[EWCP] Beta Testing Cast with Jared from team 341

[Jared Russell]

Peter:

Definitely the cRIO FRC-II. It is rated for 9-30VDC supply versus 19-30VDC for the original 8 slot model (via the 24V output from the PDB).

As for laptop specs for minimizing lag from the Kinect, unfortunately I can only offer you a couple of data points.

On our Toshiba netbook (very similar specs to the Classmate), we saw on average ~500-1000ms of latency from movement to robot response. With our core i5 laptop with 4GB of memory, we saw ~100-200ms of latency.* These numbers seem to be consistent with others doing the beta test. The difference in latency is tremendous when it comes to controlling the robot at anything above very slow speeds.

Since the "long pole in the tent" is the processing of the Kinect point cloud data (and not the FRC server application, driver station, or wireless link delay), you can actually test a candidate laptop for yourself before kickoff. If you download the Kinect SDK from here (http://kinectforwindows.org/) and install on a Windows 7 PC, you can run the "Skeleton Tracker" sample app and get a good feel for what latency you can expect for a given laptop.

* By the way, with some laptops you will notice much better performance when running on AC power than while on battery (due to power management settings on the PC).

[Akash Rastogi]

Jared, is it true that you will need Windows 7 on the laptop? (I forget where I read/heard that)

[Jared Russell]

(To use the Kinect) Yes. Currently, Windows 7 is all that is supported. I do not know of plans to support any other operating systems at this time.

[Tom Bottiglieri]

Originally Posted by Jared341

Peter:

Definitely the cRIO FRC-II. It is rated for 9-30VDC supply versus 19-30VDC for the original 8 slot model (via the 24V output from the PDB).

As for laptop specs for minimizing lag from the Kinect, unfortunately I can only offer you a couple of data points.

On our Toshiba netbook (very similar specs to the Classmate), we saw on average ~500-1000ms of latency from movement to robot response. With our core i5 laptop with 4GB of memory, we saw ~100-200ms of latency.* These numbers seem to be consistent with others doing the beta test. The difference in latency is tremendous when it comes to controlling the robot at anything above very slow speeds.

Since the "long pole in the tent" is the processing of the Kinect point cloud data (and not the FRC server application, driver station, or wireless link delay), you can actually test a candidate laptop for yourself before kickoff. If you download the Kinect SDK from here (http://kinectforwindows.org/) and install on a Windows 7 PC, you can run the "Skeleton Tracker" sample app and get a good feel for what latency you can expect for a given laptop.

* By the way, with some laptops you will notice much better performance when running on AC power than while on battery (due to power management settings on the PC).

The Kinect Server pipes over a socket, right?

I don't see why the sensor won't be connected to some mega box on the field and pumped to the teams' DS over the network. I don't think each team will be required to supply a Kinect and a beefy laptop for each match. Then again..

[Jared Russell]

Originally Posted by Tom Bottiglieri

The Kinect Server pipes over a socket, right?

I don't see why the sensor won't be connected to some mega box on the field and pumped to the teams' DS over the network. I don't think each team will be required to supply a Kinect and a beefy laptop for each match. Then again..

Yes, the Kinect server uses a simple socket that communicates directly with the Driver Station application. Currently, the Kinect server is launched when you start the Driver Station, which suggests to me that, at this time, it seems like a "bring your own laptop" CONOPS is what FIRST has in mind.

While the arrangement you speak of would make sense, I am not sure that it is in the cards for 2012. Even if it is, you will probably still want a decent laptop for pre-competition testing.

[Joe Ross]

Originally Posted by Tom Bottiglieri

The Kinect Server pipes over a socket, right?

I don't see why the sensor won't be connected to some mega box on the field and pumped to the teams' DS over the network. I don't think each team will be required to supply a Kinect and a beefy laptop for each match. Then again..

In this scenario, teams would have to upload their modified kinect server to the "mega box" on the field.

[Jared Russell]

Originally Posted by Joe Ross

In this scenario, teams would have to upload their modified kinect server to the "mega box" on the field.

That's a great point. The Kinect server application is free to be modified by the teams, which is an argument in favor of the "bring your own laptop" arrangement.