Learning plans with patterns of actions in bounded-rational agents

• Main Authors: Subagdja B., Sonenberg L.
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2005
• Subjects: Bounded Rationality; Databases and Information Systems
• Online Access: https://ink.library.smu.edu.sg/sis_research/6832
• Institution: Singapore Management University

This paper presents a model of a learning mechanism for situated agents. The learning is described explicitly in terms of plans and conducted as intentional actions within the BDI (Beliefs, Desires, Intentions) agent model. Actions of learning direct the task-level performance towards improvements or some learning goals. The agent is capable of modifying its own plans through a set of actions on the run. The use of domain independent patterns of actions is introduced as a strategy for constraining the search for the appropriate structure of plans. The model is demonstrated to represent Q-learning algorithm, however different variation of pattern can enhance the learning.