Classmate Latency and Live Video

[slavik262]

Our team ran a mini-regional today in Sussex, WI, complete with FMS and the works. While monitoring the latency between the FMS and the robots, we found that those streaming video to dashboards running on the Classmate had latency times from 120 to 200 ms, while those not streaming video to the Classmate enjoyed latency times around 30ms.

While I realize the numbers may be different in an actual competition, the moral of the story seems that without modifying the dashboard to lessen load on the Classmate's CPU or running the dashboard on another laptop as has been discussed elsewhere on this forum, running video live to your robot has a significant cost in latency. Keep in mind that this latency was from the Classmate back to the robot as well as from the robot to the Classmate, so teams running video on the Classmate must decide whether it is worth somewhat of a lag in robot control.

[Greg McKaskle]

The latency should not be as described, and previous testing with FMS lite was always done with the dashboard video turned on. We will try to reproduce the results you describe.

Greg McKaskle

[slavik262]

Originally Posted by Greg McKaskle

The latency should not be as described, and previous testing with FMS lite was always done with the dashboard video turned on. We will try to reproduce the results you describe.

Greg McKaskle

I'm sorry if the data isn't very reproducible; I understand that it could be a wide variety of things causing the delay. Seeing as we didn't inspect the code of the other teams, I don't know what resolutions they were using, what other data they were sending, etc.

[Greg McKaskle]

That is fine. I'm hoping and suspecting that it is something else, but I'd rather be certain, so we will try to reproduce it.

Greg McKaskle

[~Cory~]

If we can be any help 706 was at the mini and we did have latency issues. we fixed the problem (well we think we did -- not 100%) by patching the driver station. I can get more info from our lead programmer if you guys need it

[Greg McKaskle]

Using FMS Lite, we were unable to reproduce lag using either video or video back to the PC.

Oddly enough, we did sometimes see quite a lot of lag reported at the FMS. Running from source, we were able to see that the FMS value was correct, and ultimately this is related to the I/O board attachment. At this point, I think we can safely say that the lag exists on the status leg of the control loop and has very little affect on the control packets that control the robot. We are still looking into how the I/O driver is causing UDP delay.

Greg McKaskle

[slavik262]

Awesome. That would explain things since the lag was being reported by the FMS on our end as well. We'll look a little more into it, but your results are quite reassuring. Thanks again.

[DRH2o]

I can say that I know of two teams ( ours included ) that did have lag controlling the robot in our labs using the basic router bridge setup. We were streaming video to the classmate. The lag stopped when we disconnected the ethernet cable from the camera. We are not using FMS.

[camtunkpa]

We are also experiencing lag in our robot control system while running video feed and not connected to any FMS software.

[MishraArtificer]

Would killing off background processes in Windows XP help with the latency any? I have a good-sized list of things you can do to Windows XP to make it go faster and ease the load on the processor, from mundane things like turning off the Print Spooler (since there's no printer attached anyway), to more complex stuff like registry edits. I work at a computer repair shop here in Monroe, and I've used most of it on all of my computers, my parent's computers, as well as some customer's laptops and desktops.

[slavik262]

Originally Posted by MishraArtificer

Would killing off background processes in Windows XP help with the latency any?

The bottleneck seems to be the processing speed of the Classmate processor itself, so yes, shutting down some background stuff would probably help.