[DjScribbles]

I attempted to help our team set it up to be used at Traverse City, however, and we were even given some time on the field during lunch to test it (because there were problems the first day with us using a USB hub, apparently that was not working with the extra cable length, we were unable to test during it on setup night) (I'd like to thank the field personnel for trying to accommodate us setting it up).

However, I was severely disappointed in how it functioned, and I can't say for sure if it was due to the setup at the field or just the Kinect in general, but this is what I was seeing:

Reliability: I was frequently seeing a loss of the control enable signal (button 9) which indicates that the users arms are in valid position (extended in the same xy plane). I am a bit suspicious that this could be due to cable length.

Sensitivity: The controls I had setup the robot to use (left/right leg forward/backward) seemed extremely sensitive, making them essentially impossible to control. Depending on the posture I stood with, I would see the buttons for leg forward/backward activating sporadically.

Speed: Our driver station (not a classmate, but it had an atom) was abysmal for controlling using the kinect, it was processing 4 or 5 FPS max. We had to setup my programming laptop to get decent results (which left me unable to make programming changes outside of extended periods of downtime) and was draining the battery much more quickly, so we eventually were going to need an external power supply installed on the field.

Height: I'm about 6' with long arms, and the Kinect was barely able to pickup my elbows while I stood in the box.

The above issues pretty much killed all excitement about using the Kinect for me. When you can't drive the robot correctly because control is constantly cutting out, and the robot is picking up different gestures based on you slightly changes in stance, and it's not picking up your full range of motion...

I wanted to try out using different gestures for control (legs and head to the sides instead of legs forward/back, along with some logic to continue the last input when the control enable signal vanishes), but if that doesn't work I feel I will need to recompile the KinectServer in order to get it to be less sensitive.

All of this combined with the fact that we were doing really well in standard autonomous, kinda made me think it wasn't really worth fussing with anymore.
I'm hoping to do some off season projects with it once we can permanently unbag the robot, but until then there are lots more things to worry about.