Discrimination of Material Properties by Touch Sensors based on The Neuro Algorithm

Abstract

This paper presents a study on discriminating material properties by touch sensors based on the neuro algorithm. The authors firstly pointt out that the multi-layer artificial neural network has good results for discriminating material properties. However, it exhausts tremendous time to get synapses by the standard learning method (SLM). To solve the problem, the authors propose an improved neuro algorithm with such process as dividing all samples into several groups, learning synapses within a group by the mean learning method (MLM), and learning synapses between groups by the simulated annealing.
In the paper, a numerical simulation is also described for verifying the proposed algorithm, from which good classification resulted for some materials such as iron, aluminum, copper, brass, rubber, wood and styrene foam. We think that the study is useful to developing an artificial skin tactile sensing system in the future.