I have GPX traces and want to remove outlying observations. There are some route points that are impossible because it would require acceleration that the vehicle can not achieve.

Are there good algorithms for filtering out such extreme noisy observations?


2 Answers 2


Here are two references for linking GPS data to street centerlines:

Schussler, N. & Axhausen, K. (2009a), ‘Map-matching of GPS traces on high-resolution navigation networks using the multiple hypothesis technique’, Working paper 568. IVT, ETH Zu ̈rich, Zu ̈rich.

Schussler, N. & Axhausen, K. (2009b), ‘Processing raw data from global positioning systems without additional information’, Transportation Research Record: Journal of the Transportation Research Board 2105, 28–36.

Reference link (for those with access).

I know SF City implemented one of these algorithms to process gps tracks from a smart phone app they are using to track cyclists (cycletracks); and they released all of the code under an open-source license (because they are wonderful!). I think their issue was mainly to do with urban-canyon inaccuracies; I'm not sure if they had a problem with acceleration also.


A Kalman filter may be what you want: it takes into account predictions based on a physical model of the system. (E.g. no 10000mph cars!)

Answers to the Stack Overflow question "Smooth gps data" provide links to implementations such as ikalman github repository, as well as other approaches.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.