Diagrams are direct knowledge representation approach to dealing with topolog-ical and geometrical relations. Diagrammatic reasoning aims to perceive, revise, andmanipulate diagrams to facilitate problem solving, having effective context representa-tion, high reasoning efficiency and intelligible interpretation. While automated reason-ing using surrogate symbolic representation, diagrams are direct representation. There-fore, combining surrogate and direct representation is a natural approach to diagram-matic reasoning, which can reserve the characteristics of diagrams and maintain theefficiency of symbolic reasoning.In this paper, we present a hybrid diagrammatic reasoning architecture based onconstraint satisfaction. Followings are the main processes of the architecture.1. By means of diagram perceiving, domain dependent spatial problems are trans-formed to domain independent spatial problems, and solved in spatial problemsolver.2. Objects and spatial relations perceiving from diagrams are represented in con-straints, casting spatial reasoning in constraint satisfaction problem. Time com-plexity is analyzed.3. The effectiveness of the architecture is demonstrated with pulley block forceanalysis. |