Reasoning About Acting, Sensing, and Failure Handling: A Logic for Agents Embedded in the Real World

We are interested in agents which have to "work" in the real world. An agent which works in the real world must be able to act and to handle failures in acting. Indeed, most often an agent can not avoid failures, since the external environment is unpredictable and uncompletely known. Moreover, since the environment is not known a priori, most often the only way for an agent to work is to acquire information from the external environment. In agents embedded in the real world, information can be acquired through sensory devices. We propose a basic theory of agents which describes the fact that an agent acts, fails in acting, handles failures in acting and needs to sense the external environment as well as to handle failures in sensing. This work is built on a logical theory for acting, sensing and planning and extends it by providing a way to reason about agents acting, sensing and failure handling. This is achieved by introducing into the theory the so called "laws of motion", which describe uniformly reasoning about effects of agent's actions, perceptions and failure handling.