Dropbear OS
Dropbear OS is the core runtime that powers every Dropbear robot. Built entirely on ROS 2 Humble, it provides a modular, real-time operating layer for whole-body control, sensor fusion, and high-level task planning.
Architecture Overview
Dropbear OS is organized as a collection of ROS 2 packages, each handling a distinct responsibility. The system follows a layered architecture:
Hardware Abstraction Layer
Direct interface to actuators, IMUs, force sensors, and cameras. Provides standardized ROS 2 topics and services for all hardware. Supports MyActuator RMD series over CAN bus and custom servo drivers.
Control Layer — Corgi
Co-ordinated Robotic Gait and Intelligence. Real-time whole-body controller handling inverse dynamics, balance, and trajectory tracking at a 2ms control loop. This is the brain of locomotion.
Behavior Layer
High-level behavior trees and state machines. Coordinates complex tasks like walking, grasping, and navigation by composing primitives from the control layer.
Intelligence Layer
LLM-based task planning, reinforcement learning policies, and computer vision pipelines. Connects to on-edge inference (Jetson Orin) or cloud endpoints.
Core Packages
The ROS 2 workspace includes the following packages:
| Package | Description | Language |
|---|---|---|
| dropbear_control | Whole-body controller (Corgi), trajectory generation, balance | Python / C++ |
| dropbear_hardware | Hardware interfaces for actuators, IMU, force sensors | C++ |
| dropbear_behaviors | Behavior trees, state machines, task coordination | Python |
| dropbear_teleop | Teleoperation interfaces — web, gamepad, VR | Python |
| dropbear_simulation | Gazebo and MuJoCo simulation environments | Python / URDF / MJCF |
| dropbear_msgs | Custom ROS 2 message and service definitions | IDL |
| dropbear_safety | Safety monitors, emergency stop, joint limit enforcement | C++ |
Installation
Dropbear OS runs on Ubuntu 22.04 with ROS 2 Humble. For simulation-only development, any x86_64 Linux machine will work. For on-robot deployment, you'll need an NVIDIA Jetson Orin or compatible ARM64 board.
# Install ROS 2 Humble (if not already installed)
sudo apt install ros-humble-desktop
# Clone and build
git clone https://github.com/Hyperspawn/Dropbear.git
cd Dropbear
./scripts/install_deps.sh
colcon build --symlink-install
# Source the workspace
source install/setup.bashRunning the Stack
Launch files are provided for common configurations:
# Full simulation with Gazebo
ros2 launch dropbear_simulation simulation.launch.py
# Hardware bringup (on the physical robot)
ros2 launch dropbear_hardware bringup.launch.py
# Teleop control
ros2 launch dropbear_teleop web_teleop.launch.pyConfiguration
All tunable parameters — PID gains, joint limits, sensor calibrations, and safety thresholds — live in YAML config files under the config/ directory. The system uses ROS 2 parameter server, so values can be adjusted at runtime without restarting nodes.
🚧 Active Development
Dropbear OS is under active development. Some packages are more mature than others. Check the GitHub repos for the latest status.