TTSL: An indoor localization method based on Temporal Convolutional Network using time-series RSSI

Bing Jia; Jingbin Liu; Tao Feng; Baoqi Huang; Thar Baker; Hissam Tawfik

doi:10.1016/j.comcom.2022.07.003

TTSL: An indoor localization method based on Temporal Convolutional Network using time-series RSSI

Bing Jia, Jingbin Liu, Tao Feng, Baoqi Huang, Thar Baker, Hissam Tawfik

School of Arch, Tech and Eng

Research output: Contribution to journal › Article › peer-review

Abstract

The existing indoor location methods are mainly oriented towards the study of single Received Signal Strength Indication ( RSSI ), which does not make full use of the time information attached to RSSI, so the location accuracy is limited. In this paper, considering the correlation of RSSI in time and space, Temporal Convolutional Network (TCN) based Time Series Localization (TTSL) method is proposed by using the temporal and spatial characteristics of signals in continuous locations. The neural network model is used to extract the time fluctuation characteristics of signals in continuous locations, and the nonlinear mapping relationship between signal characteristics and time and space to location coordinates is learned. The correlation of RSSI time and location information in the trajectory is realized, and the discrete location task is transformed into a continuous time series feature discovery task. A large number of experiments were carried out in the space of approximately 1000 square meters, and a comprehensive comparison was made with the existing methods. The average location error of TTSL was 3.73 m, and the performance is found to more stable than the existing methods. The TTSL method has a relatively small dependence on data volume, eliminates spatial ambiguity, and significantly reduces the influence of noise on location results.

Original language	English
Pages (from-to)	293-301
Number of pages	9
Journal	Computer Communications
Volume	193
DOIs	https://doi.org/10.1016/j.comcom.2022.07.003
Publication status	Published - 8 Jul 2022

Bibliographical note

Funding Information:
1. Organization name: National Natural Science Foundation of China Grant numbers: 42161070 Grant numbers: 41761086 Grant numbers: 41871363

Funding Information:
3. Organization name: Key Science-Technology Project of Inner Mongolia Autonomous Region Grant numbers: 2021GG0163

Funding Information:
2. Organization name: Natural Science Foundation of Inner Mongolia Autonomous Region of China Grant numbers: 2019MS06030 Grant numbers: 2021ZD13

Publisher Copyright:
© 2022 Elsevier B.V.

Keywords

Indoor location method
Temporal Convolutional Network
Time-series WiFi RSSI
WiFi indoor localization

Access to Document

10.1016/j.comcom.2022.07.003

Cite this

@article{fbade2f93a844dcc807723d199fdf13f,

title = "TTSL: An indoor localization method based on Temporal Convolutional Network using time-series RSSI",

abstract = "The existing indoor location methods are mainly oriented towards the study of single Received Signal Strength Indication ( RSSI ), which does not make full use of the time information attached to RSSI, so the location accuracy is limited. In this paper, considering the correlation of RSSI in time and space, Temporal Convolutional Network (TCN) based Time Series Localization (TTSL) method is proposed by using the temporal and spatial characteristics of signals in continuous locations. The neural network model is used to extract the time fluctuation characteristics of signals in continuous locations, and the nonlinear mapping relationship between signal characteristics and time and space to location coordinates is learned. The correlation of RSSI time and location information in the trajectory is realized, and the discrete location task is transformed into a continuous time series feature discovery task. A large number of experiments were carried out in the space of approximately 1000 square meters, and a comprehensive comparison was made with the existing methods. The average location error of TTSL was 3.73 m, and the performance is found to more stable than the existing methods. The TTSL method has a relatively small dependence on data volume, eliminates spatial ambiguity, and significantly reduces the influence of noise on location results.",

keywords = "Indoor location method, Temporal Convolutional Network, Time-series WiFi RSSI, WiFi indoor localization",

author = "Bing Jia and Jingbin Liu and Tao Feng and Baoqi Huang and Thar Baker and Hissam Tawfik",

