The aim of this project was to develop a solution that could help SMTPerfomances, an IoT company, to fix missing point in GPS traces. They had an IoT device, the Pegase, measuring the various data on the vehicle (position, speed, etc…) but the data were sometimes lost, resulting in “holes” in GPS traces
As a team of students, we had the mission to help the company fixing these holes. It led to some nice work with API and geographical data.
Position estimation
Physical resolution
The first aspect of this problem was that holes were of different sizes: from a few tens of meters to several kilometers. For the small holes, we had enough data to apply a physical resolution.
The key was to integrate several sources of data: position and speed namely, to have a coherent model of the system. We did not any Kalman filter, as data were quite accurate, and a liner model proved enough to determine the missing data with a high accuracy.
Google Maps API
The second aspect of this project was that some data were missing for several kilometers. As the vehicles were rarely moving faster than 100 km/h, the physical resolution would not be accurate. The strategy used was to call an external API, namely Google Maps, from the script, providing the point of the trace around the hole. It was efficient around large holes, and was as well a good way to detect not precise traces.
Data completion
While the two previous systems were able to complete a GPS trace, the speed and acceleration data were still missing. By coupling the data received by Google, with a low accuracy to reduce costs, with Open Street Map (free and high accuracy), we were able to get much higher precision in traces while cutting down costs. Then, by inputting the data into the physical model we could infere the remaining missing data: speed and acceleration.
Conclusion
The project was received favorably by the company, and lead to further discussion on the actual implementation of the solution, its cost and its effectiveness. It was a great occasion to explore the field of geographical science.