Indoor Positioning based on RSSI of WiFi signals: how accurate can it be?

Campeón, Juan Pablo Grisales; Lopez, Sergio G.; Melean, S. R. de Jesus; Moldovan, H.; Parisi, Daniel R.; Fierens, Pablo Ignacio

doi:10.1109/argencon.2018.8646171

Cited by 5 publications

(2 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These use either an offline procedure or an online procedure. JP Grisales Campeón et al [ 9 ] used support vector regression (SVR), least absolute shrinkage and selection operator (LASSO), kernel ridge regression, k-nearest neighbors (KNN) [ 10 ], and other methods to filter abnormal signal strength. The best method is SVR, which gives an average positioning error of 1.24 m. H. Zhao et al [ 11 ] used signal pattern comparison, but this requires more time and cost to collect data.…”

Section: Related Workmentioning

confidence: 99%

“…The system compares (x k , y k , z k ) with M si (x i , y i , z i ) in the magnetic field strength database. According to fingerprinting algorithm, the difference values P i between (x k , y k , z k ) and M si (x i , y i , z i ) is as Equation (9). And the minimum value of P k is determined in Equation (10).…”

Section: Magnetic Field Strength Measurement and Online Calculationmentioning

confidence: 99%

See 1 more Smart Citation

A Performance Improvement for Indoor Positioning Systems Using Earth’s Magnetic Field

Yeh

Chiu

Kao

et al. 2023

Sensors

View full text Add to dashboard Cite

Although most indoor positioning systems use radio waves, such as Wi-Fi, Bluetooth, or RFID, for application in department stores, exhibition halls, stations, and airports, the accuracy of such technology is easily affected by human shadowing and multipath propagation delay. This study combines the earth’s magnetic field strength and Wi-Fi signals to obtain the indoor positioning information with high availability. Wi-Fi signals are first used to identify the user’s area under several kinds of environment partitioning methods. Then, the signal pattern comparison is used for positioning calculations using the strength change in the earth’s magnetic field among the east–west, north–south, and vertical directions at indoor area. Finally, the k-nearest neighbors (KNN) method and fingerprinting algorithm are used to calculate the fine-grained indoor positioning information. The experiment results show that the average positioning error is 0.57 m in 12-area partitioning, which is almost a 90% improvement in relation to that of one area partitioning. This study also considers the positioning error if the device is held at different angles by hand. A rotation matrix is used to convert the magnetic sensor coordinates from a mobile phone related coordinates into the geographic coordinates. The average positioning error is decreased by 68%, compared to the original coordinates in 12-area partitioning with a 30-degree pitch. In the offline procedure, only the northern direction data are used, which is reduced by 75%, to give an average positioning error of 1.38 m. If the number of reference points is collected every 2 m for reducing 50% of the database requirement, the average positioning error is 1.77 m.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Magnetic Field Strength Measurement and Online Calculationmentioning

confidence: 99%