Patrick J. Eicker CQNfSandia National Laboratories Albuquerque, New Mexico, USA

Transcription

1 L!m 3/vD--qg-183 Trends in Robotics A Summary of the Department of Energy s Critical Technology Roadmap Patrick J. Eicker CQNfSandia National Laboratories Albuquerque, New Mexico, USA I.O INTRODUCTION 1.1 The Purpose of Technology Roadmaps Technology roadmaps serve as pathways to the future. They call attention to future needs for research and development; provide a structure for organizing technology forecasts and programs; and help communicate technological needs and expectations among end users and the research and development (R&D) community. Critical Technology roadmaps, of which the Robotics and Intelligent Machines (RIM) Roadmap is one example, focus on enabling or cross-cutting technologies that address the needs of multiple U.S. Department of Energy (DOE) offices. Critical Technology roadmaps must be responsive to mission needs of the offices; must clearly indicate how the science and technology can improve DOE capabilities; and must describe an aggressive vision for the future of the technology itself. The RIM Roadmap defines a DOE research and development path for the period beginning today, and continuing through the year 22. Its purpose is to identify, select and develop objectives that will satisfy near- and long-term challenges posed by DOE S mission objectives.if implemented, this roadmap will support DOE S mission needs while simultaneously advancing the state-of-the-art of RIM. For the purposes of this document, RIM refers to systems composed of machines, sensors, computers and software that deliver processes to DOE operations. The RIM Roadmap describes how such systems will revolutionize DOE processes, most notably manufacturing, hazardous and remote operations, and monitoring and surveillance. The advances in DOE operations and RIM discussed in this document will be possible due to developments in many other areas of science and technology, including computing, communication, electronics and micro-engineering. Modern software engineering techniques will permit the implementation of inherently safe RIM systems that will depend heavily on software. I.2 DOE Needs Define the Objectives for RIM Identification of the current challenges and hture needs of the DOE forms the foundation for understanding the ways in which RIM will play a key role in enabling DOE to achieve its various mission objectives. Several cross-cutting themes are evident in the complete list of needs. Among these are: 81119s 1

2 This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any infonnation, apparatus, product, or proctss disclosed, or represents that its usc would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement. recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

3 DISCLAIMER Portions of this document may be illegible electronic image products. Images are produced from the best available original document.

4 Worker health and safety. The DOE intends to continue removing workers from the hazards of radioactive, explosive, toxic and other hazardous materials. Product quality. The advent of RIM provides the DOE with the opportunity to eliminate many design and production-related defects. Reduced cost. The capabilities of RIM have the potential to advance so rapidly that initial capital costs of RIM systems will be easily compensated for by a decrease in operating costs. Increased productivity. Remote systems of the past were characterized by the slow operation needed to ensure safety. RIM will improve safety, increase efficiency, and enable much higher facility throughputs. Illustrations of how these themes will revolutionize four of DOE s operations are provided below. RIM will Revolutionize Manufacturing DOE s nuclear weapons design and manufacturing complex consists of an integrated system of design laboratories, production plants, and operations offices, which are responsible for the design, development, re-manufacture, maintenance, and security of our nation s nuclear stockpile. Recent changes in the international political arena, including bans on nuclear testing, an increasing emphasis on environmental and worker safety, and a desire to reduce costs and improve efficiency are motivating DOE to change many of its practices. For example, to meet its future goals, DOE will have to approach manufacturing in a revolutionary manner. RIM will play a central role in this revolution, and will contribute substantially to the ability to increase efficiency and reduce costs, meet changing regulatory requirements, attract and retain a skilled workforce, and ensure safe work environments. RIM will enable a manufacturing system in which: The design-to-manufacturing process will be computer-based and use interactive, collaborative environments. Master Craftsman factories will incorporate a thorough understanding of the materials and their history through previous manufacturing steps, awareness of the intended function of the product, and the ability to accommodate past design and processing decisions to make the entire system work. Weapons products will evolve toward a solid state appearance with higher reliability and lower cost. Controlled features will be automatically inspected, with a guarantee that the inspection process will be done similarly every time. Even features that are subjective, such as surface finish and color will be systematically inspected and recorded. 2

