Team 488 using Kinect for autonomous navigation

[JohnGilb]

http://www.youtube.com/watch?v=yD5toEIT-DU

In the off-season, intrepid programmers from team 488 have been working on some fun Kinect projects. Our primary goal - to have an FRC robot navigate a unknown (and potentially changing) environment with no prior knowledge of that environment.

Our setup/process is as follows:
-Kinect running from the robot power supply
-Kinect plugged into a Laptop running RoboRealm
-Roborealm processes the depth image from the kinect and looks for obstacles
-LabVIEW program running on the cRIO queries the RoboRealm server for obstacles
-LabVIEW program determines the importance of the obstacles, based on their distance angle
-LabVIEW program sums all the forces (including the robot's desire to advance) and advances/turns as necessary to progress

In the future, we want to make a number of improvements:
-Replace the laptop with a small form factor (on-board) computer
-Integrate into a field-oriented system to give the robot higher-level goals (going to a specific point, instead of just advancing)

The Kinect has a lot of potential!

[Billfred]

John,

This is pretty awesome stuff! Especially with a goal point, I could see a number of past games where this system could be highly valuable. Keep it up.

[JohnGilb]

Yeah, we're hoping the time investment will pay off next season... but it does require a number of things (an onboard computer, good line of sight for the Kinect) that are expensive from both a cost and design perspective.

We hope to have a white paper and some libraries available for consumption before the end of the year, in case anybody else is interested in going down this path.

Overall, it has been a fun project. Has anybody else been playing around with an onboard Kinect in the offseason?

[Jared Russell]

Very cool. The Kinect is an amazingly effective sensor for mobile robots, especially considering its cost.

Now put a A*/D* path planner on there and go to town :)

[Meshbeard]

Do you think the video processing could be done by a Raspberry Pi, or is it too intensive? If it's not, you should look into some Pis.
http://www.raspberrypi.org/faqs

Team 422 is planning on doing some sort of image processing onboard the robot next year with a Raspberry Pi. Whether that will be Kinect data or just regular video data has yet to be decided.

[Paul 2912]

This may be a simple question, but I'm not able to pull up my program from 2012 Robot Rumble. When I try to open the file it says "Warning - has been deleted, renamed or moved on disk". I'm not sure how the associations with all the files got lost but I'm hoping there’s a way to restore them. There is of course the program that is still working in the CRIO on the robot, if there is a way to pull that back into the laptop?

Thanks