Hardness Recognition of Fruit and Vegetable Based on Robot Tactile Perception

Aiming at the inaccurate recognition of fruit and vegetable hardness in the damage-free grasping operation of picking robot, we propose a method called proper orthogonal decomposition-artificial neural network (POD-ANN) in this paper. First, the symmetrical clamping mechanism of two fingered manipulator is constructed, and the pressure series data set is established through the fingertip flexible tactile sensor. Second, the classification and recognition results of the collected tactile feature sequences are compared with support vector machine (SVM) and K - nearest neighbor (KNN) algorithms. The comparison results show that the POD-ANN algorithm achieves a recognition rate of 88.1% on the two finger robotic arm platform, which has higher recognition accuracy compared to the other two methods. Simultaneously, it also shows great potential in tactile classification tasks.