5 Completely remote, autonomous processing will eliminate worker exposure to radiation and increase cost effectiveness by making facilities less sensitive to regulation. Production and design records will be automatically generated and will contain manufacturing and other historical information specific to each warhead, RIM will Revolutionize Hazardous Operations By developing sensor and model-based automated processes, and by improving the capability of existing remote technology, emerging RIM will permit DOE to significantly reduce the exposure of its workers to hazards. Today, many operations in DOE depend on workers using gloveboxes, respirated suits, longhandled tools, and other means for protection from hazards. These techniques also are used in laboratory research involving hazardous chemicals. While they have successfully served DOE missions in the past, improved methods are needed as indicated by accidents which have occurred in recent years. Today, RIM technologies are beginning to make significant contributions to worker safety and protection of the environment. Recently developed mobile sensor platforms allow facility floor radiation mapping to be performed with reduced worker radiation exposure, improved data accuracy and integrity, and higher productivity. In addition, dismantlement of explosive weapon components has been accomplished with much reduced human risk using robotic operations. We can envision a future in which advanced RIM technologies will enable the near total removal of humans from exposure to hazardous environments. RIM will Revolutionize Remote Operations Remote operational techniques were originally developed to allow Manhattan Project researchers to perform tasks that involved dangerous and often lethal levels of ionizing radiation safely. Over the years, these techniques have evolved from shielding walls and long-handled tools. to the use of anthropomorphic mechanical master-slave manipulators, electro-mechanical and hydraulic manipulators, and other types of transport systems. In these systems the human operator is responsible for all of the control signals to the remote devices-a function requiring great concentration and making the operator susceptible to distraction and fatigue, particularly in continuous, production-like operations. Although advances in electronics and computers have enabled significant improvements in operator interfaces, sensory feedback, and remote manipulator controls, the unstructured and uncertain environments that are typical of these operations have limited progress. New RIM technologies will increase productivity, decrease secondary waste generation, and increase the safety of remote operations by performing a wide range of processes, including: Mapping and characterization of buried waste sites, underground storage tanks and retired facilities; Retrieval, sorting, and segregation of waste from various types of sites; Dismantlement of large facilities; Processing of heterogeneous waste streams; Inspection of critical components with higher accuracy and fewer false positives; and 8/1/98 3

6 Incorporation of human perception into remote tasks with greater fidelity and accuracy. RIM will improve the productivity, quality, cost, and safety of remote operations. In a number of instances, RIM will enable operations which are today too hazardous for both humans and remote sytems. Automation of tasks will allow human operators to perform higher level direction rather than being tied to the tedious execution of tasks. Instead, remote operators will serve as managers and organizers of the autonomous robot systems which do the work. RIM will work together symbiotically, under clear human control and supervision; the combination of human and machine will have better than human capabilities. RIM will Revolutionize Monitoring and Surveillance The United States no longer produces new fissile materials, and must dispose of large amounts of existing materials in a way that makes them difficult to reconstitute for weapons use. High levels of security must be provided and an absolute ability to monitor and track the location and fate of these materials must be ensured. Today, dealing with the existing quantities of fissile and radioactive materials requires at least some level of worker exposure. Russia and other nuclear powers have similar materials disposition needs and concerns. Domestic and international pressures to provide material disposition with greater accuracy and reliability, reduced worker exposure rates, and at lower cost will require a revolution in DOE S approach to materials handling, monitoring, and security. For example, RIM can provide the capability to process plutonium from metal to oxide, avoiding the need for exposing workers and at the same time reducing the generation of secondary waste. RIM will enable significantly higher density storage of radioactive materials because humans will no longer need access to the storage vaults. Mobile RTM will be sealed in the vaults, autonomously roam the area, respond to anomalies and communicate over wireless networks. Detailed packaging histories will be generated and downloaded to a secure database without human involvement. RIM will provide the capability to extend security levels far beyond those employed today. Automated RIM guards will communicate constantly, respond collectively and intelligently, and will show no reluctance to signal if their performance is inadequate. In addition to sights and sounds, RIM will routinely observe and record small temperature differences and the presence of organic compounds or radiation. Where necessary, RIM will generate multiple expert opinions, allowing human operators to implement appropriate responses from great distances. 2. THE SCIENCE AND TECHNOLOGY BASIS OF RIM A unified approach is necessary for discussing the R&D that lies ahead, and ultimately for managing a cohesive and integrated S&T program. During the course of defining the RIM technology plan, four basis science and technology areas were identified:

