Contract based cooperation for ambient intelligence : proposing, entering and executing contracts autonomously

Abstract

As we are introduced with the more complex, high technology devices, the need to smoothly integrate these new devices to our lives grows. In order to satisfy this need, a new kind of pervasive and ubiquitous intelligence, Ambient Intelligence, is suggested. Ambient Intelligence (AmI) describes environments that sense and react to the humans in time to help improve their living quality. Software agents are thus important in realizing such environments. While existing work has focused on individual agent’s reactions, more interesting applications will take place when agents cooperate to provide composed services to humans. When cooperation is required, the environment needs mechanisms that regulate agent’s interactions but also respect their autonomy. Accordingly, this thesis develops a contract-based approach for offering composed services. The system consists of two ontologies that keep knowledge about the environment and the domain, agents that represent the user, and the agents that provide services. Similar to the real life, individual services are usually offered by different agents. Agents that are going to take part in the contracts are selected according to their capabilities. At runtime, selected agents autonomously decide whether they enter contracts. The decision is based on the knowledge which is described in the ontologies and the internal state of the agent. After the agents agree on the contracts, they then act to fulfill their contracts. We apply this multiagent system on an intelligent kitchen domain and show how commitments can be used to realize cooperation. We study our application on realistic scenarios.