Hierarchical Proximity Sensor for High-Speed and Intelligent Control of Robotic Hand

This study proposes a hierarchical proximity information processing system and a novel proximity sensor for realizing high-speed and robust grasp control of a robotic hand. The sensor requires both fast response and advanced situation judgment abilities. Therefore, a function that digitally samples individual reaction amounts of all detection elements is added to the net-structure proximity sensor (NSPS), which extracts the sum and center position of the distribution of reaction amounts of detection elements by high-speed analog computation on the sensor circuit. To integrate these two functions, we construct a circuit design method that enables the coexistence of a multichannel A/D converter circuit on the analog computing circuit of the NSPS without disturbing the current flow for sensing; the proposed sensor is called the “hierarchical proximity sensor.” An analysis of its characteristics indicates that the sensor can be used for feedback control of the fingertip position/posture and to estimate the curvature of objects. Through an experiment conducted using a robotic hand equipped with the proposed sensor, we confirmed that the fingertip can approach an object in 0.18 s based on the high-speed analog computation information, while the information for improving the motion can be obtained by comparing the temporal change in the finger joints with the digital sampling information of the process.