Nowadays, people have more and more devices which are connected to the network. These
devices are with us most of the day and can be easily accessed providing numbers of different
functions. One of these is the GPS, which makes it possible to receive the signals of satellites
and locate a position or design a route. However, we have these intelligent technologies,
using the mentioned function indoor is still in question. Unfortunately satellite sign can not
be sensed indoor, but in many cases it would be very useful to reach its functionality. One
of the cases is navigating in parking garage. It might be wondering, why is it important to
plan a route in a parking garage? The answer is more simple than we thought, because if the
garage has an average occupancy, then the time until a car navigate to the best, free place
(for example, near to the entrance of the cinema) by own, it emits significantly more harmful
material than navigated by computer aided algorithm. Not to mention the fact that its own
costs and extra stress can be reduced.
Another feature besides the GPS is the WiFi, which is also available on most mobile devices
and does not require administrative subscription. With this technology and using a well
optimized algorithm, our system can compete with the benefits of the GPS functionality.
My essay based on the Alle shopping centers WiFi supported parking garage system called
iParking. I made simulations and tested custom positioning algorithms. My final goal was to
refine positioning algorithm estimates and to receive demonstrable better results.
From the commonly used methods I choose the fingerprint technique group for starting
point. I chose Matlab Works for the simulations, which has a lot of positive facility such as
K-nearest-neighbor search (KNN-search) algorithm, which use the pattern-fitting method
in order to find the best fitting received signal strength values and estimate the actual
mobile terminal position. The signal level database (fingerprint) was filled up with real
measurements, which are serving as reference points for the positioning algorithm. In my
essay I worked out and presented estimating improver methods, which are extensions of
the base KNN search for receiving better results. By calculating the average of the returned
positions of the KNN algorithm, using location weighs, route mapping methods, I received
better results. Next idea was to use memory and limitation of decisions by distance. In order
to evaluate the efficiency of the extension methods I introduced an index number to show the
discrepancy of the results.
Using the above mentioned methods, I compared the simulations and measure results using
different parameter setups. By analyzing the variations of the results, I made very interesting
conclusions about the presented indoor positioning methods.