Albert J. Wavering

TRICLOPS: a tool for studying active vision

The design, performance, and application of The Real-time, Intelligently ControLled, Optical Positioning System (TRICLOPS) are described in this article. TRICLOPS is a multiresolution trinocular camera-pointing system which provides a center wide-angle view camera and two higher-resolution vergence cameras. It is a direct-drive system that exhibits dynamic performance comparable to the human visual system. The mechanical design and performance of various active vision systems are discussed and compared to those of TRICLOPS. The multiprocessor control system for TRICLOPS is described. The kinematics of the device are also discussed and calibration methods are given. Finally, as an example of real-time visual control, a problem in visual tracking with TRICLOPS is examined. In this example, TRICLOPS is shown to be capable of tracking a ball moving at 3 m/s, which results in rotational velocities of the vergence cameras in excess of 6 rad/s (344 deg/s).

Parallel Kinematic Machine Research at NIST: Past, Present, and Future

Unchained from the confines of controller limitations, the industrial application of parallel kinematic machines in manufacturing is beginning to emerge. The launch of commercially available hexapod machine tools at the 1994 International Manufacturing Technology Show in Chicago represented the first radical departure in machine tool design since the introduction of numerical controls. The parallel actuator technology promises to offer manufacturers a number of advantages relative to conventional machine tools, such as a higher stiffness-to- mass ratio, higher speeds, higher accuracy, reduced installation requirements, and mechanical simplicity. Several machine tool makers in the U.S. and around the world are pursuing parallel actuator technology, while their prospective customers — manufacturers — are beginning to contemplate what the novel multi-axis machining technology might mean for their operations.

TRICLOPS: a high-performance trinocular active vision system

The design, performance, and application of the real-time, intelligently controlled, optical positioning system (TRICLOPS) are described. TRICLOPS is a multiresolution trinocular camera-pointing system which provides a center wide-angle view camera and two higher-resolution vergence cameras. It is a four-degree-of-freedom direct-drive system which exhibits dynamic performance comparable to the human visual system. The vergence degrees of freedom can achieve peak velocities in excess of 30 rad/s, and peak accelerations of 1100 rad/s/sup 2/. In an example of visual tracking, TRICLOPS is shown to be capable of following a ball moving at 3 m/s, with rotational velocities of the vergence axes in excess of 6 rad/s.<<ETX>>

The NASREM robot control system standard

Abstract The problem of robot control is approached from a systems standpoint where a complete control system must include all of the aspects involved in moving a robot, not just the algorithms in the classic controls literature. The NASA/NBS Standard Reference Model for Telerobot Control System Architecture (NASREM) provides the framework for a complete manipulator control system. It is composed of three hierarchies: task decomposition, world modeling, and sensory processing. The task decomposition hierarchy divides the task into smaller substasks. In order to achieve the desired decomposition, the task decomposition hierarchy must often access information stored in the world modeling hierarchy, which contains a workspace representation, object descriptions, robot models, etc. The sensory processing hierarchu constantly fills the world model with processed sensor information. The NASREM functional architecture was developed for NASAs Flight Telerobotic Servicer, a two-armed robot which will build and maintain the Space Station. However, the control concepts proposed by the NASREM functional architecture immediately transfer to other applications such as in manufacturing, autonomous vehicles, etc.

High-Performance Tracking with TRICLOPS

This article reviews the development of experimental visual tracking algorithms which have been implemented using TRICLOPS (The Real-time, Intelligently-ControLled Optical Positioning System), a high-performance active vision system designed and built at the National Institute of Standards and Technology (NIST). These algorithms range from triangulation using simple targets for very high-speed tracking, to model-based techniques which allow robust tracking of polyhedral targets amidst cluttered backgrounds. In all cases, effective prediction of the target motion to compensate for image processing delays is critical to achieving responsive tracking.

The real-time control system of the Horizontal Workstation robot

A 200-word or less factual summary of most significant information. If document includes a significant bi bliography or literature survey, mention it here) This manual describes the real-time control system used to control the robot in the Horizontal Workstation of the Automated Manufacturing Research Facility.

An Architecture To Support Autonomy, Teleoperation, And Shared Control

Described is an approach to the functional architecture of a telerobot so that autonomy, teleoperation, and shared control can all be supported. The system is hierarchically organized where task decomposition, world modeling, and sensory processing are explicitly represented. Goals at each level of the hierarchy are decomposed spatially and temporally into simpler tasks which become goals for lower levels. The spatial decomposition facilitates control and coordination of multi-arm robots. A description of the lowest level, the Servo Level, is presented, along with the operator control interface at that level.

Real-time control system software: Some problems and an approach

The National Bureau of Standards Center for Manufacturing Engineering is currently conducting research to help define interface and data format standards necessary to support the factory of the future. In support of this goal, NBS has implemented an experimental factory, called the Automated Manufacturing Research Facility (AMRF). The evolution of the AMRF has included the development of a Real-Time Control System (RCS) to direct robot motions according to high-level commands received from a supervisory workstation controller and environmental data obtained by local sensors. During the development of RCS, it became clear that large-scale software for real-time control is fundamentally different, and in some aspects, more complex than that intended for scientific or business applications. Specifically, software designed for control purposes must contend with a large number of non-deterministic states which complicates program synthesis and makes testing and debugging extremely difficult. This paper discusses these problems, and how they are addressed by RCS.

Experimental Evaluation of Cartesian Stiffness Control on a Seven Degree-of-Freedom Robot Arm

The programmability of Cartesian stiffness in Cartesian servo control algorithms that do not use explicit force feedback is examined. A number of Cartesian algorithms are implemented and evaluated on a commercial seven degree-of-freedom robot arm, using the NASREM robot control system testbed. It is found that Cartesian servo algorithms which use the transpose of the Jacobian and model-based gravity compensation, provide easy programmability and accurate reproduction of stiffnesses over a wide range. When dynamic behavior is a consideration, dynamic damping control, augmented to include a parameterization of the manipulator self-motion, provides superior performance and programmability.

Task Decomposition Module For Telerobot Trajectory Generation

A major consideration in the design of trajectory generation software for a Flight Telerobotic Servicer (FTS) is that the FTS will be called upon to perform tasks which require a diverse range of manipulator behaviors and capabilities. In a hierarchical control system where tasks are decomposed into simpler and simpler subtasks, the task decomposition module which performs trajectory planning and execution should therefore be able to accommodate a wide range of algorithms. In some cases, it will be desirable to plan a trajectory for an entire motion before manipulator motion commences, as when optimizing over the entire trajectory. Many FTS motions, however, will be highly sensory-interactive, such as moving to attain a desired position relative to a non-stationary object whose position is periodically updated by a vision system. In this case, the time-varying nature of the trajectory may be handled either by frequent replanning using updated sensor information, or by using an algorithm which creates a less specific state-dependent plan that determines the manipulator path as the trajectory is executed (rather than a priori). This paper discusses a number of trajectory generation techniques from these categories and how they may be implemented in a task decompo-sition module of a hierarchical control system. The structure, function, and interfaces of the proposed trajectory gener-ation module are briefly described, followed by several examples of how different algorithms may be performed by the module. The proposed task decomposition module provides a logical structure for trajectory planning and execution, and supports a large number of published trajectory generation techniques.