Validation of a Property Estimation Method Based on Sequential and Posteriori Estimation

Abstract

Robots are being developed to perform tasks in homes and factories autonomously. Several studies have examined motion generation based on haptic information, and some studies consider the environment as physical property information. However, there is a trade-off between the accuracy and the time required for the physical property estimation. Therefore, in this study, we propose a method for estimating physical properties based on training the relationship between the two estimation models. The first is a fast sequential estimation model, and the second is a highly accurate posterior estimation model. Training the relationship between the two models makes highly accurate sequential property estimation possible. Validation results showed improved accuracy of property estimation for learning samples and some untrained samples.