Abstract-This paper extends the kinematic manipulability concept commonly used for serial manipulators to general constrained rigid multibody systems. Examples of such systems include multiple cooperating manipulators, multiple fingers holding a payload, multileg walking robots, and variable geometry trusses. Explicit formulas for velocity and force manipulability ellipsoids are derived and their duality explained. Singularities are classified into two types: 1) unmanipulable singularity; 2) unstable singularity. The former is similar the singularities in serial chains where velocity manipulability ellipsoid is degenerate and force manipulability ellipsoid infinite. The latter is unique to parallel mechanisms, the velocity manipulability ellipsoid becomes infinite and force manipulability degenerate. In the case of multifinger grasp, these concepts correspond to unmanipulable or unstable grasps.

No abstract

No abstract

or robotic tasks involving contact between the robot end F effector and the environment, force feedback is frequently used to maintain the required force of interaction. Among the many force control strategies proposed in the literature, integral force feedback has been found to be the most desirable algorithm due to its robustness with respect to the measurement time delay and its removal of steady state force error. However, there has not been any serious investigation of the controller performance under large force disturbances. We have experimentally observed that large force disturbances can cause bouncing instability of a nominally stable force control system. Motivated by this observation, we augment the standard integral force controller with three robustness enhancements: integral error scaling, force setpoint scheduling, and integral windup saturation. Extensive experimentation on surfaces with different stiffness has shown the dramatic improvement of the modified controller. Modifications to Improve PerformanceMost robotic tasks involve contact between the robot end effector and the environment. Many of these tasks require the force of interaction to be held within a range, for example, handling of fragile payloads, parts insertion and removal, precision assembly, deburring and machining, etc. If the kinematic model of the robot and the environment is known exactly, then the setpoint of the position loop can be selected based on the force control objective. In general, this assumption is unrealistic, and force feedback is required for the force regulation. Many force control algorithms have been proposed in recent years, for example, [ 11-[3], to name just a few. Most of the existing algorithms can be classified as either a direct or indirect method. In the direct method, force error is used directly to servo the robot, while in the indirect method, which includes impedance control and position accommodation, force error is converted into a position error which in tum drives the robot through the position control loop. In [4], it has been shown that direct integral force feedback control [and generalized damper position accommodation, which is effectively the same as integral force feedback) possesses many desirable properties, such as robustness with respect to the force measurement delay, and removal of the steady-state force error. This observation is further confirmed by the experimental results in [3] which applies a combined feedback linearization and integral force control on a direct drive arm. A fundamental assumption in the stability analysis for the force control loop is that the arm maintains contact with the environment. This is clearly unreasonable in many applications.We present an experimental study of the performance of integral force control when the disturbance force is large enough to cause the arm to break off the contact. We have observed that when the force loop is tight [i.e., high integral feedback gain), good performance is obtained while the arm is in contact with the environ...