Learning software agent design for semiconductor tool group dispatching

Abstract

AbsiracfGExploiting the knowledge-based sofmare agent technology, this paper presents a design and its prototype implementation for learning agent-based semiconductor tool group dispatching. Core to the agent-based dispatching are decision and sequence tree-bused howledge representation models, the learning and acquisition mechanisms of dispatching policy and integration with fab information system and management. Protofype implementation and test experiments over an industry compatible environment demonstrate both feasibility and potential effectiveness of the learning sofmare agent to support tool group dispatching.