For mobile network analysis it is often desirable to classify some measurements to an indoor or outdoor environment. For example to determine the quality of service for voice calls in conjunction with the coverage it is very helpful to take the indoor/outdoor scenario into consideration. Furthermore for planning the network capacity it is interesting to take a look at the distribution of the data traffic, which could be mainly indoor or outdoor, depending on the season and the time of day.
The mobile devices – our mobile measuring instruments – offer the necessary kinds of sensors to do such a classification. The classification algorithm consists of a decision tree that evaluates the indoor / outdoor state based on a wide variety of environment information from the device. In this post we will focus on the impact of the light sensor to determine the most likely state (indoor vs. outdoor) based on a probabilistic model.
On the following chart, you can see the probability density function by using the light sensor for the classification process. How you can see, this sensor information can contribute to an indoor/outdoor decision. But note that this procedure is only practicable for hours with daylight. Also in daylight, a certain range of light sensor values is not usable for the classification, because of the difference in the probabilities, which is not significant enough to make a reliable decision.
The results of the indoor/outdoor classification taking all the different information sources into account can be evaluated in a typical day curve as shown in the following figure. Depending on the time of the day and season the outdoor probability is between 1% in night hours in winter and nearly 16% in daylight hours in summer.
Furthermore the distinction of the indoor/outdoor state for some interesting KPIs like the speedtest transfer rate (download) or the average of the signal strength in voice calls is possible. How expected in the outdoor case the average of the transfer rate and signal strength is much better for mobile networks.