Reliable context capturing for smart offices using a sensor network

In order to realize smart offices, which provides users with comfort via various digital services, we need to acquire context information about the target space. Contexts can be obtained from raw sensor data using context classification methods, such as Bayesian network. However, packet losses and disrupted communications in wireless sensor networks disables the context classification methods to collect all the necessary data, hence reduce quality of contexts. In this paper, we propose Reliable Hybrid Bayesian Inference Mechanism (RHBIM) that features in-network disruption-tolerant Bayesian inference with server-side calculation of posterior probability tables. In this paper, we show the design and implementation of the mechanism with a range of disruption-tolerance schemes, and apply the mechanism to an application that controls air conditioners based on the “comfort level” context. We also show the effectiveness of the mechanism comparing the different disruption-tolerance schemes.