Print Email Facebook Twitter AirLoc Title AirLoc: Pedestrian dead reckoning for passenger localization Author De Moes, A.F.A. Contributor Venkatesha Prasad, R.R. (mentor) Faculty Electrical Engineering, Mathematics and Computer Science Department Software Technology Programme Embedded Systems Date 2017-01-17 Abstract Travelling through airports can be quite a stressful experience for passengers. Airlines, such as KLM, want to make the journey of their passengers through airports as comfortable as possible. This thesis proposes an indoor localization system for airlines called AirLoc. The research question was: "Which airport specific indoor localization method can be developed, that is ideal for airlines and can be applied at multiple airports?". Different localization methods were discussed and pedestrian dead reckoning was chosen as the most suitable method, because it is infrastructure free and requires no additional investments. AirLoc is implemented on an Android smartphone and uses a motion model as a sensing stage, pedestrian dead reckoning as the localization method and a particle filter combined with landmarks as the refinement method. The location based services include the passenger journey and airport navigation. A particle filter was created that can be used on a complex polygon based map, by using polygon geometry and by creating special collision detection rules and optimizations, such as obstacle distance measurements. Map data of different airports can be included to make the system work at multiple airports. A complete motion model was created that does step detection, step length estimation and heading estimation dynamically. The heading estimation can be used while the phone is in different orientations, by using an initialization stage in which the GPS heading is used to determine the walking direction relative to the orientation of the phone. This estimation has an error of 30 degrees. AirLoc was compared to the Polestar system, which is based on Bluetooth fingerprinting and is used in the Schiphol app. The system was tested in two orientations, hand-held and in a pocket. AirLoc has a mean error of 6.17 meters when held in hand and an accuracy of 7.23 meters when put inside the pocket. The estimated location of AirLoc is always in the correct area and at a reachable location thanks to the collision detection of the particle filter and it removes the glass wall problem introduced by infrastructure based methods. This makes AirLoc more suitable for location based services such as airport navigation. AirLoc consumes only half of the energy consumed by infrastructure based systems, such as Polestar. A system is also proposed that calculates queuing times at the airport security, by using activity monitoring combined with an x-ray landmark. Testing shows that the queueing time could be estimated with an error of 14 seconds. Subject Indoor localizationKLMAirportSchipholParticle filterSmartphoneAndroidPDRpolygon based Map To reference this document use: http://resolver.tudelft.nl/uuid:b381c9e1-ad56-4376-a3a2-12b68938b73a Embargo date 2019-01-17 Part of collection Student theses Document type master thesis Rights (c) 2017 Moes, A.F.A. de Files PDF thesis_afademoes_digital.pdf 9.1 MB Close viewer /islandora/object/uuid:b381c9e1-ad56-4376-a3a2-12b68938b73a/datastream/OBJ/view