ROS 2 DDS vs Zenoh for Robotics Middleware Communication

[ROS 2 DDS] excels as the established, plug-and-play standard for robotics middleware because it is deeply integrated into the ROS ecosystem. For example, a standard ROS 2 talker/listener pair using Cyclone DDS over a reliable Gigabit Ethernet link can achieve sub-millisecond latency with near-zero configuration, making it the default choice for single-robot systems or lab environments where network conditions are pristine and deterministic.

[Zenoh] takes a fundamentally different approach by decoupling the data protocol from the discovery and transport layers, optimizing for minimal wire overhead. This results in a significant trade-off: Zenoh demonstrates up to 10x higher throughput and 50x lower latency than standard DDS implementations over lossy, high-latency wireless links like Wi-Fi 6 or 5G, but it requires integrating a non-native middleware bridge (rmw_zenoh) into the ROS 2 stack, adding initial architectural complexity.

The key trade-off: If your priority is seamless ROS integration and deterministic performance on a single, wired robot, choose ROS 2 DDS. If you prioritize high-throughput, low-latency communication across a distributed fleet of robots operating over unreliable wireless networks, choose Zenoh. Consider Zenoh when your scaling roadmap includes multi-robot swarms where bandwidth is the primary bottleneck.

[ROS 2 DDS] excels as the default standard for single-robot systems because it provides a mature, plug-and-play ecosystem with deep integration into the ROS 2 build, discovery, and tooling layers. For example, a standard industrial manipulator using ROS 2 Humble with Cyclone DDS can achieve reliable intra-process and local network communication with near-zero configuration, leveraging the built-in ros2 topic and ros2 bag tools. This results in a significantly lower integration burden for teams using off-the-shelf ROS 2 drivers and nodes, where the primary concern is reliability over a wired Gigabit Ethernet connection rather than extreme throughput or wireless resilience.

[Zenoh] takes a fundamentally different approach by decoupling the data exchange protocol from the ROS 2 graph layer (rmw_zenoh). This results in a superior architecture for wireless, multi-robot, and high-throughput edge systems. In benchmark tests over lossy Wi-Fi 6 links, Zenoh has demonstrated over 5 Gbps throughput with microsecond-level latency, dramatically outperforming standard DDS implementations that suffer from discovery storms and head-of-line blocking under packet loss. Its ability to natively bridge data across network boundaries—from a robot's compute board to a cloud dashboard—without complex DDS routing configuration makes it a compelling choice for Autonomous Mobile Robot (AMR) fleets and distributed sensing networks.

The key trade-off: If your priority is rapid development, maximum driver compatibility, and a battle-tested standard for a single, wired robot, choose ROS 2 with a tuned DDS implementation like Cyclone DDS. If you prioritize high-throughput data streaming over lossy wireless links, transparent multi-robot communication, and seamless cloud-to-edge integration, choose Zenoh as your middleware transport. Consider a hybrid approach where Zenoh serves as the WAN and inter-robot backbone, while standard DDS handles the deterministic, high-frequency control loops inside a single robot's compute cluster.