Recognition of Environmental Impedance Configuration by Neural Network Using Time-Series Contact State Response

Abstract

There is a need to use robots in a variety of environments. If the environment can be recognized, it will be possible to generate appropriate motions and judge the situation. The image-based environment recognition does not provide physical information such as the impedance of the target object. This paper proposes a method to recognize the configuration of objects by touching them before performing actions. By conducting simulations, the results of the recognition with a neural network (NN) using time-series data of the action were verified.