# Predictive object tracking with 2 IMUs and a stereo camera

I am quite new to this so please forgive me for any mistakes. I am working on a simulation project and have very little clue as to what to do and I hope someone can point me to the right direction.

Here is the summary of my problem. (Please keep in mind that this is a simulation, no real equipment or IMU is involved)

I have an object O and a stereo camera C. Both have an 9dof IMU attached and are moving freely in space. O and C are 3m in range. O's IMU broadcast acceleratometer info periodically. C is equipped with software to extract the 3D position of O. Find the best way to fuse the data and predict O's position in C's reference in the very near future.

My thought on solving the problem is as follows: (please point out any mistakes I have because I am honestly super new at this...)

1. Initialize the distance between O and C with an error distribution
2. Every time I receive O's imu data, I need to calculate where it is approximately with an error distribution
3. then use a nonlinear Kalman filter to fuse the 2 distributions together and find where O might be.

My question is this

1. How do I calculate where O is given approximate initial pos, and 9dof IMU data?
2. How do I combine C's IMU and O's IMU so that the measurement is done in C's reference frame?

If anyone can point me towards the right direction I will be very very grateful. Thank you so much!!