A neural approach to robotic haptic recognition of 3-D objects based on a Kohonen self-organizing feature map

The paper describes a versatile robotic haptic recognition system of 3D objects. The design methodology features a learning phase of the geometric properties of the objects, followed by the operative phase of actual recognition in which the robot explores the objects with its end-effector, correlating the sensorial data with the preceding perceptive experiences. These phases are mapped on the training and classification activities typical of the unsupervised Kohonen neural networks. The system consists of a dexterous 3-fingered, 10-DOF robotic hand. In a primary trial test, the developed prototype system has already shown a satisfactory operative level in recognizing objects