I'm working on an iOS project which needs me to do simple dead reckoning when GPS is not available. I've been given a bunch of measurement data from various phones which consists of the following:
- GPS data (x, y, z, course, speed) when available
- Accelerometer (useracceleration x, y, z)
- Gyroscope (rotationrate x, y, z)
I want to compute missing GPS positions in the time series. I understand how to project locations based on last GPS position and course values, but this gives me large errors. My question is if it is possible to obtain better results if I combine the accelerometer and gyroscope data. I understand how to integrate accelerometer values to obtain speed information, but I'm not sure how to use the gyroscope data. Any help would be greatly appreciated.
It is not a small topic. On your question: "Yes, you can obtain much better results by combining inputs from all the sensors".
This is what Kalman filter (KF) is about - the process is called sensor fusion and there are a lot topics on the net about it. The version of KF which is usually used for this purpose is Extended Kalman filter. I recommend reading this paper - it implements and explains shortly, but clearly exactly the task your want to perform:
http://www.tkt.cs.tut.fi/research/nappo_files/1_C2.pdf