ALADIn: Autonomous Linear Antenna Delay Inference on Resource-Constrained Ultra-Wideband Devices

Paper i proceeding, 2023

Enabling precise indoor localization in a cheap and small package, Ultra-Wideband (UWB) transceivers bring decimetre-accurate ranging to resource-constrained IoT devices. Due to hardware-induced signal processing delays, device-specific antenna calibration enables the most accurate ranging results. This work introduces ALADIn for estimation and calibration of antenna delays in an autonomous manner, removing the need for manual labor and external hardware or computation. Based on known geometry, our approach allows already deployed devices to utilize their ranging and computational capabilities to optimize delays and reduce ranging errors autonomously. At its heart, ALADIn combines an efficient all-to-all ranging primitive with ordinary least squares inference. We conduct both extensive simulations and on-site evaluations. Our simulation results indicate that the proposed approach performs similarly to available calibration methods while being computationally less expensive. Deployed on three testbeds, we analyze the calibration performance on up to 14 DWM1001 devices. For one, the proposed calibration reduces mean absolute error alongside the standard deviation: from uncali-brated 10.7 (7.0 SD) cm to 6.7 (4.2 SD) cm, which is also lower than the 8.1 (5.7 SD) cm of error induced by factory-calibrated values. In addition, our results highlight the quality of measurements, i.e., pairwise variances, and further exhibit the potential of excluding multipath-affected links from the estimation process. Our implementation builds on Zephyr RTOS and is released as open-source.