This thesis presents a novel method to improve the accuracy of Pedestrian Dead Reckoning (PDR) localization in indoor environments. PDR, that is widely used for indoor positioning, tracks the location of a target by integrating measurements of step length and heading provided by inertial sensors. Step length can be estimated relatively accurately, whereas the estimation of heading in indoor environments is highly problematic. A common way to estimate the heading in PDR is to use magnetometer measurements; however, unlike outdoor environments, the Earth's magnetic field is strongly perturbed inside buildings making the magnetometer measurements unreliable for heading estimation. The main purpose of this work is to estimate the heading of a group of people that are walking in the same direction by using only magnetic sensors embedded into smartphones. The novelty of this work is in the proposed method: in a fi rst phase the PDR system fi lters perturbed estimates by means of a machine learning algorithm, in a second phase it exploits collaboration among users to fuse multiple heading estimates and gain spatial diversity. Fusion approach has been realized both in centralized and in distributed way; in particular, diff erent consensus algorithms are proposed to test performance in terms of convergence time and localization accuracy at convergence. Several measurement campaigns have been conducted by means of smartphones in order to analyze magnetic sensor properties in indoor environments and evaluate the PDR performance. The encouraging results show that by the proposed approach the heading estimate error is highly reduced thus leading to noticeable enhancements in localization. This work has been carried out as a cooperation between University of New South Wales (UNSW) and Politecnico di Milano. Part of the activities has been developed during a 5-month visit at UNSW, Computer Science Engineering Department, from March to August 2014. Part of this work has been presented at the Indoor Positioning and Indoor Navigation (IPIN) 2014 Conference, October 27-30, Busan, Korea, with the paper: Marzieh Jalal Abadi, Luca Luceri, Mahbub Hassan, Chun Tung Chou, Monica Nicoli, "A Collaborative Approach to Heading Estimation for Smartphone-based PDR Indoor Localisation".
Questo lavoro di tesi presenta un nuovo metodo per migliorare l'accuratezza nella localizzazione basata su sensori inerziali o Pedestrian Dead Reckoning (PDR), in ambienti chiusi (indoor). PDR è un sistema ampiamente utilizzato per la localizzazione indoor. La posizione del soggetto da localizzare è stimata combinando le misure della lunghezza e della direzione degli spostamenti. La prima può essere stimata con buona accuratezza, mentre la stima della seconda è molto problematica in ambienti indoor. Un metodo largamente di ffuso per stimare la direzione in PDR è quello di utilizzare sensori magnetici. Tuttavia in ambienti chiusi il campo magnetico terrestre è altamente disturbato rendendo le misure del magnetometro inaffidabili per la stima della direzione di percorrenza. Lo scopo principale di questo lavoro di tesi è quello di stimare la direzione di percorrenza utilizzando il sensore magnetico incluso negli smartphone e sfruttando la cooperazione fra i dispositivi di più persone che camminano nella stessa direzione. La novit a di questo lavoro è nel metodo proposto: in una prima fase le stime corrotte dai disturbi vengono fi ltrate per mezzo di un algoritmo di machine learning; nella seconda fase le stime dei diversi utenti che hanno superato il fi ltraggio vengono combinate per produrre una stima cooperativa. Questa tecnica cooperativa è stata realizzata utilizzando sia un approccio centralizzato che uno distribuito; in particolare diversi algoritmi di consenso sono stati analizzati per validare le prestazioni in termini di tempo di convergenza ed errore nella stima a convergenza raggiunta. Diverse campagne di misure sono state condotte con smartphone in maniera tale da analizzare le proprietà del sensore magnetico in ambienti indoor e valutare le performance del sistema. I risultati mostrano che la riduzione dell'errore di stima della direzione comporta un notevole miglioramento nell'accuratezza della localizzazione. Il lavoro di tesi è stato condotto in collaborazione con University of New South Wales (UNSW). Parte delle attivit a sono state sviluppate durante un periodo di visita di cinque mesi presso il dipartimento di Computer Science Engineering di UNSW, da Marzo ad Agosto 2014. Inoltre parte di questo lavoro è stato presentato alla conferenza Indoor Positioning and Indoor Navigation (IPIN) 2014, 27-30 Ottobre, Busan, Korea, con l'articolo: Marzieh Jalal Abadi, Luca Luceri, Mahbub Hassan, Chun Tung Chou, Monica Nicoli, "A Collaborative Approach to Heading Estimation for Smartphone-based PDR Indoor Localisation".
Distributed dead reckoning for cooperative indoor localization
LUCERI, LUCA
2013/2014
Abstract
This thesis presents a novel method to improve the accuracy of Pedestrian Dead Reckoning (PDR) localization in indoor environments. PDR, that is widely used for indoor positioning, tracks the location of a target by integrating measurements of step length and heading provided by inertial sensors. Step length can be estimated relatively accurately, whereas the estimation of heading in indoor environments is highly problematic. A common way to estimate the heading in PDR is to use magnetometer measurements; however, unlike outdoor environments, the Earth's magnetic field is strongly perturbed inside buildings making the magnetometer measurements unreliable for heading estimation. The main purpose of this work is to estimate the heading of a group of people that are walking in the same direction by using only magnetic sensors embedded into smartphones. The novelty of this work is in the proposed method: in a fi rst phase the PDR system fi lters perturbed estimates by means of a machine learning algorithm, in a second phase it exploits collaboration among users to fuse multiple heading estimates and gain spatial diversity. Fusion approach has been realized both in centralized and in distributed way; in particular, diff erent consensus algorithms are proposed to test performance in terms of convergence time and localization accuracy at convergence. Several measurement campaigns have been conducted by means of smartphones in order to analyze magnetic sensor properties in indoor environments and evaluate the PDR performance. The encouraging results show that by the proposed approach the heading estimate error is highly reduced thus leading to noticeable enhancements in localization. This work has been carried out as a cooperation between University of New South Wales (UNSW) and Politecnico di Milano. Part of the activities has been developed during a 5-month visit at UNSW, Computer Science Engineering Department, from March to August 2014. Part of this work has been presented at the Indoor Positioning and Indoor Navigation (IPIN) 2014 Conference, October 27-30, Busan, Korea, with the paper: Marzieh Jalal Abadi, Luca Luceri, Mahbub Hassan, Chun Tung Chou, Monica Nicoli, "A Collaborative Approach to Heading Estimation for Smartphone-based PDR Indoor Localisation".| File | Dimensione | Formato | |
|---|---|---|---|
|
2014_12_Luceri.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
8.27 MB
Formato
Adobe PDF
|
8.27 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/102682