Objective
Our final objective is to estimate the position of the mobile. But, to achieve this objective, som previous estimations are needed:
- Channel Estimation.
- Time of Arrival Estimation. Two techniques are developed: Maximum Detection and Normalized Minimum Variance.
- Angle of Arrival Estimation. Two techniques are developed: Besson-Stoica and MUSIC.
- Kalman Filter: To track the speed and position of the mobile
Channel Estimation
The channel estimation is performed by the correlation of the received signal and the midamble sequence.
On the next figure you can see an example of a channel estimation in a LOS case.
One of the most important topics is the array calibration. This task is performed at the eigth TMS320C40 processors.
TOA Estimation
The TOA is computed using the channel estimation from the four antennas. We have developed two algorithms:
- Maximum detection: All the channel are combined by the addition of their squares.
The TOA arrival estimated, is the maximum value of this addition.
- NMV: This is a variation of the Capon Estimatior.
In the next image is the square sum of the previous example:
It can be seen how this algorithm works well in LOS situations with a low multipath.
The next figure shows a comparisson between the two estiamtors in a high multipath situation.
The blue line is the square addition, and the black line is the NMV estiamtor.
It can be seen that the NMV estimator can detect the different arrivals.
The next two links are two videos showing an evolution when using the NMV estimator.
AOA Estimation
An espatial correlation matrix is computed from the channel estimation filtered at the TOA estimated. Both algorithms
use this correlation matrix to compute the AOA:
- Besson-Stoica: With a very low computational cost
- MUSIC: More robust, but at a very high computational cost
The next graph shows a comparisson between the two estimators:
Position Tracking
The final position is tracked using a kalman filter. There are some different versions of this filter. The next videos
are examples of the result:
- Example 1: Blue line, true position; Black Line, A Kalman filter; Red Line, Final Version of the Kalman filter.
- Example 2: Red line, true position; Blue line, Estimated position
| 
