In recent years, Indoor Positioning Systems (IPS) have gained significance across various applications, including asset tracking, monitoring, interior navigation, and location-based services. The use of Wi-Fi-based technology is a popular choice for IPS due to its cost-effectiveness and widespread accessibility. The Wi-Fi signal’s Round Trip Time (RTT) measurement, using the Fine Time Measurement (FTM) protocol, offers fewer ranging errors in Line of Sight (LOS) conditions. However, Wi-Fi RTT ranging measurements encounter higher-ranging errors in Non-Line of Sight (NLOS), multipath, and interference scenarios. This study examines the error in ranging measurements for different scenarios such as LOS, Glass, Metal, and Wall blocking scenarios. To address these challenges, we propose a method that combines calibrated RTT range and processed Received Signal Strength (RSS) feature values in constructing a fingerprinting map to enhance positioning accuracy. The methodology includes the development of a range compensation model for RTT range calibration, utilizing a Gaussian filter for RSS measurement values processing, and creating a classifier model to distinguish between LOS and Wall scenarios. This integrated approach reduces noise in measurement values, and the Gaussian Process Regression (GPR) algorithm is utilized to predict the final location of the user. Our proposed method achieved a positioning error of 0.79 m, surpassing the performance of RTT fingerprinting by 17.71 %, RSS fingerprinting by 49.68 % and trilateration methods by 29.46 %.

Indoor Localization Systems (ILS) based on Wi-Fi which utilize Wi-Fi routers installed within indoor spaces, are particularly popular because of their low cost and availability. However, noise signals and multi-path problems can affect traditional Wi-Fi-based approaches, leading to high localization errors. To address these challenges of this approach, the 802.11mc protocol introduced a Fine Timing Measurement (FTM) frame, which adopts the two-way ranging technique. Using calibrated Round Trip Time (RTT) to remove the range offset in RTT at the initiator end and deep learning for indoor location recognition has become increasingly important in this context. In our work, we propose an indoor location method that combines calibrated Wi-Fi RTT fingerprinting with a range-based technique that utilizes a Deep Neural Network (DNN) regression and the Random Forest (RF) regression algorithm model. This approach leverages calibrated RTT to provide accurate distance measurements in indoor environments, resulting in more accurate distance measurements than the existing methods, such as Received Signal Strength (RSS) approach and Wi-Fi FTM based on DNN. The experiment area has 3 Access Points (APs) in a $9.55 \mathrm{m}\times 7.27\mathrm{m}$ room with a $1 \times 1\mathrm{m}$ grid covering 36 Reference Points (RPs). We trained the proposed model and predicted the location of a new fingerprint in the test dataset, achieving a Mean Squared Error (MSE) of 0.12 m and 0.11 m in reference and non-reference points, respectively. This shows how effectively the proposed model works for obtaining less localization error indoor positioning using calibrated Wi-Fi RTT fingerprinting.

With the emergence and development of various indoor positioning technologies, Wi-Fi fingerprint-based positioning has become one of the most widely used indoor positioning technologies due to its high performance in indoor positioning systems. This paper introduces a positioning approach for Wi-Fi fingerprints based on a stacked ensemble model. In our method, support vector regression (SVR) and XGBoost algorithms are employed to construct stacked ensemble model. Our proposed method can enhance the accuracy and robustness of indoor positioning in comparison with the existing Wi-Fi fingerprint-based positioning method. Firstly, the outlier of the raw RTT range is removed, and it is calibrated using a quadratic polynomial. Then, the Wi-Fi RTT range and its standard deviation values are used as environmental features and coordinates as output labels. The training dataset is imported into the base learner SVR and XGBoost, respectively, and its predicted output is used to create a new dataset. Finally, the new dataset is used in the linear regression model of the me-ta learner to predict the final position obtained. We simulated the complex environment of indoor positioning in our experiments, compared with existing machine learning methods and existing stacked ensemble models. The experimental results show that the proposed method effectively improves positioning accuracy.

With the development of wireless sensor network (WSN) technology, WSN based indoor localization technology has received more and more attention. Wi-Fi based ranging methods has become a current research hotspot in the field of localization due to their low cost and low complexity. However, the use of Wi-Fi based ranging methods is subject to large errors due to noise and multipath effects. To address this challenge, in this paper, the indoor positioning of RTT-RSS fusion is investigated by the improved ranging method. Firstly, the logarithmic attenuation model based on RSSI is improved by quadratic polynomial and the RTT is calibrated by linear model. And then the adaptive weighting method is adopted to fuse the RTT and RSS. Finally, the positions of the target nodes are obtained by the trilateral localization method. In the experimental results, the complex environment of indoor localization is simulated in the corridor and lounge respectively. And the results show that the method proposed in this paper can effectively improve the localization accuracy compared with the traditional method.

Indoor positioning using commodity WiFi has gained significant attention; however, achieving sub-meter accuracy across diverse layouts remains challenging due to multipath fading and Non-Line-Of-Sight (NLOS) effects. In this work, we propose a hybrid Graph–Temporal Convolutional Network (GTCN) model that incorporates Access Point (AP) geometry through graph convolutions while capturing temporal signal dynamics via dilated temporal convolutional networks. The proposed model adaptively learns per-AP importance using a lightweight gating mechanism and jointly exploits WiFi Received Signal Strength (RSS) and Round-Trip Time (RTT) features for enhanced robustness. The model is evaluated across four experimental areas such as lecture theatre, office, corridor, and building floor covering areas from 15 m × 14.5 m to 92 m × 15 m. We further analyze the sensitivity of the model to AP density under both LOS and NLOS conditions, demonstrating that positioning accuracy systematically improves with denser AP deployment, especially in large-scale mixed environments. Despite its high accuracy, the proposed GTCN remains computationally lightweight, requiring fewer than ${10}^5$ trainable parameters and only tens of MFLOPs per inference, enabling real-time operation on embedded and edge devices.

With the development of wireless communication technology research, location-based services (LBS) have emerged due to provide useful information to users based on the current user locations or a given location-indoor positioning technology has become a hot spot for research. Indoor positioning technology based on WIFI signals attentions owing to WIFI networks cover various indoor environments. Received signal strength (RSS) based indoor positioning technology has become an effective indoor positioning technology as a method that utilizes the wireless signal propagation model. However, due to the complexity and variability of the indoor environment, RSS values are highly vulnerable to noise and multipath interference, RSS-based WIFI fingerprinting can effectively avoid this problem. During the construction of RSS fingerprint maps, fingerprint data collection is time-consuming and laborious, how to use a small amount of fingerprint data to achieve high-precision positioning becomes a difficult point for fingerprint positioning technology. To solve this problem, a high-precision indoor localization method based on spectral normalization for generative adversarial networks (SNGAN) is proposed. The generative adversarial network with spectral normalization not only retains the original information to the greatest extent during the training process, but also makes the generated data closer to the distribution of real data and speeds up the training. Collecting a small amount of fingerprint data achieves higher localization accuracy.