Revision history [back]

Integrating acceleration twice to get position is terrible. I think you're discovering just how bad it actually is.

additionally, your current approach integrates the acceleration without taking into account any possible rotation of the sensor, which means that your results will probably be terrible if the sensor rotates at all.

If you have gyro or compass data from the 3dm_GX2, you may be able to feed all of your data into something like a kalman filter to get better results.