Raspberry Pi as Coprocessor

Technical Control System
1

We plan on using the Raspberry Pi to do our vision processing. How would we output Python script results (after processing) and receive on RoboRio (Java)? I looked into pynetworktables and am confused on how to receive the data from the RoboRio.

2

GRIP has some examples. Specifically samples/frc_find_red_areas.

Basically, you just publish your results to NetworkTables with the pi connected to the roborio/radio/switch with an ethernet cable, then you can read that from your robot program.

3

Hi, I am working on the same thing, but in Java. Is there documentation for Network tables in java?

4

Javadocs here

5

Thanks, could you take a look at this code I found on another thread:


//========================================
//Robot Code
// declare object
NetworkTable cameraTable;

// getinstance (table is a singleton, if one does not exist then one will
// be created otherwise returns reference to existing instance
cameraTable = NetworkTable.getTable("camera");

// get data element from the table
// default values returned if no data has been provided from another source yet

cameraTable.beginTransaction();
Boolean imageFound = cameraTable.getBoolean("found", false );
Double imageOffset = cameraTable.getDouble("offset", -1.0);
Double imageDistance = cameraTable.getDouble("distance", -1.0);
cameraTable.endTransaction();

System.out.println( "found = " + imageFound);
System.out.println( "offset = " + imageOffset);
System.out.println( "distance = " + imageDistance);

//=======================================
// Driver Station PC
// camera image processing is done in C# application on DS

// declare it
private NetworkTable table;

// Init NetworkTable
NetworkTable.setIPAddress("10.6.23.2");  // ip of crio
table = NetworkTable.getTable("camera");


// in the image processing loop

table.beginTransaction();
table.putBoolean("found", true);
table.putDouble("distance", distance);
table.putDouble("offset", offset);
table.endTransaction();

Looks like what I need, but is it missing something (ie setTeam, etc.)?

6

That example is 4 years old. All you have to do on the pi:


NetworkTable.setClientMode();
NetworkTable.setTeamNumber(1512);
ITable visionTable = NetworkTable.getTable("vision");

That’ll set up the pi and a table specifically for vision. If you launch OutlineViewer from Eclipse (WPILib > Launch OutlineViewer), you’ll be able to see the values update in realtime.

Then to publish the results from the vision processing


visionTable.putBoolean("Have target", haveTarget);
visionTable.putNumber("Angle to target", angleToTarget);
...

In your robot program, you’d also declare a table for vision just like we did on the pi. To read the values the pi is publishing, use the corresponding ‘get’ methods


haveTarget = visionTable.getBoolean("Have target", false); // default to false
targetAngle = visionTable.getNumber("Angle to target", 0); // default to 0
...
1 Like
7

Thank you so much!

8

Does anyone know of a vision processing tutorial?

Sent from my Pixel using Tapatalk

9

Look up PyImageSearch. Extremely useful!

Sent from my SM-N920V using Tapatalk

10

I thought the Pi had issues running GRIP, are those gone now?

11

We are not using GRIP but OpenCV

12