Before anything else: my intention here is not to turn this thread into a complaint fest.
DDS and ROS 2 are topics that periodically raise questions for me - and for others as well - so I’d like to collect and consolidate existing insights. Positive experiences, success stories, and best-case user datapoints about DDS are especially welcome!
My goal is to gather constructive feedback or identify where such information has already been summarized. I noticed several ROSCon DE 2024 talks in Strasbourg touched on related challenges, which inspired me to start this thread.
My question: What solid resources already exist - talks, guides (e.g., Clearpath’s ROS 2 networking material), previous discourse threads, etc. - that I could collect into one place?
I’d really appreciate neutral or positive, technically grounded feedback.
Thanks in advance for your help!
DDS is more like a protocol with an open specification of it, and we, as a users, can choose which vendor implementation to use (Cyclone DDS, Fast DDS, rtpstalk, …).
In case of Zenoh (@kydos please correct me) there is only one vendor right now - ZettaScale
Great initiative! Adding to the list of resources, I’d like to share some insights from our research group (DGIST CSI Lab).
While the middleware landscape is heavily driven by brilliant developers, our group (about 20 members) approaches these challenges from an academic perspective—focusing on mathematical modeling and experimental validation to support the community.
Here are three resources from our recent work that tackle ROS 2 wireless robustness:
Theory (PLA): A probabilistic model that turns network tuning from guesswork into math, accurately predicting latency and jitter.
Practice (Large Data): A drop-in XML profile we developed to solve bottlenecks (IP fragmentation, buffer bursts) for heavy sensor data over Wi-Fi (stable 30Hz even with packet loss).
We are continuously researching middleware, including DDS and Zenoh. Please stay tuned for our upcoming research on DDS discovery analysis and novel security vulnerabilities.
We are always open to collaboration or technical discussions!
Thanks! To answer your question: while there is a minor overhead in the publish process time, the trade-off is acceptable for the performance gains.
We are currently preparing a detailed End-to-End guide covering ROS 2 application optimization, multi-pub/sub scenarios, and socket buffer sizing. We fully empathize with the struggles many robot developers and researchers face with complex network stacks. We are working hard to release this to the ROS community as soon as possible!
Cheers from Franconia