Multiprocessor control of robotic manipulators

Abstract

SIERA (System for Implementing and Evaluating Robotic Algorithms) is a multiprocessor system that has been developed to support research in robotic algorithms. It consists of the tightly-coupled, bus-based Real Time Servo System (RTSS) and the loosely-coupled, link-based Armstrong Multiprocessor. This hybrid architecture was selected to combine the benefits of both systems. The tightly-coupled RTSS provides the low communication overhead that is necessary for real-time control. The loosely-coupled Armstrong Network provides expandability and reconfigurability, and can perform the high-level planning required to coordinate robots and other devices in an intelligent manner. An important research contribution of this dissertation is the demonstration that a hybrid architecture, such as the one employed in SIERA, is both desirable and feasible. Overhead on the real-time system is reduced by performing many of its operating system functions, and the less time-critical robot computations, on the loosely-coupled system. SIERA is a flexible research system that allows virtually all functions related to operation of the robot (e.g. control, kinematics, trajectory planning) to be interactively modified. Furthermore, the control law can be changed while the robot is in operation. This is important because a robot interacting with its environment must be able to adapt to new situations. The modular construction of the hardware and software allows SIERA to be applied to different robots. Presently, SIERA is used to control an IBM 7565 Cartesian robot and a Puma 560 revolute robot. Various position and force controllers have been implemented using SIERA. This thesis presents some research results in the area of compliant control to illustrate operation of the system. A damping controller with saturation is designed to control position or force for a single axis. The concept of dual-drive control is subsequently developed to allow desired velocity and force constraints to be satisfied without prior knowledge of the task. The damping controller with saturation provides the force control algorithm, and allows the desired velocity to be easily incorporated. Implementation results for the two-dimensional case, which includes surface-following and crank-turning experiments, demonstrate the potential of these algorithms.