Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging the Doppler Effect for Channel Charting

Florian Euchner, Phillip Stephan, Stephan ten Brink

TL;DR

This work presents a Doppler effect–based loss for Channel Charting that requires only frequency synchronization across a small number of distributed BS antennas, enabling global-coordinate channel charts in indoor environments. By formulating a pairwise log-likelihood that leverages inter-antenna phase differences to cancel CFO, and training a Siamese neural network on CSI-derived features, the method learns a forward charting function that maps CSI to a two-dimensional global map without relying on angular or time synchronization. The approach demonstrates robust localization in a 14 m by 14 m indoor area using just four antennas and outperforms dissimilarity-based baselines, with promising implications for practical deployments where phase/time sync is limited. The results suggest avenues for extending the framework to frequency-shift–based dissimilarities and delay-Doppler features, potentially broadening the applicability of Channel Charting in real-world networks.

Abstract

Channel Charting is a dimensionality reduction technique that reconstructs a map of the radio environment from similarity relationships found in channel state information. Distances in the channel chart are often computed based on some dissimilarity metric, which can be derived from angular-domain information, channel impulse responses, measured phase differences or simply timestamps. Using such information implicitly makes strong assumptions about the level of phase and time synchronization between base station antennas or assumes approximately constant transmitter velocity. Many practical systems, however, may not provide phase and time synchronization and single-antenna base stations may not even have angular-domain information. We propose a Doppler effect-based loss function for Channel Charting that only requires frequency synchronization between spatially distributed base station antennas, which is a much weaker assumption. We use a dataset measured in an indoor environment to demonstrate that the proposed method is practically feasible with just four base station antennas, that it produces a channel chart that is suitable for localization in the global coordinate frame and that it outperforms other state-of-the-art methods under the given limitations.

Leveraging the Doppler Effect for Channel Charting

TL;DR

This work presents a Doppler effect–based loss for Channel Charting that requires only frequency synchronization across a small number of distributed BS antennas, enabling global-coordinate channel charts in indoor environments. By formulating a pairwise log-likelihood that leverages inter-antenna phase differences to cancel CFO, and training a Siamese neural network on CSI-derived features, the method learns a forward charting function that maps CSI to a two-dimensional global map without relying on angular or time synchronization. The approach demonstrates robust localization in a 14 m by 14 m indoor area using just four antennas and outperforms dissimilarity-based baselines, with promising implications for practical deployments where phase/time sync is limited. The results suggest avenues for extending the framework to frequency-shift–based dissimilarities and delay-Doppler features, potentially broadening the applicability of Channel Charting in real-world networks.

Abstract

Channel Charting is a dimensionality reduction technique that reconstructs a map of the radio environment from similarity relationships found in channel state information. Distances in the channel chart are often computed based on some dissimilarity metric, which can be derived from angular-domain information, channel impulse responses, measured phase differences or simply timestamps. Using such information implicitly makes strong assumptions about the level of phase and time synchronization between base station antennas or assumes approximately constant transmitter velocity. Many practical systems, however, may not provide phase and time synchronization and single-antenna base stations may not even have angular-domain information. We propose a Doppler effect-based loss function for Channel Charting that only requires frequency synchronization between spatially distributed base station antennas, which is a much weaker assumption. We use a dataset measured in an indoor environment to demonstrate that the proposed method is practically feasible with just four base station antennas, that it produces a channel chart that is suitable for localization in the global coordinate frame and that it outperforms other state-of-the-art methods under the given limitations.
Paper Structure (12 sections, 10 equations, 5 figures, 1 table)

This paper contains 12 sections, 10 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Dataset and System Model
  3. Doppler Effect-Based Loss Function
  4. Introduction: Doppler Effect and Cumulative Phases
  5. In Discrete Time, Comparing Phases
  6. Log-Likelihood Function
  7. Uncertainty
  8. Foward Charting Function Training
  9. Evaluation
  10. Baselines and Optimal Affine Transforms
  11. Results
  12. Outlook

Figures (5)

  • Figure 1: Information about the environment the dataset was measured in: The figure shows (a) a photograph of the environment, (b) a top view map and (c) a scatter plot of colorized "ground truth" positions (only used for evaluation, not training) of datapoints in $\mathcal{S}_\mathrm{train}$.
  • Figure 2: Single BS antenna, unsynchronized UE moving at velocity $v$ away from BS: Received uplink phase is affected by both CFO and Doppler shift.
  • Figure 3: Two BS antennas, UE moving at velocity $v$ between BS antennas. The change in differential uplink phases, caused by the Doppler effect, is now unaffected by CFO. $\lambda$ denotes the wavelength. The figure assumes $t_1 = 0$.
  • Figure 4: Learned channel charts, FCF applied to $S_\mathrm{train}$. Colors of points are preserved from color gradient applied in Fig. \ref{['fig:groundtruth-map']}.
  • Figure 5: Empirical CDFs of absolute localization errors (evaluated on $\mathcal{S}_\mathrm{train}$)