Journal Article

Neural Network-Based LoRa Received Signal Strength Indicator Fingerprint Identification for Indoor Localization of Mobile Robots

Barai, Chandan, Mazumder, Subhabrata, Pandey, Hemendra Kumar, Chandra, Abhijit, Sarkar, Meem, Sarkar, Ushnish, Samanta, Tapas

2026

Overview

This paper presents an indoor self-localization framework for mobile robots, an essential component for automation in Industry 4.0 and smart environments. We evaluate a Received Signal Strength Indicator (RSSI) fingerprinting technique utilizing Long-Range (LoRa) technology to overcome the challenges of congested indoor settings. To optimize communication parameters, the Structural Similarity Index Measure (SSIM) was employed to select the most effective spreading factor, while the entropy of the RSSI database was calculated to verify fingerprint stability. For positional prediction, a Multi-layer Perceptron (MLP) neural network was developed to classify the location of the target within a grid-based experimental setup, featuring cells spaced 60 cm apart. The MLP achieved a validation accuracy of 91.8 percent during training and demonstrated high precision in classifying grid regions within a signal-dense environment. For scenarios where slow-moving robots (5 cm/s) are required, like radiation mapping, this method provide highly accurate high-level localization data.These results suggest that the proposed LoRa-MLP integration provides a robust, low-power solution for high-accuracy indoor positioning systems (IPSs) in modern industrial infrastructure.

• Publisher: MDPI AG,Multidisciplinary Digital Publishing Institute (MDPI)
• Subject: Accuracy / Classification / Communication / Deep learning / entropy of fingerprint database / Global positioning systems / GPS / Laboratories / Localization / Location based services / LoRa / Machine learning / MLP / Neural networks / Robots / RSSI fingerprint / SSIM / time series RSSI preprocessing / Wireless access points