Could HLT use the gyroscope output from the iPod Touch 4g and iPhone 4? That would give you full inertial navigation functionality. If that can update at 100Hz then it might be better to calculate position and velocity with INS and correct for drift with GPS rather than just using GPS. Even if you don't use it, it would be nice to have the data included in the output file so it could be used with another program.
How does the recalculate function work? Do you throw out the accelerometer data completely or do you correct it by comparing GPS calculated accelerations with the accelerometer readings? Presumably the accelerometer readings are only off because of a problem with the calibration. For example, if the starting line in an autocross isn't level, will the accelerometer calibration be wrong?
Inertial Navigation with iPod Touch 4g and iPhone 4
Re: Inertial Navigation with iPod Touch 4g and iPhone 4
Hi,
That is a very good point. Both the gyro and the accelerator can be used to improve / interpolate and even extrapolate GPS fixes. The prototype I have running is not yet production ready (it works with acceleration) but is planned to be further developed. It certainly would allow a completely different approach for calibration and synchronization of sensor input in addition. The current calibration sensor depends on several conditions to be met. One is the point you raised, you need to stand on a flat surface during the first step of calibration (which does not need to be the standing start point, but any time after having started and mounted LT). The other is you need to drive straight during step #3 (visible in the calibration assistant). The later is not as important as the first one. In case lat / lin acceleration are not around zero while standing, it makes sense to reset calibration (usually in the Calibration View, but possible in the Video View - touch the circle - too). LT v16 will auto recalibrate if LT detects calibration has been wrong.
Finally, recalculate works with GPS only - assuming the corner radius + speed it calculates lateral acceleration. Lineal acceleration is calculated from the speed change. Both are not too accurate but far better than nothing
- Harry
That is a very good point. Both the gyro and the accelerator can be used to improve / interpolate and even extrapolate GPS fixes. The prototype I have running is not yet production ready (it works with acceleration) but is planned to be further developed. It certainly would allow a completely different approach for calibration and synchronization of sensor input in addition. The current calibration sensor depends on several conditions to be met. One is the point you raised, you need to stand on a flat surface during the first step of calibration (which does not need to be the standing start point, but any time after having started and mounted LT). The other is you need to drive straight during step #3 (visible in the calibration assistant). The later is not as important as the first one. In case lat / lin acceleration are not around zero while standing, it makes sense to reset calibration (usually in the Calibration View, but possible in the Video View - touch the circle - too). LT v16 will auto recalibrate if LT detects calibration has been wrong.
Finally, recalculate works with GPS only - assuming the corner radius + speed it calculates lateral acceleration. Lineal acceleration is calculated from the speed change. Both are not too accurate but far better than nothing
- Harry
-
gplracerx
- 20 or more Posts ★★★
- Posts: 836
- Joined: Thu May 03, 2012 5:26 am
- Location: Kingsport, TN USA
Re: Inertial Navigation with iPod Touch 4g and iPhone 4
I think I would do a moving fourth order polynomial fit for latitude vs time and longitude vs. time. I believe there's an exact solution for five points. The first derivative at the center point would be the velocity and the second the acceleration as vectors NS and EW. Then you could calculate the lateral and lineal accelerations parallel and perpendicular to the velocity vector. That may not be much more complicated than fitting a circle. I may try that and see if there's a linear correlation to the accelerometer data.