Foxglove bridge on lower bandwidth

[tomdeblauwe]

Hello,

I am using foxglove bridge 0.8.5 (because that is in meta-ros for jazzy) and I was wondering about a feature that would be very helpful on low bandwidth.

Now, afaik, if you subscribe to a few topics, then the bridge will push the messages over the websocket as soon as they are received on the bridge. But, when you have low(er) bandwidth, it's not working very well. If you have for example an imu topic at 200hz and a pointcloud topic at 10hz and a jpeg topic at 1hz. If you are checking the situation on a robot for debugging, it is no use to have, per second, a lot of imu and pointclouds topics coming in, which are then choking the bandwidth for the jpeg image, resulting in no jpeg image update. So it would be great that you could get, on demand, 1 imu update + 1 pointcloud update + 1 jpeg update. So what would actually be nice is that there is like a "step" button in the live view, and when you press it, it downloads for each subscribed topic 1 new message and then goes to "pause" again. This would really be helpful in low bandwidth conditions.

I could of course put some node just before the foxglove bridge, that then publishes some topics on-demand when I call a service on that node. But it would be nicer if it would be in foxglove I think because it know better about the bandwidth available in the send buffer.

I saw that there was already a similar discussion, but it was more about trying to do things automatically. However, maybe a step button would be simpler to implement and be very useful already.

Best regards
Tom,

[gasmith]

I could of course put some node just before the foxglove bridge, that then publishes some topics on-demand when I call a service on that node.

This sounds like the right way to go for now. Or alternatively a node that throttles output down to a desired publication frequency.

One of the things we've talked about is the ability to configure that kind of throttling natively, as something that could be either configured on the device, or controlled dynamically from the client side. Maybe both! Would that be useful in your situation, or is there a hard requirement on the stepping mechanism you've described?

[tomdeblauwe]

Yes, controlled from the client side sounds good to me, then the client could request topics on-demand in some sort of round robin style. Of course, a pause button would be great for instance when you see a situation that you want to analyze further and don't need updates anymore because that would interfere with your analysis. A step button would still be nice to have I think, but I mention it just as an "easy" first solution to the problem. By the way: the user interface elements are actually already visible in foxglove 😅 but they are greyed out in live view.