Bayesian graphical models for indoor localization in MTC deployment scenarios
Hilleshein, Henrique (2021-02-12)
Hilleshein, Henrique
H. Hilleshein
12.02.2021
© 2021 Henrique Hilleshein. Tämä Kohde on tekijänoikeuden ja/tai lähioikeuksien suojaama. Voit käyttää Kohdetta käyttöösi sovellettavan tekijänoikeutta ja lähioikeuksia koskevan lainsäädännön sallimilla tavoilla. Muunlaista käyttöä varten tarvitset oikeudenhaltijoiden luvan.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:oulu-202102271237
https://urn.fi/URN:NBN:fi:oulu-202102271237
Tiivistelmä
Herein, we propose and assess an iterative Bayesian-based indoor localization system to estimate the position of a target device. We describe the Bayesian network and then build graphical models for various measurement metrics, namely Received Signal Strength (RSS), Time Difference of Arrival (TDOA), and Angle of Arrival (AOA) which are collected by the distributed receivers in the network area. The estimations are carried out by Markov chain Monte Carlo (MCMC) methods which approximates the target’s position using the Bayesian network model and measurements collected by the receivers. We employ an iterative method by using previous estimations of the target’s position as prior knowledge to improve the accuracy of the subsequent estimations, where the prior knowledge is used as the prior distributions of our Bayesian model.
In our results, we observe that the proposed iterative localization system improves the performance of the Bayesian TDOA-based localization system by increasing the respective estimate accuracy. Furthermore, we show that the number of measurements collected by the receivers and the selected prior distribution also affect the performance of the proposed iterative mechanism. In fact, the number of measurements increases the accuracy of the mechanism, while its benefit diminishes with more iterations as the mechanism progresses. Regarding the prior distribution, we show that it can lead to good or bad estimations of the target’s position, and therefore, needs to be carefully chosen considering the measurement metric and the mobility of the target node.
In our results, we observe that the proposed iterative localization system improves the performance of the Bayesian TDOA-based localization system by increasing the respective estimate accuracy. Furthermore, we show that the number of measurements collected by the receivers and the selected prior distribution also affect the performance of the proposed iterative mechanism. In fact, the number of measurements increases the accuracy of the mechanism, while its benefit diminishes with more iterations as the mechanism progresses. Regarding the prior distribution, we show that it can lead to good or bad estimations of the target’s position, and therefore, needs to be carefully chosen considering the measurement metric and the mobility of the target node.
Kokoelmat
- Avoin saatavuus [31657]