7 Perception Reasoning Action, and Novel interfaces and integration systems. Action Science and Technology for RIM The ability to move and manipulate objects in space is a key capability of RIM, and the science and technology that enables it is of critical importance in DOE operations. RIM of the future will be required to implement many different operations on materials of varying hazards and forms. Physically handling objects with traditional remote manipulators requires extreme care and focus on the part of the operator and is thus slow, inefficient, and becomes unsafe if the operator becomes fatigued or loses concentration-as happens regularly. New manipulation technologies are needed to automate much of the manipulation and decision-making currently performed by a human operator. Twenty years from now, the goal for RIM is to develop more sophisticated robotic structures, actuators, controllers, human-machine interface controls, and auxiliary systems. Such devices and tools will include grasping systems and tactile hands, sensors, inspection and vision systems, cutting, digging, surface removal, cutting tools, and more. General requirements for the robotic machines of the future include accommodating task-appropriate payload, precision, speed, and reach. Over the next twenty years, the RIM research challenges include the development of: Dexterous manipulators. General design techniques for systems with highly complex, nonlinear components. Monitoring systems and design for fault tolerance for maintenance and error recovery. New structural and assembly concepts, control elements, and actuators with higher power-to-weight ratios. Perception Science and Technology for RIM Perception systems provide a means for RIM to gather information about the working environment-information that permits operations such as manufacturing processing, navigation, monitoring, and manipulation to be accomplished safely and precisely. As such, they are an essential component of the RIM of the future. Recent developments in sensor technologies promise a new generation of devices that are more sensitive, more accurate, and extend the realm of perception to a broader range of phenomena. These sensors will also enable more efficient processing of information, a function vital to the operation of RIM systems. Faster and better perception systems will enable quicker autonomous site characterization by teams of RIM; improved perception technologies will be able to discriminate between solvent waste, radioactive waste, and mixed waste, and respond accordingly. Increases in perception accuracy and efficiency will enable DOE S weapon system programs to prevent manufacturing defects, enable smaller production runs, facilitate more flexible manufacturing operations, and reduce costs. 8/1/98 5

