C. K. Ji (Australia)
intelligent resource-bounded agents, plan management,chance discovery, belief and expectation
For real-time and rational adaptation to dynamic and uncertain environments, a resource-bounded agent should focus not only her deliberation onto her commitments to the future, but also her attention onto relevant features. Such features include opportunities or risks whose attraction to agent's attention takes a crucial but probably potential role, since chances are often difficult or impossible to predict but nevertheless necessary to get ready for. Hence the agent's plan management must allow for autonomous anticipation for unexpectedly significant situations, which is defined as the meta-level of autonomous chance discovery. This paper presents a three-level structure of attentive perception to extend IRMA into AP-IRMA that focuses agent's attention toward chances for effective plan management, which integrates learning and planning, and as well reduces the entire procedure into several functions with respective simple computations in parallel.
Important Links:
Go Back