Combining Activity Recognition and AI Planning for Energy-Saving Offices
Ilche Georgievski, Tuan Anh Nguyen, Marco Aiello · 2013
Energy-saving offices require autonomous and optimised control of integrated devices and appliances with the objective of saving energy while the occupant comfort and productivity are preserved. We propose an approach that analyses and controls an office space and accounts for the objectives of energy-saving offices. The approach considers ontology-based occupant activity recognition using simple sensors to process the context information, and employs Artificial Intelligence planning to control appliances. The approach is evaluated in a semi-simulated setting. The activity recognition strategy is tested in an actual living lab and shows recognising accuracy of about 80%. The planning technique is able to cope efficiently under a simulated and increasing number of offices and recognised activities. The overall solution shows intriguing potential for energy saving in the order of 70%, given mostly sunny days and a provisional set of devices for experimentation.