The paper introduces a model based on Semantic Web technologies and on a Ubiquitous computing paradigm, that supports passive adaptation of smart environments to user preferences, without explicit intervention. Passive adaptation differs from the classical active command and control as involved devices interact by specifying preferences. Adaptation to such preferences depends on the internal smart devices policies, and on the active constraints that the device must satisfy. Smart devices, or as they are called by the authors, smart objects publish, in a web-like fashion, information about context, capabilities and constraints. Context information includes information about the perceived environment (either through directly connected sensors or through devices bounded to the smart object). Capabilities describe how can the smart object influence ambient parameters, through its effectors. Finally, constraints declare preferences on the environment state and drive the smart object behavior towards the desired user goals. Adaptation Profiles, i.e. models for automatic passive adaptation of the environment, are composed of two main parts a pre-condition part, that specifies requirements that must be met to activate the profile, and a post-condition part in which the desired patterns in the future environment states are specified. A SOAM orchestrator agent is responsible for the management of Adaptation Profiles, by evaluating profiles against context information and by generating and propagating constraints to the proper smart objects. SOAM is currently being implemented by using the Mika VM on ARM7 UNC20 platforms.

This paper describes a generic ontology for modeling context in AmI. The ontology is organized in 4 main areas involving respectively user, environment, platform and service modeling. User modeling takes into account user profile (preferences), mood and role and relates this information with user tasks and with devices adopted by users to accomplish these tasks. The environment is modeled taking into account simple conditions such as temperature, pressure, humidity, and other environment-related variables such as time and location. The latter is further subdivided in relative and absolute locations. The Platform ontology branch describes the hardware and software platform used to run AmI processes. It encompasses notions about input and output devices, about available resources (power, memory, cpu, etc.), about available software (operating systems, virtual machines, rendering engines and middlewares) and about provided services. Finally, services are described by linking the OWL-S ontology, i.e., they are defined in term of Service Profile (human-readable description of provided service, including service inputs and outputs), Service Model (what happens when the service is carried out) and Service Grounding (implementation details).