Safety of Agents (SagA)

THE THEME of this project are the future information agents , also known as ubiquitous or pervasive agents. Ubiquitous information agents have computation abilities and the power to act on their physical environments. Examples include complex systems such as robot companions, self-driving cars (e.g., Stanford or Google), simpler agents as in smart elderly homes and AAL, mobile travel devices, smart refrigerators, or electronic transactions agents. Ubiquitous agents operate in the living environments of people, interacting with (helping) people in their daily life. In consequence, the safety of the decisions and actions of such information agents is of paramount importance. THE OBJECTIVE is to automatically control ubiquitous agents for achieving a desired goal under the restrictions of safety contracts. This would imply a paradigm shift, from the current rule-based systems to goal-based systems . Example: consider a system that shall contain patients in a certain area, unless they are in view of a nurse. • A rule-based system would require the human expert to describe through rules how to restrict the movements of the patients and nurses in order to achieve this goal. • A goal-based system would simply infer/synthesise suitable rules by itself from the goals set by the user.