Tracking time-variant cluster parameters in MIMO channel measurements: algorithm and results
Author
Summary, in English
This paper presents a joint clustering-and-tracking framework to identify time-variant cluster parameters
for geometry-based stochastic MIMO channel models.
The method uses a Kalman filter for tracking and predicting cluster positions, a novel consistent
initial guess procedure that accounts for predicted cluster centroids, and the well-known KPower-
Means algorithm for cluster identification.
We tested the framework by applying it to three entirely different sets of MIMO channel measurement
data obtained by different channel sounders: indoor measurements conducted at 2.55
GHz, outdoor rural measurements at 300 MHz, and outdoor sub-urban measurements at 2.0 GHz.
The time-variant cluster parameters of interest are: (i) cluster movement, (ii) change of cluster
spreads, (iii) cluster lifetimes, and birth and death rates of cluster.
We find that clusters show significant movement in parameter space depending on the environment.
The spreads of individual clusters change rather randomly over their lifetime, with a
standard deviation up to 150% of their mean spread. The cluster lifetime is approximately exponentially
distributed, however additionally one has to account for long-living clusters coming from
the line-of-sight path or from major reflectors.
for geometry-based stochastic MIMO channel models.
The method uses a Kalman filter for tracking and predicting cluster positions, a novel consistent
initial guess procedure that accounts for predicted cluster centroids, and the well-known KPower-
Means algorithm for cluster identification.
We tested the framework by applying it to three entirely different sets of MIMO channel measurement
data obtained by different channel sounders: indoor measurements conducted at 2.55
GHz, outdoor rural measurements at 300 MHz, and outdoor sub-urban measurements at 2.0 GHz.
The time-variant cluster parameters of interest are: (i) cluster movement, (ii) change of cluster
spreads, (iii) cluster lifetimes, and birth and death rates of cluster.
We find that clusters show significant movement in parameter space depending on the environment.
The spreads of individual clusters change rather randomly over their lifetime, with a
standard deviation up to 150% of their mean spread. The cluster lifetime is approximately exponentially
distributed, however additionally one has to account for long-living clusters coming from
the line-of-sight path or from major reflectors.
Publishing year
2007
Language
English
Document type
Conference paper
Topic
- Electrical Engineering, Electronic Engineering, Information Engineering
Conference name
3rd COST2100 Management Committee Meeting, 2007
Conference date
2007-09-10 - 2007-09-12
Conference place
Duisburg, Germany
Status
Published