Vision co-processor Location
 

I been testing some vision using a second laptop on the network wifi running Grip and has been the most successful experience. I tried on roborio and on the driverstation of runing GRIP with undesirable results. So tonight I ordered a Kangaroo mini PC. I wonder do we need to put a network switch on the robot with the Kangaroo or can we install the switch and Kangaroo on the driverstation table. I like the switch and kangaroo on the driverstation table is there reason not to do this? I figured that to be FRC event legal it can't be wifi.

Re: Vision co-processor Location
 

You are correct, it can not be WiFi during FIRST events and likely not during off season events either.

May teams including Citrus Circuits used a Kangaroo this year. I believe they ran it on-board the robot.
With the current radio, there is a second port that allows the co-processor to connect. Alternatively you can put a switch on the robot to allow connecting the Kangaroo and a programming laptop to the RoboRio at the same time.

Re: Vision co-processor Location
 

You can use both the Kangaroo and Switch on the driver station as long as it is in accordance to the driver station rules in the manual. Personally, I would advise putting the coprocessor on the robot itself to avoid FMS bandwidth restrictions. If you do put it on the robot, be sure to pay attention to the FRC regulations for wiring auxiliary processors, i.e. they cannot directly control actuators, and the RoboRIO's ethernet connection must go directly to the router without an intermediate switch.

Re: Vision co-processor Location
 

I understand the FMS limit idea. Two hurtles we have, one we are using an IP camera so not enough ports on curent switch, two robot weight current bot weight is 119.999lbs.

Re: Vision co-processor Location
 

For issue #1:
You can setup your extra switch to attach to the second port on the router. This will provide you with ports for your coprocessor, camera and any other ethernet devices on your robot.

For issue #2:
Weight when it comes to stuff like this is a fairly non-issue. Any reasonable coprocessor and/or switch will not weigh anything more than a pound or two, which even then is on the heavier side of the spectrum when it comes to coprocessor selection.

Re: Vision co-processor Location
 

Helium balloon to reduce robot weight?

As to your original question. We placed a 5 port switch on the robot this year and attached it to the 2nd port on the radio. Other options include USB to Ethernet adapter or Ethernet emulation over USB and connecting the kangaroo board directly to the RoboRIO. You can then attach the network camera directly to the Kangaroo. I do recommend developing a way to use USB cameras in the long run.

If you can't run the Kangaroo on the robot, I recommend using the driver station as your co-processor, assuming it has the power to do so. Do be sure to take into account network lag in your control loops.

Re: Vision co-processor Location
 

Completely agree that the robot side is preferable due to significant expansion of data speed, and reduced latency. In recent years, we've used USB cameras directly connected to a pi for vision processing. If you want to take it to the extreme, only the extracted information (e.g. height, width, and center of goal) needs to go on the network at all.

If you do opt to put vision code on the driver station side, it may be just as effective and much more convenient to install it directly on the driver station computer, especially if you're using a laptop significantly more powerful than the classmate.

Re: Vision co-processor Location
 

If you know how you will be using the data, you could even calculate distance to the target and number of degrees to center of target. Then just send that data. What you choose to send on the network, or via serial of one flavor or another, is all dependent on what data you want, and how you want the data delivered.

Re: Vision co-processor Location
 

On the contrary, if you just send the target location, you will actually be more re-usable. We used our 2013 vision code for targeting in 2014 with minimal modification to the code on the pi; I suspect that we could have simply transferred the pi from one robot to the other. (And yes, the team-specific code was and still is in a publically accessible github repository.)

Re: Vision co-processor Location
 

We used this switch this year with no problems whatsoever. It is 5v, so just power it from the VRM.

A raspberry pi is 0.0683433 pounds. I do not see that putting you over the weight limit, and even if it did, shortening all your wires and zip ties would probably put you back under it. The switch does weigh a little more than the pi, but still not enough to really make a huge difference in your overall bot weight. Overall solution shouldn't be more than a pound.

Re: Vision co-processor Location
 

The Kangaroo does not have a ethernet port? What gives?

Re: Vision co-processor Location
 

Most teams will buy a usb to ethernet adapter to solve this issue. Another option might be this new option from kangaroo. It come with an ethernet port built in.

Re: Vision co-processor Location
 

yep getting one on order.

Re: Vision co-processor Location
 

I wonder if there will ever be different docks available to break out more options.

Re: Vision co-processor Location
 

I'm going to give this a try. Solves two current issues. 1 no ethernet on Kangaroo. 2 using an IP camera, roborio, and co-proccessor with last year's radio. I say last year's because who knows what will be in store for us next year.

https://www.amazon.com/gp/aw/d/B00SA..._st_dp_summary