Multifeature-Based Outdoor Fingerprint Localization With Accuracy Enhancement for Cellular Network.

Localization technology is a critical element in obtaining spatial data for the Internet of Things (IoT) and represents one of the most promising development areas for the next generation of IoT. In this regard, this article proposes a multifeature-based outdoor fingerprint localization technique with accuracy enhancement for the cellular network. The fingerprint collection scenarios are carefully designed to include diverse urban environments and seasonal characteristics. Based on these scenarios, a new set of cellular network parameters is introduced as a multifeature composition of fingerprint, resulting in marked improvements in localization accuracy. Furthermore, to alleviate the interfering information brought by multifeature, an adaptive bistage feature processing (BFP) method is proposed. At the stage of location matching, a hybrid model combines deep learning and $K$ -nearest neighbors (KNN) algorithms is implemented to enhance the localization accuracy. Finally, a unique error detection (UED) method is proposed to check the predicted real-time fingerprint position. The experimental results demonstrate that the proposed technique achieves a median localization error (MLE) of about 8 m and an average localization error (ALE) of 15.4 m in a complex urban outdoor environment, improving localization accuracy by 41.6% compared to other state-of-the-art fingerprint localization techniques. The proposed technique shows the potential to be an effective alternative to outdoor IoT nodes utilizing other localization sensors.