Proposal of a Micromachined Tactile Sensor Having Four Stories and Its Information Processing Method using Module Networks

Imitating the human skin structure, an arrayed type tactile sensor composed of many force sensing elements distributed on four stories is proposed. The information processing method for this sensor using neural network (NN) and Support Vector Machine (SVM) is investigated. Since a practical sensor is under development, the finite element method (FEM) simulation is employed to simulate the data from sensing elements of this tactile sensor. While some feature vectors extracted from raw data are usually input to NN, the raw data of all sensing elements’ outputs are input to the NN directly in this study. First, contact force magnitude and its direction are recognized using two stage NNs. Second, the shape of contact object is recognized, confirming that SVM is equivalently effective compared to NN. In both cases of force recognition and shape recognition, a sensor having four stories is confirmed to be superior to that having one story in the viewpoint of recognition accuracy. To summarize, the novelties of this study on tactile sensor are 1) using four stories, 2) using raw data of all sensing elements and inputting them to NN and SVM, and 3) proposing two stage NNs and module networks.