This paper developed five control modes for telemanipulators with extra kinematic resources. The modes fall into the categories of manual control, intelligent assistance, and shared control. All five control modes have been implemented and tested on actual telemanipulator hardware. The manual control modes could be applied immediately to D&D tasks. Implementing the shared control modes in actual D&D tasks requires supplying the computer with reliable data describing the telemanipulator’s environment. Clearly, the location of obstacles is of primary importance. Intelligent assistance falls between shared and manual control. This mode is useful when there is some information about the location of obstacles, but the information is either limited or not guaranteed reliable.Testing of the algorithms included computer simulation for the DAWM and hardware implementation in mock D&D tasks with another dual-arm robot having 17 DOF. This paper described a number of the implementation details. Notable among these details is a closed-form
reverse position analysis for the Schilling manipulator geometry. Though the Schilling geometry does not include a spherical wrist, the analysis requires solving only second order polynomials. Other notable details include the geometric transformations for a constraint-tracking technique that dramatically improves the solution speed of
the shared control algorithms.
The testing revealed the efficiency of the control mode depends upon the task at hand. Joint control is best for extricating the robot from difficult configurations, but is not suitable for precisely controlling the robot’s EEF. Decoupled-Cartesian is a general-purpose mode best suited for material transport and tool deployment. The self-motion mode allows for some optimization and is a useful
extension of decoupled-Cartesian control. These first three modes are useful when there is no reliable information about the exact location of obstacles. The advisor mode is useful when there is limited information about the location of obstacles. The coupled-Cartesian is the most efficient mode in terms of task performance, but requires extensive and reliable information about the robot’s environment.
