Breadcrumbs Beta 1

Micro controllers
Vector graphics

For about 2 years I have had Google Location History active on my cell phone. This has so far collected over 680 thousand data points, and I have now (finally) gotten around to writing a small c++ program to visualize this data set as a kind of density map.

Collecting data with location history is very convenient (as I would typically carry my cell phone around anyway), but there are some limitations in the data set which affect which types of data visualization that can be made to work. For example the sample rate is rather slow, usually one point per minute. Most of the collected data points are actually not from the GPS. Instead the location data is more often based on wireless networks. Continously running the GPS all the time would simply drain the battery of the cell phone too quickly. An advantage of wireless network based data is that location data is generated even when you are indoors, provided that there are wireless networks in range. This property makes the data well suited to be presented as a density map. The wireless location data typically has lower precision than the satellite based data. I had initially assumed that the errors would be uniformly distributed, but this has turned out not to be quite true. For example, close to the river you can get systematic errors, presumably because all wireless networks in range are to one side. In areas where there are no wireless networks nearby you will end up with voids (or near voids) in the data. Some of the errors are less easy to explain, but sometimes rather amusing (No, I did not pop out for a quick lunch in Warsaw, and I am pretty sure I didn't sleepwalk to Saudi Arabia). I have filtered out the most obvious artifacts in the data, and lately I have used My Tracks or Android GPS Test to improve the quality of the data, when moving about.

The visualization program is at the moment rather basic, but the result so far can be seen in the images below. Clicking an image takes you to a larger version of that image.

breadcrumbs beta 1:6
breadcrumbs beta 1:5
breadcrumbs beta 1:4
breadcrumbs beta 1:3
breadcrumbs beta 1:2
breadcrumbs beta 1:1

This project is not intended to be about surveillance or privacy. Location History is an opt-in service, the accumulated data is available only to me, and the data can be erased as soon as I have made the KML export. The visualizations I present here are also rather abstract. No coordinates are presented, nor are any times presented. Rather than being about surveillance/privacy, this project is about my facination with patterns, and I have been interested in quantified self style projects long before the first time I ever heard the term quantified self.