Joint Torque Sensory in Robotics

Motivated by the need for accurate joint torque sensing in robots, we designed a new torque sensor, based on the hollow hexaform geometry. Its key features were its extremely high sti ness and its insensitivity to the set of support forces and moments which persist in a robot joint. These features permit to mount the sensor directly in the joints of a robot manipulator leading to accurate joint torque sensing and to a compact and modular design. The structure of the sensor also exhibits strain concentration to torsion loads which maximizes the sensitivity to torsion without sacrificing torsional sti ness. Other design issues such as practical shape consideration, material properties and overloading also considered. The sensor geometry was analyzed and optimized using the finite element method. The sensor was tested extensively to confirm its design goals, and is well suited as a torque-sensing device in robots or other industrial high performance motion control applications. A quantitative comparison with di erent types of sensors is shown in table 1. The table indicates that our sensor's performance characteristics compare very favorably. The applications of adaptive control in conjunction with joint-torque sensory feedback was used for dynamic motion control of manipulators. The control system had the advantages of requiring neither the computation of link dynamics nor the precise measurement of joint torques, i.e., the torque sensor's gains and o sets are unknown to the controller. The adaptive controller could also tune all the joint parameters including the rotor inertia, twist.