Applying Deep Neural Network (DNN) for large-scale indoor localization using feed-forward neural network (FFNN) algorithm

Belay, Abebe; Yen, Lei; Lin, Hsin‐Piao; Yayeh, Yirga; Li, Yun Ruei; Jeng, Shiann-Shiun; Berie, Getaneh

doi:10.1109/icasi.2018.8394387

Cited by 28 publications

(10 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wi-Fi fingerprinting relies on a set of fingerprints taken at well-known positions for the position estimation; i.e., Wi-Fi FP requires a reference dataset (or radio map) to operate. Different well-known methods tackle this problem, including the Nearest Neighbour (NN) algorithm k-NN [7], Gaussian kernels [8], Bayesian models [9], Neural Networks [10] and, even, Deep Learning [11,12].…”

Section: Introductionmentioning

confidence: 99%

A Comprehensive and Reproducible Comparison of Clustering and Optimization Rules in Wi-Fi Fingerprinting

Torres-Sospedra

Richter

Moreira

et al. 2022

IEEE Trans. on Mobile Comput.

View full text Add to dashboard Cite

Wi-Fi fingerprinting is a well-known technique used for indoor positioning. It relies on a pattern recognition method that compares the captured operational fingerprint with a set of previously collected reference samples (radio map) using a similarity function. The matching algorithms suffer from a scalability problem in large deployments with a huge density of fingerprints, where the number of reference samples in the radio map is prohibitively large. This paper presents a comprehensive comparative study of existing methods to reduce the complexity and size of the radio map used at the operational stage. Our empirical results show that most of the methods reduce the computational burden at the expense of a degraded accuracy. Among the studied methods, only k-means, affinity propagation, and the rules based on the strongest access point properly balance the positioning accuracy and computational time. In addition to the comparative results, this paper also introduces a new evaluation framework with multiple datasets, aiming at getting more general results and contributing to a better reproducibility of new proposed solutions in the future.

show abstract

Section: Introductionmentioning

confidence: 99%

A Comprehensive and Reproducible Comparison of Clustering and Optimization Rules in Wi-Fi Fingerprinting

Torres-Sospedra

Richter

Moreira

et al. 2022

IEEE Trans. on Mobile Comput.

View full text Add to dashboard Cite

show abstract

“…The colors from blue and red represent weak and strong RSS values, respectively. The average coverage area was over 1000 m2 per AP, which was larger than the majority of the existing indoor localization works (e.g., approximately 60 m2 per AP in [35] and 40 m2 per AP in [38]. For practical engineering practice, an AP density of 100 to 400 m2 per AP is commonly used for meter-level WiFi or BLE based localization).…”

Section: Tests and Resultsmentioning

confidence: 99%

Wireless Fingerprinting Uncertainty Prediction Based on Machine Learning

Гао

et al. 2019

Sensors

View full text Add to dashboard Cite

Although wireless fingerprinting has been well researched and widely used for indoor localization, its performance is difficult to quantify. Therefore, when wireless fingerprinting solutions are used as location updates in multi-sensor integration, it is challenging to set their weight accurately. To alleviate this issue, this paper focuses on predicting wireless fingerprinting location uncertainty by given received signal strength (RSS) measurements through the use of machine learning (ML). Two ML methods are used, including an artificial neural network (ANN)-based approach and a Gaussian distribution (GD)-based method. The predicted location uncertainty is evaluated and further used to set the measurement noises in the dead-reckoning/wireless fingerprinting integrated localization extended Kalman filter (EKF). Indoor walking test results indicated the possibility of predicting the wireless fingerprinting uncertainty through ANN the effectiveness of setting measurement noises adaptively in the integrated localization EKF.

show abstract

“…WOA algorithm is combined with BP neural network, and WOA algorithm is used to adjust the parameters of BP neural network. Then the RSSI value data and corresponding parameters collected at different distances are taken as the input value of BP neural network, and the coordinates of unknown nodes are taken as the output value of BP neural network, so as to establish WOA-BP neural network model and complete the node positioning [21][22][23][24][25][26].…”

Section: Using Woa-bp To Optimize Indoor Environment Attenuation Modelmentioning

confidence: 99%

A Node Localization Algorithm based on WoaBp Optimization

Lang¹,

Sun²,

Jun³

et al. 2021

IJWMT

View full text Add to dashboard Cite

With the rapid development of 5G technology, the era of interconnection of all things has arrived. At the same time, a variety of hardware and software are getting more and more location information through sensors, and the accuracy of location information is increasingly important. Because traditional positioning relies on satellite signals, it achieves good results outdoors without obstruction, but indoors, due to the obstruction of various walls, such as Beidou satellite navigation system and U.S. Global Positioning System, it is difficult to meet the accuracy requirements for indoor positioning. Therefore, how to improve the positioning accuracy of indoor nodes has become a research hotspot in the field of wireless sensor. In order to improve the indoor positioning accuracy, this paper combines artificial neural network, intelligent optimization algorithm and node positioning to improve the accuracy of indoor positioning. One of the essences of the neural network is to solve the regression problem. Through the analysis of indoor node positioning, it can be concluded that the accuracy of distance-based positioning method lies in finding the relationship between signal strength and distance value. Therefore, the neural network can be used to regression analysis of signal strength and distance value and generate related models. In order to further improve the accuracy and stability of indoor node positioning, a method combining whale optimization algorithm with neural network is proposed. By using the whale optimization algorithm to find the optimal parameters of the neural network model, the training accuracy and speed of the neural network are improved. Then, using the excellent fitting ability of the neural network, the mapping relationship between RSSI value and distance value of indoor nodes is fitted, and the corresponding regression analysis model is generated, which can minimize the noise problem caused by abnormal signal attenuation and reduce the indoor positioning error. Finally, the data is processed by the neural network to get the parameters needed in the positioning algorithm. The experimental results show that the node positioning model based on the optimized neural network and the single optimization algorithm has significantly improved the positioning accuracy and stability.

show abstract

Applying Deep Neural Network (DNN) for large-scale indoor localization using feed-forward neural network (FFNN) algorithm

Cited by 28 publications

References 12 publications

A Comprehensive and Reproducible Comparison of Clustering and Optimization Rules in Wi-Fi Fingerprinting

A Comprehensive and Reproducible Comparison of Clustering and Optimization Rules in Wi-Fi Fingerprinting

Wireless Fingerprinting Uncertainty Prediction Based on Machine Learning

A Node Localization Algorithm based on WoaBp Optimization

Contact Info

Product

Resources

About