{"title":"A Location-Blind Spatial Regression Framework for IoT Monitoring Systems Based on Location Distribution and Spatial Correlation","authors":"Koki Kanzaki;Koya Sato","doi":"10.1109/LSENS.2024.3441104","DOIUrl":null,"url":null,"abstract":"This letter presents a spatial regression framework that does not rely on absolute positions, such as those obtained from the Global Navigation Satellite System. Typical spatial analysis methods, which depend on precise sensor location data, can result in increased sensor costs and reduced accuracy in challenging environments, such as indoor or underwater settings. Our framework circumvents the need for positioning functions at the sensors by estimating the locational relationships among sensors on relative coordinates based on the probability distribution of the sensor locations and spatial correlations of the sensed data. The server then performs spatial analysis on the relative coordinate system using a regression method, such as Gaussian process regression (GPR) or inverse distance weighting. We validate our approach with two open datasets: a meteorological dataset from the Japanese Meteorological Agency and Intel Lab Data. In both datasets, our results demonstrate that the proposed method can realize spatial regression analysis with less than 10% accuracy degradation in terms of median root mean squared error compared to GPR on absolute coordinates.","PeriodicalId":13014,"journal":{"name":"IEEE Sensors Letters","volume":null,"pages":null},"PeriodicalIF":2.2000,"publicationDate":"2024-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10632575","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Sensors Letters","FirstCategoryId":"1085","ListUrlMain":"https://ieeexplore.ieee.org/document/10632575/","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
This letter presents a spatial regression framework that does not rely on absolute positions, such as those obtained from the Global Navigation Satellite System. Typical spatial analysis methods, which depend on precise sensor location data, can result in increased sensor costs and reduced accuracy in challenging environments, such as indoor or underwater settings. Our framework circumvents the need for positioning functions at the sensors by estimating the locational relationships among sensors on relative coordinates based on the probability distribution of the sensor locations and spatial correlations of the sensed data. The server then performs spatial analysis on the relative coordinate system using a regression method, such as Gaussian process regression (GPR) or inverse distance weighting. We validate our approach with two open datasets: a meteorological dataset from the Japanese Meteorological Agency and Intel Lab Data. In both datasets, our results demonstrate that the proposed method can realize spatial regression analysis with less than 10% accuracy degradation in terms of median root mean squared error compared to GPR on absolute coordinates.