Energy saving is essential for life on Earth. Hence, it is fundamental in developing new important ambient intelligence (AmI) systems, for example, applications in smart homes, health, transportation, emergency, education, and the workplace. Artificial intelligence (AI) is also utilized for smart automation and human-computer interactions. Current AmI applications are mostly for network protocol unification and multimedia streaming, yet they only marginally apply to real energy savings.

The paper presents a well-crafted three-layered architecture (hardware, middleware, and application) for the authors’ automated comfort and energy smart “International Hellenic University” (IHU) project. At the hardware layer, there are three sets of efficiently deployed devices. Smart plugs (sensor/actuator encrypted ZigBee network) devices are used to control and measure appliances’ power consumption. Smart clampers devices are used to wirelessly measure a whole building’s consumption. Finally, sensor boards on a ZigBee network are deployed to report back (via ZigBee-to-TCP/IP gateways) temperature, luminance, and humidity. The middleware service-oriented layer handles (1) the hardware platform and data heterogeneity, where every cluster of devices runs an instance of the middleware’s web server; (2) sensor fusion/analysis, to provide presence detection; and (3) deliberation about energy management by actuators. Its main challenge is abstracting different hardware devices’ programming, network management, and topology/properties to the application layer. The application layer uses the authors’ previous work, service-oriented Web standards middleware, aWESoME-S, to support/unify a variety of platform-independent applications. Two complementary and mutually exclusive rule-based agents are presented for better energy management: a reactive Wintermute agent and a deliberative, defeasible logic (DeL) agent; analysis showed that the DeL agent was favored.

This paper presents an enhancement over the existing middleware that features semantic web services, semantic annotations for the Web service definition language (SAWSDL), which supports the required higher-level intelligence. Another main contribution is the well-designed and thorough analysis of the aforementioned three-layered agent-based energy-saving smart IHU system and its real-world IHU project. Although the IHU project integration of the reactive and deliberative rule-based agents resulted in only a small four percent daily energy savings, it is still very advantageous to guarantee the enforcement of all energy-saving behavior, and ensure that energy-wasting behavior is kept to a minimum.