Verbal Representation of Object Collision Prediction Based on Physical Common Sense

Abstract

Recently, prediction models for the real world have been widely proposed. It is thought that when humans predict the real world, they capture the observed object’s change points and understand its situation in language. Based on this approach, in this study, we extended several prediction models and constructed change point prediction models that predict collision timing from changes in object motion. Moreover, the predictions of these models were generated as sentences containing environmental physical common sense, and the accuracy was verified.