KOS SDK

Software Development Kit (SDK) for KOS robots that bundles together libraries, tools, APIs for developers to build, test, and deploy Z-Bot and K-Bot applications.

Overview

Pre-built modules for sensors, policies, motion planning, perception, simulation, and tests for rapid prototyping, acceptance testing, and problem diagnosis.

Features

• Tests: Comprehensive unit tests for all robot components such as IMU, servos, camera, speaker, microhpone, led.
• Simulation: KOS-SIM compatbility. https://github.com/kscalelabs/kos-sim
• Locomotion: PPO based locomotion.
• Manipulation: ACT and IK based manipulation.
• Perception: Object detection, human pose estimation, etc.
• Algorithms: Motion planning, ZMP, balance control, etc.
• Tools: Keyboard teleoperation, telemtry data logging.

Requirements

• Milk-V Image with kos version 0.6.1
• Python 3.11 with pykos version 0.7.1

Getting Started

1. Set up Git LFS and pull large files

# Install Git LFS
git lfs install

# Pull large files (URDF models, neural networks, etc.)
git lfs pull

2. Clone the repository

git clone git@github.com:kscalelabs/kos-sdk.git

3. Make sure you're using Python 3.11 or greater

python --version  # Should show Python 3.11 or greater

4. Install dependencies

python -m venv .venv
source .venv/bin/activate
pip install -e .

5. Run the tests

make test

Additional Tests

Check that the URDF and MJCF models are realistic:

# To check the URDF model:
ks robots urdf pybullet zbot-v2 --fixed-base

# To check the MJCF model:
ks robots urdf mujoco zbot-v2