AI for Improved Indoor Positioning via Multi-Band Channel Charting

Sammanfattning

Channel charting is an unsupervised positioning method that learns a low-dimensionalspatial representation from channel state information (CSI), which describes howwireless signals propagate between a user and base stations. Unlike fingerprinting,it does not require ground-truth position labels during training. This thesis investigateswhether using CSI from two frequency bands improves channel charting comparedwith conventional single-band CSI. Simulated CSI was generated in a streetcanyonenvironment at 3.5 GHz and 12 GHz using Sionna RT. Three dual-bandchannel charting methods were evaluated: averaging dissimilarities between CSIsamples, multiplying similarity scores from both bands, and aligning two separatelytrained networks. The methods were compared with single-band channel chartingand with supervised fingerprinting baselines. The results show that dual-band fusionimproves channel charting performance across both chart-quality metrics andpositioning accuracy. The best channel charting result was obtained with similaritymultiplication, reducing the mean absolute error from 6.67 m for the best single-bandreference to 5.86 m. Dual-band fingerprinting also improved performance, reducingthe mean absolute error from 1.09 m to 1.02 m, although the relative improvementwas smaller than for channel charting. The gains were strongest in non-line-of-sightconditions, where the best channel charting error decreased from 7.64 m for the bestsingle-band reference to 6.54 m with dual-band similarity multiplication. These resultsindicate that multi-band CSI provides complementary spatial information andis especially useful for unsupervised channel charting in challenging propagationenvironments.