Using hidden markov models to improve floor level localisation

Publication Type:
Conference Proceeding
Australasian Conference on Robotics and Automation, ACRA, 2015
Issue Date:
Full metadata record
The focus of this paper is on estimating the oor level of a robot/person moving in a multi-oor environment. It demonstrates how in- formation about transitions between oors can be employed within a probabilistic framework to improve the accuracy of oor level estimation. This is achieved by combining a simple linear classifier with a Hidden Markov Model that captures the two basic motion patterns in a multi-oor environment: Within-oor and be-Tween oors, switching from one to the other as oor transition events are detected. Through real-world experiments, we demonstrate the ability of this framework to produce accurate oor level estimates using only RSSI (Received Signal Strength Indicator) measurements, even when operating in an environment with as little as five WiFi access points per oor.
Please use this identifier to cite or link to this item: