Fingertip Flexible Tactile Sensor Design And Manipulation For Dexterous Robot Hand

Through utilizing information of necessary sensors, dexterous robot hand can effectively fulfill grasping and manipulating tasks for many objects. The information of contact state and location from fingertip tactile sensor is critical to improve the practical manipulation capacity(especially rolling contact manipulation capacity) of dexterous robot hand. Therefore this dissertation develops a fingertip tactile sensor for HIT/DLR Hand II multifingered robot hand that is developed by Harbin Institute of Technology and German Aerospace Center Institute of Robotics, and investigates the manipulation control based on this tactile sensor.To study the crosstalk problem of piezoresistive array tactile sensor, the most simplified circuits to weaken the crosstalk are firstly structured based on the voltage feedback method and the zero-potential method, respectively; then, the effects of analog switch and multiplexer on-resistance values on measure accuracy of each circuit are simulated and analyzed; finally, the performances of these circuits are comparatively analyzed. This provides some guidances to design a crosstalk weakened circuit for the follow-up fingertip tactile sensor.A flexible sticking fingertip tactile sensor based on Piezoresistive principle is developed for the robot hand in this paper. The tactile sensor has 36 units. To develop undevelopable surface, a method of combining inserting darts with finite element analysis is proposed. This method firstly inserts darts to properly split undevelopable surface, then uses finite element analysis software to approximately develop the split surface. The tactile sensor that its shape is designed using this method can fine cover the three-dimensional fingertip surface. To meet the high integration requirements of the robot hand, a embedded signal acquisition and processing system for this sensor is designed based on a Programmable System-on-Chip(PSo C) system. To test the sensor, a tactile sensor performance test system is designed. This tactile sensor system can provide necessary information of contact for the robot hand in the process of grasping and manipulation.In order to improve accuracy of kinematic and mechanical models of the finger, kinematic and mechanical models are constructed based on position information of the tactile sensor. Comparing to models based on virtal fingertip, the proposed models can be effective to improve detecting accuracy of contact position and force in Cartesian space. This establishes the foundation of mathematics for the robot hand control.Cartesian impedance control based on the tactile and torque sensors of the finger is investigated to ensure the flexibility of the finger when contacting with environment. Firstly, an effective algorithm that can be used to calculate the internal force and torque of any point in any link of a serial robot is proposed. This algorithm is derived by Newton-Euler equations using a hypothetical truncated principle of the link. In the process of the derivation, the situations that whether the serial robot is static or applied an external force on its end and whether each joint of the serial robot is a rotary joint are taken into account; then the dynamic model of the link torque sensor space are derived by using this algorithm. By translate the Cartesian force that is calculated with considering the tactile information after the link torque sensor dynamic compensation into adjusting position, velocity and acceleration vectors, two kinds position-based Cartesian impedance control schemes based on the tactile and torque sensors are built. Experimental results not only show that the algorithm and the proposed Cartesian impedance controllers are effective, but also show that the Cartesian impedance control based on the tactile and torque sensors can effectively improve the dynamic position tracking precision of the fingertip. This establishes the foundation for the robot hand manipulation.When dexterous robot hand manipulates unknown objects, the rolling contact may occur, which may induce manipulation failing. To solve this problem, firstly, kinematic and mechanical models of dexterous robot hand manipulation based on the fixed contact point assumption; then a strategy of varying contact point manipulation is proposed based on fingertip tactile sensor. This strategy utilizes fingertip tactile sensor to detect contact position and to update the parameters of the end-link size of the finger when rolling occur without considering object geometric shapes. Experimental results show that the strategy can effectively improve the rolling contact manipulation performance when dexterous robot hand manipulates unknown objects.