Use of goals for creating and enacting dynamic contracts in ambient intelligence

Abstract

Ambient Intelligence (AmI) systems support everyday lives of humans by sensing and reacting to the environment when necessary. Devices of the AmI world cooperate with each other to achieve a common task by using the information they capture from the environment and through their internal knowledge. They make their decisions based on their reasoning processes and strategies. To increase the satisfaction of the user, several types of applications cooperate with each other such as audio, visual and health care related applications. This cooperation is hidden from the user. This thesis proposes to view AmI devices as autonomous agents with di erent capabilities and goals and the entire AmI system as a distributed multiagent system. Goals refer to the states that the agent desires to achieve. The proactive behavior of the AmI devices are modeled with the goals of the agents. Each agent within the system is motivated to satisfy its goals. The agent selects actions that will have positive consequences on satisfying its goals. The agent decides and initiates relevant actions to satisfy its goals one by one. If the agent has the capability to achieve its goal, it just executes it. If the agent's capabilities are inadequate for satisfying its goal, then it asks for help from other agents. Agents use contracts to interact with each other, which are regulated with a commitment-based methodology. We provide algorithms and reasoning rules to help agents create and respond to the commitments in di erent situations. We apply our approach on an intelligent kitchen domain consisting of smart kitchen devices such as a co ee maker, a refrigerator and a user agent representing the resident of the kitchen. We demonstrate how our approach can accommodate several realistic cooperation scenarios.