Agent-based smart environments
My first activities in Nokia Research (1999-2000) were in building multi-agent-based smart environments as well as experimenting with different control interfaces. Initially, I created a smart environment where X10-based devices could be controlled through a gesture-based interface built by adapting a handwriting recognition algorithm to work with a touchpad. For each device brought into the smart environment an agent was created. Each device had an associated profile, represented in RDF. The environment had the following types of agents: System Agents – taking care of agent life cycle issues; Entity Agents – agents for entities; Specialized Agents – distributed yellow-pages – function-based; and Macro Agents (or Context Agents) – agents having a plan for a specific context (e.g. Presentation). RFID tags were used to associate entities to their profiles. In building the multi-agent environment I tested various agent platforms but I eventually selected JADE as the most appropriate. The multi-agent design was also used in the later work focused on home environments: agents were added for access control, negotiating preferences, learning, and various services.