Experiments in the dynamic and strategic control of cooperating manipulators

Abstract

This thesis comprises an experimental study of a semi-autonomous robotic system with multiple manipulators. Four topics are studied in detail: hierarchical real-time system design, conceptual operator command, dynamic control of cooperative manipulators, and integrated real-time vision. The goal is to study simultaneously the dynamic and strategic issues of cooperative robotic manipulation. This work focuses not only on the various subsystems, but also on their interfaces and interactions. The system is structured as a four-level hierarchy. At the highest level, an iconic object-only user interface allows an operator to direct the conceptual activities of the system. The operator commands only motions; the arm actions required to effect these motions need not be specified. An event-driven tabular finite state machine provides strategic command. This technique encourages modular design of multi-process programs, provides an intuitive task programming environment, and naturally manages the concurrent system interactions. The dynamic controller implements object impedance control--an extension of the impedance control concept to cooperative-arm manipulation of a single object. This controller presents an intuitive behavior specification interface, and provides good dynamic performance both for free-motion positioning and environmental contact tasks. A real-time point-tracking two-dimensional vision system locates and tracks passive targets. Groups of targets can be identified and tracked as individual objects. The system can track multiple objects at 60 Hz with sub-millimeter resolution. The design was verified by experimental implementation: a multi-processor real-time computer controls a dual-arm planar manipulator system. Experimental results are presented, showing the system locating and identifying a moving object, catching it, and performing a simple cooperative assembly. These operations are controlled by a remote user entering only high-level conceptual relations. Results from dynamic control experiments show excellent dynamic trajectory tracking performance, while also permitting control of environmental contact forces.