Dataflow-Oriented Robotic Architecture (dora)

A Rust-based robotics dataflow middleware positioned as a lower-latency alternative to ROS.

dora-rs is an open-source robotics middleware framework with its core written in Rust. It splits a robot program into a number of nodes and describes who sends data to whom as a dataflow graph; nodes pass data to each other through shared memory and Apache Arrow (a columnar in-memory data format), avoiding copies wherever possible and thereby cutting transmission latency for large data such as camera images and point clouds. Nodes can be written in Python, Rust, or C/C++, which makes it fairly friendly toward Python nodes running AI models. It targets the same class of problems as ROS 2 — inter-process communication and system orchestration — and provides a bridge for interoperating with ROS 2, and is often benchmarked against ROS 2 on latency.

ExampleA YAML file declares a three-stage dataflow — “camera node → VLA inference node → arm control node” — and dora starts each node and passes images and actions between them.

Also called

dora

Related

Robot Operating System 2 · Middleware · Zero-Copy (Shared-Memory IPC) · Eclipse iceoryx · Publish-Subscribe · Eclipse Zenoh

Sources

dora-rs GitHub
dora-rs 官网 (Chinese)