Recognition Of Object’s Softness And Hardness Based On Tactile Information Of Robotic Hand

With the rapid development of technology, intelligent robot has been being applied in many fields such as medicine, military, agriculture and industry. In practice, robot’s ability to grab object is particularly important. Hardness, as one of the most important physical properties of an object, affects the controlling function of the robotic hand to grasp object to a certain degree. People can identify the hardness of an object accurately through complex tactile sensation system, and then use moderate strength to grasp. However, this is not a simple task for a robotic hand. Most of the robotic hands only have the basic feedback of pressure. It is difficult to identify the hardness of an object directly, which affects the robotic hand to grasp the objects with different hardness. A feedback system of tactile information is built on the platform of bionic mechanical arm in this paper, which enables the arm to identify the hardness of the target object and guarantees stable grasping. Specific work is as following:(1) With the demand of recognizing the target object’s hardness, we develop a feedback system of tactile information, including finger sensor, sensor signal collection circuit, microcontroller, robotic hand movement control interface and sensor data display, etc. The sensor signal acquisition and data communication are implemented in the hardware unit of the tactile information feedback system. Sensor signals are amplified through signal processing circuit and the DMA transmission mechanism is used to realize real-time collection and digitization. Bluetooth module is used for data communication by microcontroller, which simplifies the circuit connection. In this case, tactile signal could be transmitted to PC wirelessly. What’s more, the software unit of the tactile information feedback system mainly completes the design of PC program. On PC program interface, we can control robotic hand to grasp, monitor and storage tactile sensor data real-timely, and debug the whole system freely.(2) We study robotic hand’s ability to recognize the hardness of target object with the tactile information feedback system. Four levels of hardness are set as the standard to identify different objects in the experiment and different objects are grasped repeatedly according to the standard. At the same time, tactile sensor signals are collected and analyzed. Three types of data are chosen as eigenvalues to identify the hardness of the target objects. Then K-nearest neighbor algorithm is used to fuse these eigenvalues. This enables the robotic hand to distinguish different unknown objects by recognizing the hardness. Compared with single sensor information, multi-sensor information is more comprehensive, which obviously increases the accuracy of hardness recognition. Finally, combined with the recognition of object’s hardness, the stable grasping control system is designed. From this experiment, we found that when grasping soft object, its deformation degree is decreased significantly. The result shows that the control system enables the robotic hand to apply moderate force according to the hardness to grasp objects stably.