Resolving Users' Behavior Modeling Ambiguities in Fuzzy-Timed Smart Homes Using Only RFIDs

摘要

Context aware systems successfully exploit the knowledge of the users’ actions?presumably in concordance with the environmental parameters of the smart space that they dwell in such as, their location inside the smart space, current time, etc?in providing ambient-intelligent services like automatic, reactive lighting-appliance control and proactive temperature control. In the past, this kind of intelligence has been realized using high-end sensing technologies like video cameras, microphones and environmental sensors such as pressure sensing pads, ambient light-level sensors, temperature sensors, etc. In this paper, we attempt to realize location, time and behavior aware smart spaces using only Radio Frequency Identification Device (RFID) Technology. RFIDs have been effective in object and people tracking, but using RFIDs results in ambiguities in inferring users’ activities accurately. We propose to resolve this ambiguity using Bayesian Belief Networks (BBNs). We employ a learning system that models time as fuzzy slots to assist users’ location prediction. Our system uses an unobtrusive, negative reinforcement learning (NRL) technique that learns users’ behaviors without querying the users of their actions?as typically been done in previous implementations to improve prediction accuracy. The key contributions of our work are that we have proposed a novel method for modeling users’ behaviors using RFID technology and shown the experimental results of the same.