Outdoor Accuracy Evaluation of DecaWave’s DWM1002 PDoA Kit Measurements

Accurate localization in Wireless Sensor Networks (WSNs) is critical for applications such as emergency response, traffic monitoring, and industrial process control. Range-based methods generally outperform range-free ones, and Ultra Wide-band (UWB) technology is particularly promising due to its high bandwidth and robustness against multipath effects. This paper evaluates the DecaWave DWM1002 Phase Difference of Arrival (PDoA) kit, an experimental commercial UWB platform extending the earlier DWM1001 Time Difference of Arrival (TDoA) boards, through a real-world outdoor test-bed spanning multiple distances and angles. Our objective is not to propose a new localization algorithm, but to characterize and correct structured, repeatable PDoA bias that affects outdoor deployments. The raw measurements show systematic errors, with short ranges often underestimated and longer ranges increasingly overestimated, deviating significantly from the centimeter-level accuracy claimed by the manufacturer. The observed bias is highly consistent across repetitions, indicating good precision but poor accuracy and motivating lightweight calibration. To mitigate these biases, we explored regression-and Machine Learning (ML)-based calibration models. The best-performing ML approach reduced the average error to around 12 cm, achieving an improvement of roughly 70% over the uncalibrated data. These results indicate that while the kit is unsuitable for safety-critical localization in its raw form, lightweight calibration makes it viable for non-critical tasks such as tracking or crowd monitoring.