Lumos Technology Series: Self-Learning

Personalisation is one of the most pleasant gifts of technology in the recent years. Right from our smartphones to restaurants and daily deals, we love to be understood and receive a service just the way we wanted. But strangely enough this is something that’s completely amiss when it comes to our homes. Even though we spend almost 60–70% of our lives in our homes, our homes still have no idea what to do the next morning for us to be comfortable.

With Lumos Smart Switches, we intend to change this. Lumos Switches not only can take decisions on pre-set factors, but also learn from users to understand their behaviour. The fact that we had a very limited amount of on device memory in each switch, implementing this Self-learning was an interesting technical challenge. We had to re-think our data structures and the switch-server interaction to come with an optimum solution that doesn’t eat away a ton of data while giving the most personalised experience.

After all the work, this is where we stand today:

  • Each switch uses less than a few mbs of data a month. Translate that to 70 switches, we are talking about roughly 200Mb of data in a month for the entire home which is less than what your phone uses to probably just update a couple of apps.
  • The entire time taken for data to travel from your smartphone to our servers and back to the switch takes less than a second typically.
  • Each switch responds to change in conditions in milliseconds. (You will frequently experience, that the moment you close your windows during the day, your lights will automatically go on!)
  • Complex decisions like what temperature the A.C should be or the speed for the fan is decided by the server at lighting fast rate and communicated to device in a few seconds.

There are several other optimisations that we have done on the switch level to ensure the best experience. Feel free to ping us at contactus@getlumos.co to know more about them.

Follow us at Medium & Twitter to stay updated with our work!