Beacon against other 

adaptive algorithms

 Email

Quick Navigation

Intro

Sensor Geometry

SMF versus RLS

DOWNLINK

Beacon against RLS

The influence of gamma: Gamma is the threshold which indicates if an update of the parameter estimate is required or not. Gamma is directly related to the frequency of updates and a smaller gamma implies more frequent updates. In the time- invariant case, a small gamma usually yields a better final solution. It had been proven that the error is asymptotically upper bounded by the factor gamma.

     But the problem is different in the time variant case. In that case, the better error rate is not necessarily obtained with a small gamma: To explain that with simple words, 'It's sometimes better not to update than to update in a wrong direction'  

   The gamma is also related to the forgetting factor. The algorithm doesn't try to find out every step the best solution (And this is a part of the design specification and a feature enjoyed by SMF algorithms : a Trade-off between the quality of the result you want to reach and the frequency of updates, which is actually the price you are ready to pay in terms of computations). If the algorithm delivers a not so bad solution, then it will not even try to improve this one. (Because we're already in the tolerance margin).  

The Next movie shows the comparison Beacon / RLS by tracking a MS. The gamma is chosen to be equal to 0.07, which seems to be a good compromise between accuracy in Tracking and gain in terms of frequency of update.

MOVIES: You can either show these movies by clicking on the following links or download them as Zip files and then open them in your own computer. We strongly advise you to download them in Zip format if your Internet connection is slow because of the size of these documents:

movie (13,26 Mo)

movie.zip (693 Ko)

      < >