LIGATE: LiDAR-IMU Guided GAT-Transformer for Bluetooth AoA Localization With AIoT-Driven Edge Deployment

Abstract

Accurate and scalable indoor localization is essential for real-time tracking of materials, tools, and mobile assets in industrial environments. Among existing technologies, Bluetooth Angle-of-Arrival (AoA) is a hardware-cost-effective solution for large-scale asset localization. However, existing AoA-based methods face key limitations, including high computational complexity, costly ground-truth acquisition, and inefficient deployment. This article presents LIGATE, a LiDAR-Inertial Measurement Unit (IMU)-guided framework that integrates a graph attention network (GAT) and a Transformer-based localization model for Bluetooth Low Energy (BLE) AoA, optimized for Artificial Intelligence of Things (AIoT) edge deployment. First, a LiDAR-IMU fusion system automatically generates high-precision ground truth without requiring labor-intensive manual calibration. Second, the GAT-Transformer directly maps raw AoA signals to spatial coordinates, bypassing intermediate angle-to-position conversion errors. Third, the system is specifically designed for real-time edge deployment, reducing latency and supporting scalable industrial applications. Evaluated in both a controlled pilot environment and a warehouse-like industrial environment, LIGATE achieved postfiltered mean localization errors of 0.169 and 0.335 m, respectively, with an end-to-end latency of 24.515 ms on an RK3576 edge device. These results demonstrate that LIGATE provides accurate, robust, and deployable BLE AoA localization for industrial edge intelligence.