Skip to main content

Production Patterns / Reference Architectures

Production patterns covers the middleware, runtime stacks, and reference architectures used to actually ship robots — message buses, motion planners, navigation stacks, fleet orchestration, and the observability tooling that keeps deployed systems debuggable. It is the layer between a trained policy and a robot a customer can use.

From an engineering standpoint, this is where the cost of operating a fleet is set. Choices about middleware (DDS profiles, QoS), real-time guarantees, security (authentication, encryption on the data bus), and telemetry shape on-call load, incident response time, and the ability to ship updates safely. Most of the reliability differences between robotics teams trace back to decisions made — or skipped — at this layer.

When choosing components, anchor on a stable middleware (ROS 2 with a supported DDS vendor is the default), pick planners and navigation stacks that match your kinematics and environment, and invest early in observability (structured logs, replayable bag files, dashboards). For multi-robot deployments, fleet orchestration and security profiles stop being optional — design them in from the start.

Start here

ROS 2 is the default backbone: production-grade middleware with QoS, security, and real-time profiles, and the substrate most other entries in this list build on.

  • ROS 2 — De facto middleware for production robot software, with QoS, security, and real-time profiles.
  • NVIDIA Isaac ROS — GPU-accelerated ROS 2 perception and navigation packages for production robots.
  • MoveIt 2 — Production-grade motion planning framework integrated with ROS 2.
  • Nav2 — ROS 2 navigation stack with behaviour trees, planners, and recovery patterns.
  • Open-RMF — Open-source framework for multi-robot, multi-vendor fleet orchestration.
  • DDS Security (OMG Spec) — Reference standard for authenticated, encrypted robot data buses (used by ROS 2).
  • Foxglove — Observability and visualisation stack for robotics telemetry, logs, and replay.
  • micro-ROS — ROS 2 client stack for microcontrollers and embedded real-time robot components.
  • ros2_control — Standardized hardware abstraction and controller framework for production robot actuation.
  • BehaviorTree.CPP — Widely adopted behavior-tree runtime for deterministic task orchestration.
  • Eclipse Cyclone DDS — High-performance DDS implementation commonly deployed in ROS 2 production stacks.
  • Fast DDS — Industrial-grade DDS middleware with configurable QoS and security support.
  • MCAP — Modern log container format for robotics telemetry, replay, and long-term data retention.
  • Zenoh — Data-centric middleware for distributed robotics over constrained and heterogeneous networks.
  • rosbag2 — Standard ROS 2 recording and replay pipeline for debugging and incident analysis.