With the ongoing development of Indoor Location-Based Services, accurate location information of users in indoor environments has been a challenging issue in recent years. Due to the widespread use of WiFi networks, WiFi fingerprinting has become one of the most practical methods of locating mobile users. In addition to localization accuracy, some other critical factors such as cost, latency, and users' privacy should be considered in indoor localization systems. In this study, we propose a lightweight Convolutional Neural Network (CNN)-based method for edge devices (such as smartphones) to overcome the above issues by eliminating the need for a cloud/server in the localization system. To enable the use of the proposed model on resource-constraint edge devices, post-training optimization techniques including quantization, pruning and clustering are used to compress the network model. The proposed method is evaluated for three different open datasets, i.e., UJIIndoorLoc, Tampere and UTSIndoorLoc, as well as for our collected dataset named SBUK-D to verify its scalability. The results demonstrate the superiority of the proposed method compared to state-of-the-art studies. We also evaluate performance efficiency of our localization method on an android smartphone to demonstrate its applicability to edge devices. For UJIIndoorLoc dataset, our model with post-training optimizations obtains approximately 99% building accuracy, over 98% floor accuracy, and 4 m positioning mean error with the model size and inference time of 60 KB and 270 us, respectively, which demonstrate high accuracy as well as amenability to the resource-constrained edge devices.

翻译：随着室内位置服务的持续发展，在室内环境中获取用户的精确位置信息近年来一直是一个具有挑战性的问题。由于WiFi网络的广泛应用，WiFi指纹已成为定位移动用户最实用的方法之一。除了定位精度外，成本、延迟和用户隐私等其他关键因素也应纳入室内定位系统的考量范围。在本研究中，我们提出了一种轻量级卷积神经网络方法，用于边缘设备（如智能手机），通过消除定位系统中对云端/服务器的依赖来克服上述问题。为了在资源受限的边缘设备上使用所提出的模型，我们采用量化、剪枝和聚类等训练后优化技术对网络模型进行压缩。该方法在三个不同的公开数据集（即UJIIndoorLoc、Tampere和UTSIndoorLoc）以及我们自行收集的SBUK-D数据集上进行了评估，以验证其可扩展性。结果表明，所提方法相较于现有最先进研究具有优越性。我们还在一部安卓智能手机上评估了定位方法的性能效率，以证明其在边缘设备上的适用性。对于UJIIndoorLoc数据集，经过训练后优化后的模型实现了约99%的建筑识别准确率、超过98%的楼层识别准确率以及4米的位置平均误差，模型大小与推理时间分别为60 KB和270微秒，这证明了该方法的高精度及其对资源受限边缘设备的适用性。