3 - A Domain-Independent Approach

Abstract

Problem solving with building augmented generalizations by generating extended recurrences' (BAGGER) situation calculus rules can be viewed as transforming and expanding situations until one is found in which the goal is known to be achieved. The BAGGER system has two types of inference rules: (1) inter-situational rules that specify attributes that a new situation will have after application of a particular operator, and (2) intra-situational rules that can embellish BAGGER's knowledge of a situation by specifying additional conclusions that can be drawn within that situation. Each inter-situational inference rule specifies knowledge about one particular operator. However, operators are not represented by exactly one inference rule. A major inference rule specifies most of the relevant problem-solving information about an operator. However, it is augmented by many lesser inference rules that capture the operator's frame axioms and other facts about a new situation. The advantage of the STRIPS approach is that the system can always be assured that it has represented all that there is to know about a new situation.