file list on the right.
How to use
These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.
README
🐾 Quadruped Locomotion Policies
This repository contains learned locomotion policies for quadruped robots, trained and deployed in the Isaac Lab simulation environment.
🚀 Overview
Each locomotion policy takes as input an observation vector and outputs joint torques to control the robot.
Flat Terrain Policy
Observation: 48-dimensional vector
Action: 12-dimensional torque command
Rough Terrain Policy
Observation: 48-dimensional base observation + 187-dimensional height map
(Total: 235 dimensions)
Action: 12-dimensional torque command
These policies are suitable for deployment in simulation or transfer to real hardware, and were trained to ensure robustness and agility on both flat and uneven terrains. 🤖 Supported Robots
Each policy is associated with a specific quadruped robot. You can find more information about each robot and its associated policy below:
Hound-1 – https://dynamicrobot.kaist.ac.kr/hound1.html
Hound-2 - https://dynamicrobot.kaist.ac.kr/hound2.html
📂 Structure
policies/ ├── hound1/ │ ├── flat.pt │ └── ... ├── hound2/ │ ├── flat.pt │ └── ...
📎 Notes
Policies are exported in PyTorch format (.pt) and ONNX format (.onnx).
They are compatible with the Isaac Lab environment and can be loaded into simulation scripts directly.
Contact torque-level control is used across all policies.
🛠️ Usage
Example code snippets or instructions for loading the policy into Isaac Lab can be provided here.
Model info
Created: 2025-06-17
Updated: 2025-08-15
View on Hugging FaceFiles (44)
Related models
More from Kyu3224
No other models by this author.