Next Generation Intelligent Spaces

A building equipped with sensors is a use case of heterogeneous data sources distributed naturally across spaces or zones. Lack of spatiotemporal awareness can lead to excessive sensors or non-optimal distribution across a building. We extend the problem of optimal sensor placement to find the minimal sensor group that can robustly approximate the rest of the sensors values. The solution uses a distributed learning step to parallelize the predictive capability of candidate groups generated by evolutionary computing techniques. Additionally, the system models the cost and energy consumption of the sensor group when performing a multi-objective optimization. We experiment on 24 zones from a seven-storied building in Thailand to establish a trade-off between virtualization accuracy, energy footprint, and installation cost. The work exploits the concept of "less is more" to bring down the capital and recurring expense for a smart building solution