Title

Online Map Matching for Passive Indoor Positioning Systems

Published In

Proceedings of the 8th ACM SIGSPATIAL Workshop on GeoStreaming

Document Type

Citation

Publication Date

2017

Subjects

Passive indoor positioning systems (PIPS)

Abstract

Passive indoor positioning systems (PIPS) enable non-invasive tracking of mobile users using enterprise network infrastructure. PIPS-generated location estimates tend to be noisy, infrequent, and trail actual locations. Timely discernment of user intent, such as which hallway a user is walking on, or when the user made a turn requires online map matching, i.e., given PIPS-generated location estimates and a floor map, determining which among many possible indoor path segments (e.g., a hallway) a mobile user is currently on. Existing map matching approaches apply geometric, topological and semantic constraints to constrain the location estimates on to the predefined path segments a user can take. However, no prior work has investigated which combination of these constraints produces robust map matching results for PIPS. In this paper, by analyzing the existing constraints, the trajectories of the location estimates collected on two different building floors with PIPS deployments, and by performing a preliminary study of human mobility on three other building floors, we find that a point-to-curve proximity constraint, a travel distance constraint, and a number of turns constraint can be applied to perform robust map matching. We also propose a novel Hidden Markov Model-based map matching algorithm to apply these constraints. To account for trailing location estimates, we propose a more suitable accuracy metric to measure the accuracy of map matching algorithms for PIPS.

Rights

Copyright © 2021 ACM, Inc.

DOI

10.1145/3148160.3148161

Persistent Identifier

https://archives.pdx.edu/ds/psu/35632

Share

COinS