I want to share my experiences yesterday while testing new fixes to PIR assisted tracking... I was having problems with the PIR assisted tracking because of some hardware issues related to PIR sensors (detailed writeup about the issue is on my other post at: http://myrobotlab.org/content/pir-sensor-retriggering-issues 

yasterday Grog came up with a solution to the problem by adding a configurable Debounce Delay to the py script which will ignore short false triggers generated by the sensor. I upgraded to the latest MRL but had issues with non-Mega Arduino to upload new MRLcomm.ino, Grog once more spotted the problem and Gael and I managed to upload the new MRLcomm.ino to my ProMini and Gael's Uno without problem... Then I tested the PIR assisted tracking adding the new debounce delay feature to the py script. It worked but slowed down the tracking considerably. even adding a small 1000msec delay caused the system to slow down to 2fps facetracking, weirdly like watching my previous moves in front of the cam... However PIR stayed on and tracking did not interrupt... I soon realized that tracking actually never stop, even when no human around, that is, PIR sensing was not working... Seeing that PIR tracking didnot work as supposed to be with the new MRL, I decided to revert to my previous 1881 version (which turned out to be the fastest for my config). After doing that, and testing my last working config, decided to add the new debounce delay to the script on MRL 1881 which has the old MRLcomm.ino and not supposed to work, the debounce delay function worked with the old version too.. However, the tracking was terribly slow as with 1965 and PIR state  was not taken into account once delay is activated. Then I realized that once you activate a debounce delay in py script it stays there and doesnt let the script to monitor PIR status... Also adding debounce delay creates lags which is described in the config... That is, if you set a debounce delay of 10000msec it interrupts the tracking for 10secs... So, it is not useful as is for the time being, and if you add the delay then the tracking is not PIR assisted anymore.... so sad.... :(

hairygael

10 years 10 months ago

Mmmh bad news. I was hoping to get it working with InMoov2.0.

I did see the issue about the tracking not stoping even if the PIR could not detect humans and I thought it was related to my personal script having a written problem. I guess this gives me the answer.

So what next?