note = "Funding Information: 1. Organization name: National Natural Science Foundation of China Grant numbers: 42161070 Grant numbers: 41761086 Grant numbers: 41871363 Funding Information: 3. Organization name: Key Science-Technology Project of Inner Mongolia Autonomous Region Grant numbers: 2021GG0163 Funding Information: 2. Organization name: Natural Science Foundation of Inner Mongolia Autonomous Region of China Grant numbers: 2019MS06030 Grant numbers: 2021ZD13 Publisher Copyright: {\textcopyright} 2022 Elsevier B.V.",

year = "2022",

month = jul,

day = "8",

doi = "10.1016/j.comcom.2022.07.003",

language = "English",

volume = "193",

pages = "293--301",

journal = "Computer Communications",

issn = "1873-703X",

publisher = "Elsevier",

}

TY - JOUR

T1 - TTSL

T2 - An indoor localization method based on Temporal Convolutional Network using time-series RSSI

AU - Jia, Bing

AU - Liu, Jingbin

AU - Feng, Tao

AU - Huang, Baoqi

AU - Baker, Thar

AU - Tawfik, Hissam

N1 - Funding Information: 1. Organization name: National Natural Science Foundation of China Grant numbers: 42161070 Grant numbers: 41761086 Grant numbers: 41871363 Funding Information: 3. Organization name: Key Science-Technology Project of Inner Mongolia Autonomous Region Grant numbers: 2021GG0163 Funding Information: 2. Organization name: Natural Science Foundation of Inner Mongolia Autonomous Region of China Grant numbers: 2019MS06030 Grant numbers: 2021ZD13 Publisher Copyright: © 2022 Elsevier B.V.

PY - 2022/7/8

Y1 - 2022/7/8

N2 - The existing indoor location methods are mainly oriented towards the study of single Received Signal Strength Indication ( RSSI ), which does not make full use of the time information attached to RSSI, so the location accuracy is limited. In this paper, considering the correlation of RSSI in time and space, Temporal Convolutional Network (TCN) based Time Series Localization (TTSL) method is proposed by using the temporal and spatial characteristics of signals in continuous locations. The neural network model is used to extract the time fluctuation characteristics of signals in continuous locations, and the nonlinear mapping relationship between signal characteristics and time and space to location coordinates is learned. The correlation of RSSI time and location information in the trajectory is realized, and the discrete location task is transformed into a continuous time series feature discovery task. A large number of experiments were carried out in the space of approximately 1000 square meters, and a comprehensive comparison was made with the existing methods. The average location error of TTSL was 3.73 m, and the performance is found to more stable than the existing methods. The TTSL method has a relatively small dependence on data volume, eliminates spatial ambiguity, and significantly reduces the influence of noise on location results.

AB - The existing indoor location methods are mainly oriented towards the study of single Received Signal Strength Indication ( RSSI ), which does not make full use of the time information attached to RSSI, so the location accuracy is limited. In this paper, considering the correlation of RSSI in time and space, Temporal Convolutional Network (TCN) based Time Series Localization (TTSL) method is proposed by using the temporal and spatial characteristics of signals in continuous locations. The neural network model is used to extract the time fluctuation characteristics of signals in continuous locations, and the nonlinear mapping relationship between signal characteristics and time and space to location coordinates is learned. The correlation of RSSI time and location information in the trajectory is realized, and the discrete location task is transformed into a continuous time series feature discovery task. A large number of experiments were carried out in the space of approximately 1000 square meters, and a comprehensive comparison was made with the existing methods. The average location error of TTSL was 3.73 m, and the performance is found to more stable than the existing methods. The TTSL method has a relatively small dependence on data volume, eliminates spatial ambiguity, and significantly reduces the influence of noise on location results.

KW - Indoor location method

KW - Temporal Convolutional Network

KW - Time-series WiFi RSSI

KW - WiFi indoor localization

UR - http://www.scopus.com/inward/record.url?scp=85134705147&partnerID=8YFLogxK

U2 - 10.1016/j.comcom.2022.07.003

DO - 10.1016/j.comcom.2022.07.003

M3 - Article

SN - 1873-703X

VL - 193

SP - 293

EP - 301

JO - Computer Communications

JF - Computer Communications

ER -

TTSL: An indoor localization method based on Temporal Convolutional Network using time-series RSSI

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Embargoed Document

Fingerprint

Cite this