Not kidding. You read it right. What if you lost your way from living room to bedroom? AI can help you navigate inside your home. Here is some fun with AI.
The satellite-based Global Positioning System (GPS) has changed our lives. From location tagging, to find shops, to driving using maps to self-driven cars, multiple use cases leverage this “external” positioning system. We are getting addicted to GPS based navigation to the point we are losing our ability to just find a place on our own. Increasingly gone are the days where our mind had a direction and positioning sense in a new city.
Now you reach your home. GPS fails inside your house. Cell signals vary and can’t precisely locate where you are. How do you navigate from your living room to your bedroom? An “internal” positioning system will help.
Be it external or internal, a signal is key. What is an internal signal in today’s world? Of course it is WiFi!!! Electromagnetic waves all around us.
Run a WiFi scan on your mobile and these days you find multiple access points happily beaming a signal to you at various signal strengths. Whether you connect or not the signal exists. That’s the magic key. It is like a fingerprint. Can we build a map of indoors using these signals?
Here are two such scans at one location each in my house.
Quality indicates signal strength your scanner is receiving. Each point has a different number of signals and at different strengths. That’s a vector. A vector of numbers depicting signal strength. In this case they are [6,6,6,7,7,8,8,9,11,11,11,28,30,30,30,30,81,83,87] and [57,8,8,51,8,8,911,10,10,1010,12,30,0]. This is our equivalent of latitude and longitude.
Once there is a vector there is AI… heart skips a beat.
Now walk around the house getting 1000s of these. Add an AI algorithm to train a classifier/ predictor. Now, whenever you are at a point the scanner scans and basis the combination of signals predicts where you are.
Those are the simple steps. Of course one has to ensure data is cleaned, scaled to the same size, take readings at different points in time to remove any vagaries in signal and turning on/off of modems by neighbors, do outlier detection, add some noise for robustness etc. All these go towards improving accuracy.
What do we get?
Here is a vectored map of my house.
Multidimensional vector reduced to 2 dimensions using PCA (Principal Component Analysis)
Here is a house WiFi map with the algorithm drawing boundaries or walls if you will.
Boundaries auto-generated by code
You can observe common walls for example between two rooms. The map roughly (not accurately) depicts the floor plan but haven’t mapped all the rooms hence it’s not complete. The inter-connections are well captured like living room connects to the kitchen directly but not the kid bedroom.
For fun here is the house in 3D vector space.
PCA(3D)… XYZ coordinate map
What do we do with this? The machine can now predict where one is at home. Internal positioning games of-course is possible… a new form of the treasure hunt. It is fun. Try it.
One can tag items to this WiFi coordinate map to create an internal search engine. Don’t lose those keys at home anymore. The internal positioning system can change how you navigate buildings, malls etc.
There is just too much data out there… text, images, videos, genomes and, electromagnetic waves. This is adding to a lot of clutter. Each of this is a vector. How do we leverage these vectors and de-clutter life?
Our life is a vector. Let’s De-Clutter it
In the age of AI… Our life is a vector. Let’s de-clutter it using code & AI.