Affiliations: Computer Science Research Institute and School of Computing and Mathematics, Faculty of Computing and Engineering, University of Ulster, Newtownabbey, Northern Ireland, UK
Correspondence:
[*]
Address for correspondence: Michael Poland, Smart Environments Research Group, University of Ulster, Shore Road, Newtownabbey BT37 0QB, Co. Antrim, Northern Ireland, UK. Tel.: +44 2890 36 8918; E-mail: poland-m@email.ulster.ac.uk.
Abstract: Smart homes are living spaces facilitated with technology to allow individuals to remain in their own homes for longer, rather than be institutionalised. Sensors are the fundamental physical layer with any smart home, as the data they generate is used to inform decision support systems, facilitating appropriate actuator actions. Positioning of sensors is therefore a fundamental characteristic of a smart home. Contemporary smart home sensor distribution is aligned to either a) a total coverage approach; b) a human assessment approach. These methods for sensor arrangement are not data driven strategies, are unempirical and frequently irrational. This Study hypothesised that sensor deployment directed by an optimisation method that utilises inhabitants' spatial frequency data as the search space, would produce more optimal sensor distributions vs. the current method of sensor deployment by engineers. Seven human engineers were tasked to create sensor distributions based on perceived utility for 9 deployment scenarios. A Pure Random Search (PRS) algorithm was then tasked to create matched sensor distributions. The PRS method produced superior distributions in 98.4% of test cases (n=64) against human engineer instructed deployments when the engineers had no access to the spatial frequency data, and in 92.0% of test cases (n=64) when engineers had full access to these data. These results thus confirmed the hypothesis.
Keywords: Sensor optimisation, smart homes, pure random search
