How can the Internet of Things shape the future of our cities and make them smarter? This was my take when hearing about the “Trackathon” organized by Deutsche Telekom AG in Bonn. A team of Splunkers (the “Traffic Ninjas”) participated in wiring up some sensors and collecting data to find answers to the challenge’s main question - “How can transport companies track the mobility flow of their passenger flows via WiFi monitoring?”
The idea behind WiFi monitoring is that most wireless devices like mobile phones send out (beaconing) signals to discover connection points, and in doing so tell other devices that they are there. Effectively, these devices are saying “hi!” to each other. However, due to the specifics of WiFi communication and recently introduced privacy security mechanisms, there are challenges to capturing a meaningful picture of what’s really going on ‘over the air’. For example, channel switching and MAC address anonymization can cause difficulties in getting a meaningful dataset to analyze. With this in mind, the Splunk “Traffic Ninjas” worked with a setup of sensors like the ESP8266 dual Sensor Pack, using NTP time at transmission over a TCP push connection. The following diagram wraps up the whole setup.
But let’s revisit the question and see what KPIs were set for us to analyze:
How many persons per coach are on a train or bus?
How long do they stay there ( what’s the dwell time)?
How many connecting passengers are there that change from one transportation to another?
How many people are on which line, and in which part of the city?
With Splunk, the answers to these questions were clearly displayed by statistics and visualizations, providing the transportation company a real-time view of the inforamtion that's most important to know.
One example of such a finding, is this visualization (Maps+) showing a heatmap of a bus line through the city of Bonn, and how many people are on the bus in certain areas of the city. But how can you tell how many people are on which line? This is quite easy to derive from the underlying statistics, but of course, our “Traffic Ninjas” didn’t stop there. Splunk’s machine learning capabilities allows for a shortcut to set up predictive analytics or time series forecasting using the interactive assistants in the Machine Learning Toolkit. This allowed the team to answer questions such as ‘how is this bus route is used over the year?’ and ‘can we predict an expected volume of passengers in the future?”
After hacking for 24 hours, our “Traffic Ninjas” won 2nd place! Special thanks to Bernd Sommer, Henri Mak, Dirk Beerbohm and Holger Sesterhenn for your passion and curiosity to pioneer the IoT, and pave the way to smarter cities which most of us tend to live in.