Dashboard Robot SDRead Communication Problems

[nstephenh]

Our robot writes the value it gets from the camera to the dashboard, so we can do vision processing on a laptop. The robot then reads the value from the dashboard and uses that data to tell if the goal is hot.

[Alan Anderson]

Originally Posted by nstephenh

Our robot writes the value it gets from the camera to the dashboard, so we can do vision processing on a laptop.

That is not a very good way to do it. Given the network architecture we use, I would go so far as to call it a spectacularly poor way to do it.

Your dashboard program should be getting the image directly from the camera, without forcing the robot program and Smart Dashboard communication to be an intermediary. The default Dashboard already connects to the camera and reads the image. You just need to feed that image to your vision processing.

[prometheusLKR]

Originally Posted by Alan Anderson

That is not a very good way to do it. Given the network architecture we use, I would go so far as to call it a spectacularly poor way to do it.

Your dashboard program should be getting the image directly from the camera, without forcing the robot program and Smart Dashboard communication to be an intermediary. The default Dashboard already connects to the camera and reads the image. You just need to feed that image to your vision processing.

We tried doing that, and because of the internal buffering of the camera, if we use the dashboard Read MJPG just when we want to get a new image, the image is not current, and if we read from the camera continuously, we have bandwidth issues (even at low resolution and frame rate, it resulted in our robot losing communication with the field at the NC Regional). We seem to have been able to solve this issue by running the "MJPG Loop" vi in Periodic Tasks, and when we need an image, we get the image string and communicate it using a network tables variable. This has resulted in being able to consistently get a current image, something we were never able to accomplish getting the image directly from the camera in dashboard code.

This is, however, not the problem being discussed in the thread, and as an update on that issue (not getting correct values with SD Read Boolean and Read Numeric), we seem to have been able to fix the problem by switching from SD Read and Write vi's to NT Read and Write vi's.

~another programmer from OP's team

[Greg McKaskle]

Just to clarify a few things. The SD and NT functions work almost identically except that the SD functions automatically add "SmartDashboart/" to the variable name in order to keep it separate from the other types of Network Table variables. The storage, transmission approach and other aspects are the same.

I believe the size limit of strings is actually pretty small. It is implementation dependent, but cannot be larger than 65k characters. So in some implementations you may be able to store an image, but that may not work consistently.

I'm not sure I have enough info to give a suggestion on how to setup your camera, but the camera supports both JPG read of a single image and an MJPG session which sends images without asking for each. If using MJPG, you need to read them as fast as they are sent or they will lag behind. You can ask the camera for a very low framerate. Because it was not often needed and caused confusion, I believe the JPG option is in vi.lib, but not in the palettes. If you want to use it, it is in the same folder as the other camera VIs.

Greg McKaskle

[prometheusLKR]

Originally Posted by Greg McKaskle

Just to clarify a few things. The SD and NT functions work almost identically except that the SD functions automatically add "SmartDashboart/" to the variable name in order to keep it separate from the other types of Network Table variables. The storage, transmission approach and other aspects are the same.

I believe the size limit of strings is actually pretty small. It is implementation dependent, but cannot be larger than 65k characters. So in some implementations you may be able to store an image, but that may not work consistently.

I'm not sure I have enough info to give a suggestion on how to setup your camera, but the camera supports both JPG read of a single image and an MJPG session which sends images without asking for each. If using MJPG, you need to read them as fast as they are sent or they will lag behind. You can ask the camera for a very low framerate. Because it was not often needed and caused confusion, I believe the JPG option is in vi.lib, but not in the palettes. If you want to use it, it is in the same folder as the other camera VIs.

Greg McKaskle

OK, thanks. I'll try and see if I can get the JPG option working sometime, although I think we're going to stick with network tables for championships since it's been working consistently and we wouldn't have much time to test out a new approach before championships.