Paper Info
Title
An Enhanced Direction Calibration Based On Reinforcement Learning For Indoor Localization System
Abstract
In this paper, we propose an advanced direction calibration method for the smartphone-based indoor localization system on the basis of map information and reinforcement learning (RL). Currently, the direction estimated by pedestrian dead reckoning (PDR) is biased due to the low-precision sensor in smartphone and magnetic field distortion in indoor environment. Thus, the direction calibration methods draw increasing attention. Since the movement of pedestrian is restricted by the indoor environment, the map information could be used to correct the heading of pedestrian and then improve the localization performance. Furthermore, since the tracking of pedestrian can be modeled as a Markov decision process. we propose a novel direction calibration algorithm based on deep Q-network (DQN). Different from the traditional direction calibration algorithms that usually rely on image processing, the proposed method use DQN to find an optimal policy to determine the moving direction. We conduct experiments in a realistic representative office environment to reveal the validity of the proposed direction calibration algorithm. The experiment results indicate that the proposed algorithm can remarkably alleviate the cumulative error, and improve the accuracy, stability and robustness of the indoor positioning system.
Year
DOI
Venue
2020
10.1109/WCNC45663.2020.9120672
2020 IEEE WIRELESS COMMUNICATIONS AND NETWORKING CONFERENCE (WCNC)
DocType
ISSN
Citations 
Conference
1525-3511
0
PageRank 
References 
Authors
0.34
0
3
Name
Order
Citations
PageRank
Qiao Li100.34
Xuewen Liao28815.95
Zhenzhen Gao39819.18