EKF IMU GPS Fusion Simulation

I need an Extended Kalman Filter that fuses raw IMU and GPS data to deliver reliable navigation states inside an autonomous-vehicle 3D simulation. I already have an EKF designed in MATLAB, I however require an improvement on it (I've all the files required). My workflow is centred on IPG Carmaker, so the filter must compile, run and stream its results in real time inside that environment; the same code should also remain portable enough to drop into CARLA for future tests.

The filter has to provide:
• position and velocity estimates in the global frame
• full orientation (roll, pitch, yaw)
• live statistics on sensor noise and estimated error covariance

A simple on-road scenario in Carmaker will serve as the proving ground: the vehicle must hold a trajectory with typical urban dynamics while the EKF maintains stable state estimates under artificial sensor noise and occasional GPS drop-outs. I will supply the IMU/GPS message format, vehicle model, and a baseline scenario file.

Acceptance is complete when:
1. Source code (C++ or MATLAB/Simulink S-Functions) builds without errors, plugs into Carmaker, and runs faster than real time.
2. Use High fidelity Physics based simulators such as CARLA, IPG Carmaker for 3D simulation including (x, y, heading)
3. Include Full EKF state-space equation and matrix definitions for reproducibility
4. Logged outputs match ground-truth within tolerances I will share (≤0.3 m RMS position, ≤0.5 deg RMS attitude).
5. Plots and a short report document filter tuning, assumptions, and observed sensor error statistics including RMSE.
6. Detailed report on work done.

If you have prior experience coupling EKFs with vehicle simulators such as Carmaker, please tell me which interfaces you used (e.g., FMI, ROS Bridge, TCP sockets). That background will help us start quickly.