DORA (Dataflow-Oriented Robotic Architecture) is a Rust-based middleware for building real-time, multi-AI, and multi-hardware robotic applications. It targets roboticists and AI engineers seeking a high-performance, composable, and distributed framework for complex systems, offering significant speed improvements over ROS2.

How It Works

DORA models applications as directed graphs (pipelines) of isolated processes called nodes. Communication between nodes leverages Apache Arrow for efficient, zero-copy data transfer via shared memory locally and Zenoh for distributed systems. This dataflow approach promotes modularity and simplifies configuration, with nodes managed by a coordinator.

Quick Start & Requirements

• Install: pip install dora-rs-cli or via cargo install dora-cli.
• Prerequisites: Python >= 3.7, Rust (for building nodes), OpenCV (for camera nodes), CUDA (for GPU acceleration), Zenoh (for distributed setups).
• Resources: Requires a webcam for the YOLO example.
• Docs: Website, Guide

Highlighted Details

• 10-17x faster than ROS2.
• Supports Python, Rust, and C/C++ APIs.
• Extensive node library for hardware, AI models, and algorithms.
• Experimental Zenoh integration for distributed dataflow.

Maintenance & Community

• Active development with recent additions of Phi4, Magma, Qwen2.5, and Zenoh support.
• Accepted into GSoC 2025.
• Community channels: Discord, GitHub Discussions.

Licensing & Compatibility

• Licensed under Apache-2.0.
• Compatible with commercial and closed-source applications.

Limitations & Caveats

• ROS2 bridge is marked as unstable.
• Support for Windows, iOS, and Android is experimental or unsupported.
• Zenoh integration is experimental.