Physics problem solver towards mental simulation for real world oriented intelligence

Abstract

General intelligent robots in daily life environment must observe and model external world, understand users' instructions and intentions. In such situations, robots always face to ambiguity of information. Conventional approaches to solve the ambiguity were using common sense database/ontology, or asking users; however developers'/users' cost is too huge. In this paper, we propose an approach in which physics simulation as mental simulation solves the ambiguity instead of descriptive common sense. The target should be focused on physics problems in entrance examination, because physics problems in examination shares important element in general intelligent robots. we state that the required functions of mental simulation for general intelligent robots in real-world can be reduced to functions for physics problem solver. We also propose a framework of software configuration of the physics problem solver.