Automated Model Generation and Simulation

초록

Understanding or predicting the behavior of a complex physical system requires the use of a model, which is often handcrafted by the person studying the system. In this paper we discuss the problem of automating the process of model building and simulation, and an implemented system called Oracle. Given a description of a problem involving a physical system, Oracle automatically identifies relevant model fragments, instantiates them for the particular entities and physical phenomena in the problem, composes the instantiated fragments to form a model, and transforms the model into an executable simulation program. Knowledge of physical phenomena is represented with general model fragments which can be shared and reused by many models. Oracle has been successfully tested on several physics and mechanical design problems. The contributions of our approach are (1) it can automatically model and simulate mechanical systems with multiple moving objects in arbitrary configurations from first principles, and (2) the approach is domain-independent and extensible to multiple physical phenomena.