This paper reports our experience with crowd monitoring technologies in the challenging real-world conditions of a modern, open-space museum. We seized the opportunity to use the NEMO science center as a testbed, and studied the effectiveness of neighborhood discovery and density estimation algorithms in a network formed by visitors wearing bracelets emitting RF beacons. The diverse set of conditions (flash crowds in open spaces vs. single person booths) revealed three interesting findings: (i) state-of-the-art density estimation fails in 80% of the cases, (ii) RSS-based classifiers fail too, because their underlying assumptions do not hold in many scenarios, and (iii) neighborhood discovery can obtain exact information in an energy-efficient way, provided that static and mobile nodes are differentiated to filter out “passers by” clobbering the true popularity of an exhibit. The overall lesson from the experiment is that today’s algorithms are quite far from the ideal of monitoring popularity in a privacy-preserving and energy-efficient way with minimal infrastructure across the set of heterogeneous conditions encountered in practice.
Acm Digital Library | Paper
Authors: Marco Cattani and Ioannis Protonotarios and Claudio Martella and and Joost Van Velzen and Marco A. Zuniga and Koen G. Langendoen
Research groups: Embedded Software group, Delft University of Technology and Large-scale Distributed Systems group, VU University Amsterdam
Conference: International Conference on Embedded Wireless Systems and Networks (EWSN)