Learning-Based WiFi Fingerprint Inpainting via Generative Adversarial Networks

TL;DR

This work addresses the challenge of completing WiFi fingerprint maps with incomplete survey data by predicting the fingerprint vector f_u ∈ \mathbb{R}^n at unseen locations. It introduces two learning-based inpainting models: IAP, which combines GPR initialization with a Variational AutoEncoder, and I2AP, an end-to-end, multi-channel GAN that leverages k-nearest-neighbor context and a specialized discriminator to preserve inter- and intra-AP correlations. Across three datasets, I2AP consistently achieves the lowest inpainting error and yields improved downstream positioning, demonstrating the importance of modeling spatial relationships and multi-AP channels in WiFi-based localization. The approach offers a practical pathway to reduce survey costs while maintaining or enhancing localization performance in complex indoor environments.

Abstract

WiFi-based indoor positioning has been extensively studied. A fundamental issue in such solutions is the collection of WiFi fingerprints. However, due to real-world constraints, collecting complete fingerprints at all intended locations is sometimes prohibited. This work considers the WiFi fingerprint inpainting problem. This problem differs from typical image/video inpainting problems in several aspects. Unlike RGB images, WiFi field maps come in any shape, and signal data may follow certain distributions. Therefore, it is difficult to forcefully fit them into a fixed-dimensional matrix, as done with processing images in RGB format. As soon as a map is changed, it also becomes difficult to adapt it to the same model due to scale issues. Furthermore, such models are significantly constrained in situations requiring outward inpainting. Fortunately, the spatial relationships of WiFi signals and the rich information provided among channels offer ample opportunities for this generative model to accomplish inpainting. Therefore, we designed this model to not only retain the characteristic of regression models in generating fingerprints of arbitrary shapes but also to accommodate the observational outcomes from densely deployed APs. This work makes two major contributions. Firstly, we delineate the distinctions between this problem and image inpainting, highlighting potential avenues for research. Secondly, we introduce novel generative inpainting models aimed at capturing both inter-AP and intra-AP correlations while preserving latent information. Additionally, we incorporate a specially designed adversarial discriminator to enhance the quality of inpainting outcomes.

Figures (4)

• Figure 1: IAP model architecture.
• Figure 2: I2AP model architecture.
• Figure 3: WiFi fingerprint datasets in our validation: (a) The campus-multi-fp dataset has six missing scenarios, A, B, C, D, E and F, each containing one or two unsurvey regions represented by squares. (Red point = grid point; x = non-grid point; solid triangle = 2.4 GHz AP; hollow triangle = 2.4+5 GHz AP) (b) The UJIIndoorLoc dataset has 5 floors, where data points with the same color are on the same floor. (c) The MallParking-fp dataset has two missing scenarios, marked as A and B.
• Figure 4: Comparisons of inpainting errors on Campus-multi-fp (L1 loss, $k=4$, and $dim(v_i) = 30$).