Smart Home MEGA Sensor!

[BraveSirRobbin]

Saw this link on the HomeSeer boards and thought it would be of interest to Cocooners.
 
This is a 'mega' sensor being developed that incorporates various inputs such as EMI, heat, sound, etc... and develop profiles that can be recognized for a specific item or event.
 
https://www.engadget.com/2017/05/12/smart-home-mega-sensor-track-whole-room/
 

 

[upstatemike]

I like the potential for the first order synthetic sensors but begin to get skeptical with second order and beyond. An analogy in my mind is if you have a sensor connected to the input of  Stargate or similar controller you can have actions based on changes at the input and have good confidence in the reliability of the system because the controller is monitoring the input directly. If instead you use a PC based system that maintains a state table representing the various inputs it is connected to and then drives actions based on changes in the state table you are one level removed from the actual condition of the input and there is a chance that some error will cause the state table to get out of sync with actual state of the input. This can lead to incorrect activation or failure of activation of actions triggered by the state table.
 
I feel like there is a reasonably high risk that misinterpretation of data from the first order synthetic sensors (maybe multiple simultaneous activities cause one to action to mask another) could lead to incorrect data or actions in the second order (and beyond) synthetic sensors.

 

[kwschumm]

I like the potential for the first order synthetic sensors but begin to get skeptical with second order and beyond. An analogy in my mind is if you have a sensor connected to the input of  Stargate or similar controller you can have actions based on changes at the input and have good confidence in the reliability of the system because the controller is monitoring the input directly. If instead you use a PC based system that maintains a state table representing the various inputs it is connected to and then drives actions based on changes in the state table you are one level removed from the actual condition of the input and there is a chance that some error will cause the state table to get out of sync with actual state of the input. This can lead to incorrect activation or failure of activation of actions triggered by the state table.
 
I feel like there is a reasonably high risk that misinterpretation of data from the first order synthetic sensors (maybe multiple simultaneous activities cause one to action to mask another) could lead to incorrect data or actions in the second order (and beyond) synthetic sensors.

 
To put it simply, if a value is stored in two separate locations and they differ, which value is correct?

 

[BraveSirRobbin]

I see your point Mike.  What happens if someone is standing in front of an item (and they weren't while develping the profile).  Or, the item gets moved, changes operation, etc...  
 
Like you, I'm wondering about accuracy percentage over the long term.  Obviously, you would only want to monitor items not 'critical' and can stand a certain amount of inaccuracy.
 
On the other hand, maybe these profiles that are developed with multiple sensor types are accurate.  Heck, think back five years ago, and we would have never thought voice recognition would be as accurate as we now have with Alexa (this is one of the major reasons I'm getting back into automation as I've always wanted reliable voice control over lights and such)

 

[pete_c]

Here is a quickie clip and post picture of the megasensor.
 
You can DIY something similiar today with an RPi and or Arduino's