Agents dealing with Time and Uncertainty

摘要

Situated agents in the real world need to handle the fact that events occur frequently, as well as the fact that the agent typically has uncertain knowledge about what is true in the world. The ability to reason about both time and uncertainty is therefore very important. In this paper, we develop a formal theory of agents that can reason about both time and uncertainty. The theory extends the notion of agents described in [10, 21] and proposes the notion of temporal probabilistic (or TP) agents. A formal semantics for TP-agents is proposed - this semantics is described via structures called feasible TP-status interpretations (FTPSI's). TP-agents continuously evaluate changes (in the state of the environment they are situated in) and compute appropriate FTPSI's. For a class of TP-agents called positive TP-agents, we develop a provably sound and complete procedure to compute FTPSI's.