Crowdsourcing agents for smart IoT.
Journal
Proceedings of the second international conference on Human-agent interaction, HAI '14, Tsukuba, Japan, October 29-31, 2014
Pages
103
Date Issued
2014
Author(s)
Abstract
Activity recognition is a key capability for a smart environment to offer timely services and intelligent interactions with people, especially with the growing number of connected devices. While logging data from connected sensors is no longer beyond reach, it is still quite difficult to collect the labels required by machine learning approaches to activity recognition. In this research, crowdsourcing agents are designed to acquire status labels from people situated in the environment. Experiments on crowdsourcing in a typical building on campus have been conducted to improve air conditioning and space utilization. In particular, we will discuss how crowdsourcing agents in the form of simple physical objects can significantly improve user engagement as well as data quality. Collaboration among cyber-physical agents can lead to better user experience and overall performance.
SDGs
Type
conference paper