Indoor tracking trajectory data similarity analysis with a deep convolutional autoencoder

Abstract Building energy consumption is influenced not merely by the energy-saving performance of hardware but also by the occupants’ indoor behaviours. Advances in indoor positioning technologies can generate large volumes of spatial trajectory data on the occupants, which can reveal the distribution of the occupants or be interpreted to reflect the occupants’ behaviours. This calls for systematic research on new computing technologies to identify information from trajectory data rather than from visualizations or statistics. Due to the imperfections and complexities of trajectory data, few robust techniques are available for similarity comparisons, which are critical for further clustering and pattern mining. In this work, we propose a novel means of evaluating similarities in occupant trajectory data based on the use of a convolutional autoencoder (CAE). Trajectory data can be compared and their feature vectors extracted with the CAE in an unsupervised manner. We applied this approach to high-precision tracking data collected from an ultra-wide band (UWB) indoor positioning system (IPS) installed in an exhibition hall. The calculated results show that our approach offers great advantages in terms of its application, robustness, and flexibility.

[1]  Christos Faloutsos,et al.  Efficient Similarity Search In Sequence Databases , 1993, FODO.

[2]  Heng Tao Shen,et al.  Searching trajectories by locations: an efficiency study , 2010, SIGMOD Conference.

[3]  Dino Pedreschi,et al.  Trajectory pattern mining , 2007, KDD '07.

[4]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[5]  O. T. Masoso,et al.  The dark side of occupants’ behaviour on building energy use , 2010 .

[6]  En Zhu,et al.  Deep Clustering with Convolutional Autoencoders , 2017, ICONIP.

[7]  Lei Chen,et al.  On The Marriage of Lp-norms and Edit Distance , 2004, VLDB.

[8]  W. Eric L. Grimson,et al.  Learning Semantic Scene Models by Trajectory Analysis , 2006, ECCV.

[9]  Li Li,et al.  Sequential Behavior Pattern Discovery with Frequent Episode Mining and Wireless Sensor Network , 2017, IEEE Communications Magazine.

[10]  Lei Chen,et al.  Robust and fast similarity search for moving object trajectories , 2005, SIGMOD '05.

[11]  Angela Lee,et al.  The impact of occupants’ behaviours on building energy analysis: A research review , 2017 .

[12]  Hwan-Seung Yong,et al.  Mining Spatio-Temporal Patterns in Trajectory Data , 2010, J. Inf. Process. Syst..

[13]  Xing Xie,et al.  Mining Individual Life Pattern Based on Location History , 2009, 2009 Tenth International Conference on Mobile Data Management: Systems, Services and Middleware.

[14]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[15]  Ravi S. Srinivasan,et al.  From occupancy to occupant behavior: An analytical survey of data acquisition technologies, modeling methodologies and simulation coupling mechanisms for building energy efficiency , 2017 .

[16]  Nan Li,et al.  Measuring and monitoring occupancy with an RFID based system for demand-driven HVAC operations , 2012 .

[17]  Ada Wai-Chee Fu,et al.  Efficient time series matching by wavelets , 1999, Proceedings 15th International Conference on Data Engineering (Cat. No.99CB36337).