A Multiagent System Architecture is proposed, which is a framework that facilitates combining multiple agent into a team, and is well suited to addressing the unique requirements for dynamic scheduling in hot strip mill. Several problem-solving agents using different strategies were incorporated, which can be classified to four kinds: construction agents, improvement agents, destruction agents and evaluation agents. Three kinds of construction and improvement techniques were introduced including: mathematical algorithms, domain dependent heuristics and search algorithms. The prototype system has been implemented in C++ Builder, according to the object-oriented paradigm. With minor modifications, these techniques will provide near-optimal schedules for the difficult existing scheduling problems at hot strip mills.
展开▼
