Wolfgang Paetsch

Input-dependent stability of joint torque control of tendon-driven robot hands

The input-dependent stability observed during torque control experiments using the first joint of the Darmstadt-HAND is discussed. Friction and compliance existing in tendon-sheath drive systems introduce a hysteresis nonlinearity between the joint torque output and the actuator displacement. Although this transmission characteristic is close to the well-known backlash behavior of the gears situated between a motor and a load shift, this hysteresis loop exhibits input-dependent characteristics in the backlash region of the transmission system, with springlike behavior within a portion of the backlash region. Experiments confirmed that there is a close relationship between the input-dependent backlash characteristics and the input-dependent stability. Based on these experiments, the authors describe the transmission characteristic using a simple model and explore the system stability using sinusoidal-input-describing-functions (SIDF). A nondimensional stability-criterion-map that successfully predicts the experimental results is presented. >

A three fingered, multijointed gripper for experimental use

A three fingered, multijointed robot gripper for experimental use is presented. The mechanics as well as the control architecture is designed for this special purpose. The gripper system provides the basic means in terms of position and force control to perform experiments about grasping and object motion in a useful way. The gripper can be used to develop and evaluate different approaches of stable grasping and object manipulation. Results of the control of the gripper on joint level, the Cartesian behaviour of the fingers and some experiences with the grasping and manipulation experiments using the presented system are reported.<<ETX>>

Input-dependent stability on joint torque control of robot hand

Input-dependent stability was observed during torque control experiments using the first joint of the Darmstadt-HAND. The friction and compliance existing in tendon-sheath driving systems bring a hysteresis characteristic into the dependence of joint torque output on actuator displacement. While this transmission characteristic is close to the well-known backlash behavior appearing for the gears situated between a motor and load shaft, input-dependent characteristics were found to exist in the backlash region of the transmission system in the case of the use of a practical tendon length. The observed characteristic shows springlike behavior even in the backlash region and it also depends on input. Through experiments, it was confirmed that there are close relationships between the input-dependent backlash characteristics and the input-dependent stability. Based on the experiments, a description is given of the transmission characteristic in simple model equations, and the system stability is explored by using the sinusoidal-input-describing-function technique. A nondimensional stability-criterion map explaining the experimental results is shown.<<ETX>>

Intelligent Real-Time Control of a Multifingered Robot Gripper by Learning Incremental Actions

Abstract Learning control systems are expected to have several advantages over conventional approaches when dealing with complex, high-dimensional processes. One example is the task of controlling grasping operations of a multifingered, mul-tijoined robot gripper, which has been designed and implemented at our robotics lab (the Darmstadt-Hand). The Advanced Gripper Control with Learning Algorithms-AGRICOLA- presented in this paper is able to maintain a stable grasp even if disturbances are applied. Also it works for objects of different sizes for which the grasping has not been learned. Compared to the conventional stiffness approach the performance of the learning system is equal but the design is much easier, since less knowledge about the gripper-hardware has to be taken into account. The main part of the learning control loop is an associative memory storing the grasping behaviour as determined by the choice of an objective function.

INTELLIGENT REAL-TIME CONTROL OF A MULTIFINGERED ROBOT GRIPPER BY LEARNING INCREMENTAL ACTIONS

Learning control systems are expected to have several advantages over conventional approaches when dealing with complex, high-dimensional processes. One example is the task of controlling grasping operations of a multifingered, multijoined robot gripper, which has been designed and implemented at our robotics lab (the Darmstadt-Hand). The Advanced Gripper Control with Learning Algorithms -AGRICOLA- presented in this paper is able to maintain a stable grasp even if disturbances are applied. Also it works for objects of different sizes for which the grasping has not been learned. Compared to the conventional stiffness approach the performance of the learning system is equal but the design is much easier, since less knowledge about the gripper-hardware has to be taken into account. The main part of the learning control loop is an associative memory storing the grasping behaviour as determined by the choice of an objective function.