抄録
This paper proposes an algorithm that estimates a reaction torque loaded to a grasping jaw of a wire actuated robotic forceps and compensates for an amount of a wire elongation that subsequently occurs. The algorithm is constructed by the combination of Reaction Force Observer (RFOB) and an optimization problem based on a gradient descent method. The conventional RFOB is used to estimate an external torque which includes not only the reaction torque but also an acceleration of the wire elongation. Then, the amount of the wire elongation is directly obtained from the differential equation of the wire elongation. On the other hand, as a solution of the optimization problem, the acceleration of the wire elongation is obtained. Thus, the reaction torque is derived by subtracting the acceleration of the wire elongation from the estimated external torque by the RFOB. Our previous algorithm required an assumption that the motor torque and the reaction torque are regarded as constant, although in practice they fluctuate. The New algorithm succeeded in eliminating this assumption, and resulted in great improvement on accuracy of estimating the reaction torque. The improvement was ensured through simulation works.
