Data mismatch between ROS publisher and Simulink subscriber model
2 views (last 30 days)
Show older comments
Pavan Dayala
on 31 May 2019
Commented: Pavan Dayala
on 6 Jun 2019
I'm publishing the data to ros subscriber in matlab and writng the published data and subscribed data to a file respectively. The problem is that the data is not accurate when comparing two files. I think the problem is in matlab to write the ros data to a file with as the ros publishes data in high speed.
0 Comments
Accepted Answer
Cam Salzberger
on 3 Jun 2019
Hello Pavan,
What is the difference between the two? What data is being stored (the whole ROS message, or an extracted data field from it)? Is it simply dropped messages that are published but apparently not seen by the subscriber?
If it is the latter, then I have a suspicion about what is happening. Simulink executes its simulation on a (typically) fixed timestep. Every timestep, the Subscriber block will output the most recently receive message, indicating if it is a new message or one that has been output before. If your publisher is sending data faster than Simulink is running its timesteps, it will not pick up every message. Even if MATLAB is running slower, if the network communication happens in such a way that two messages arrive during the same timestep in Simulink, only one will be output.
Commonly, Simulink is used in feedback control systems or similar, where it needs the latest sensor or position data to determine the next action. From this point of view, always operating on the latest message received makes more sense than operating on all messages received in order. What are you planning on using the Subscriber in Simulink for though, that you require every message?
-Cam
5 Comments
Cam Salzberger
on 5 Jun 2019
Yeah, then that workaround won't do it for you. And the message queue option only applies if it is taking too long to process each message (such that a queue is building up). For Simulink models, the processing time is minimal. The issue is that the underlying subscriber is handling all the messages, but your model isn't looping fast enough to pick them all up. You could try to optimize your model to reduce the processing time that any given timestep takes?
Again, though, it may be that you may not need all your messages. Are you able to answer my earlier questions about the message contents, or is it proprietary information or something like that?
More Answers (0)
See Also
Categories
Find more on Services and Actions in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!