Tactile sensing based softness classification using machine learning

The research on tactile sensors and its wide applications have received extensive attention among researchers very recently, especially in the two fields-Medical Surgery (Minimally Invasive Surgery-MIS) and Fruit and Vegetable Grading Industry. This paper proposes the implementation of a robotic system which can distinguish objects of different softness using machine learning approach, based on different parameters. Two piezoresistive flexible tactile sensors are mounted on a two fingered robotic gripper, as robotic arm can perform repetitive tasks under a controlled environment. A PIC32 microcontroller is used to control the gripping action and to acquire pressure data. Decision Tree and Naive Bayes methods are used as intelligent classifiers using feature vectors, obtained from the time series response of tactile sensors during grasping action for grading the objects. From the analytical point of view it is observed that Decision Tree based approach is better than the Bayesian approach.