Knowledge structures for planning in realistic domains

摘要: Abstract In this paper we discuss techniques for representing and organizing knowledge that enable a planning system to work effectively in large realistic task domains. We consider the following engineering issues: (i) generating manipulating complex plans; (ii) managing bases; (iii) operating with an incomplete base and/or world model. A theme reflected throughout is policy of developing common representations useful many modules system. believe these will be other artificial intelligence systems similar demands on reasoning skills. present uniform descriptions objects actions, including partial indefinite descriptions, domains, trace process includes alternative solutions, (iv) organizational mechanism both as well gathered during (e.g. constraints among actions contained plan solution).