Quote:
Originally Posted by 7h0m54
I've always wanted to try doing something like this. I look forward to testing this on a practice bot while Mechanical and CAD are doing their thing this season.
|
If you have any questions on how to set it up, send me a PM, or better, post it here since someone else might have the same question.
Currently we found the program typically gives good precision (playbacks are generally within about 3 inches of each other) but less than ideal accuracy (though the playbacks are close to each other, they center around a point that may be up to around 8 inches from the spot the driver drove the robot to). Essentially what we found was that it may take several tries to get a recording that accurately reflects the setpoint. However, once the recording is produced, it is usually consistent with itself (barring mechanical problems).
(This is another thing we want to work on, we haven't really messed with the recording and playback rates enough to determine the ideal rates and number of datapoints. The sensor feedback we will likely add during the season, assuming that Kinect doesn't replace autonomous, will also help with accuracy)