8 The capabilities of RIM perception technologies are being dramatically affected by developments in electronics and photonics processing capabilities. Emerging RIM sensor technoloa, including novel acoustic sensors; chemistry labs-on-a-chip; hyperspectral imagers; and nanoscale devices for measuring pressure, temperature and nanosize particles will be highly valuable for robots deployed in DOE applications. However, improvements in sensory devices can only be fully exploited if methodologies for integrating such diverse sensors are developed. At present, the integration of perception capabilities into RIM systems is more of an art than a science. Thus, one of the ultimate goals for perception R&D is to be able to simply plug-in a new sensor anywhere in a RIM system, and for the system to be able to assimilate it automatically. The capability to easily integrate new sensors into RIM systems will result in significant reductions in the cost of developing, deploying and operating RIM systems for practical DOE applications. Reasoning Science and Technology for RIM Reasoning is the smarts of an intelligent machine. It is the capacity to reason that provides the complex connection between perception and action. Without reasoning, machines are relegated to perform static, repeated actions that do not respond or adapt to a changing environment. Reasoning for RIM is more challenging than standard computer reasoning because of the tight coupling that exists between the RIM and the physical world, because RIM must move about in hazardous environments and manipulate materials with the highest standards of safety, and because the real-world applications envisioned for RIM will require them to make intelligent and safe decisions on their own, without having to be explicitly guided by a human. Advances in communication, memory and foresight, algorithms, and the ability to cooperate andlor act autonomously will substantially increase cost and time savings, productivity and safety in DOE activities. For example: In the near-term, reasoning will enable small lot production operations. DOE will be able to provide a RIM with a model of a part and a model of the production operation and expect the RIM to use its reasoning capabilities to automatically produce a program to be employed by the production equipment. In the long-term, reasoning will enable teams of RIM devices to autonomously cooperate and communicate with each other to remove hazardous material from buried waste sites and separate it into discrete packages. These systems will completely plan and execute their coordinated actions, adapting as needed in response to a dynamic environment or a changing mission. With sufficient technical advances, these systems will be robust, reliable, and flexible, and will be automatically designed and programmed. Novel Interfaces and Integration for RIM RIM components that provide perception, reasoning and action capabilities must be easy to integrate and easy to use. The integrated RIM of the future will offer many benefits:

9 They will offer interfaces that are as intuitively understandable as the best of today s personal computers and applications programs; They will be easy to bring quickly into a state of safe and reliable operation; and They will be easy to program for specific, individual operations. An intuitive human-computer-machine interface for RIM does not yet exist. Robotics engineers, still program with outdated computer languages and non-intuitive hand held devices called teach pendants. This is one reason that robotic systems for the DOE remain expensive on a per-unitprocessed basis. Making RIM accessible to non-specialists will involve much more research in the area of better interfaces, including virtual reality, and devices for manipulation of virtual and real objects. In addition, the RIM systems of the future will require operating systems (OS) that provide all the usual capabilities provided by an OS, are nearly invisible to the user, assure the safety requirements of the DOE, and support plug and play. We are far from having such operating systems today. However, if the RIM vision is to be achieved, novel interfaces and integration systems must make operation of RIM as simple as pointing and clicking on today s personal computer. 3. DELIVERING NEW CAPABILITIES TO DOE DOE s offices do not think of RIM in the context of its reasoning or perception capabilities-rather their interest is in RIM as a deliverer of processes; in essence asking the question, what can RIM help me do better? To answer this question, we can conceive of RIM as providing DOE with four types of processes/capabilities: Mapping and modeling Non-contact operations Contact Operations, and Multi-operations Mapping and Modeling. Although important and relevant research on collective intelligence and emergent behavior is being conducted within DOE s research community (and elsewhere), reasoning using computer models is viewed as the means to achieve the envisioned intelligence for the foreseeable future. The watchwords are: If you have a computer model, use it to reason about the process being delivered; if you don t have a computer model, gather the information to build a model. A program plan for R&D must have an element which addresses the requirements for modeling

10 Non-contact Operations. One type of function of a RIM is to deliver and use sensors which do not require contact for their successful operation. RIM that inspect products during and after manufacture are examples, as is the monitoring of a materials storage vault. Contact Operations. In many cases, RIM must physically manipulate materials. The reason for differentiation from non-contact operations is the greater danger of unsafe operating conditions. For example, in a contact operation, much care must be taken to assure that the right amount of force-and no more-is applied. Contact operations are more complicated and involve more safety concerns than non-contact operations. Multi-operations. These are complex operations which require many contact and non-contact operations to work together in an integrated way. RIM will be required to map and generate models of objects to be sorted, direct graspers to pick up objects, etc. Multi-operations are the most complex of the operations discussed here. In general, each type of operatiodprocess described requires the some of the capabilities of the previous type. The increase in complexity implies that development will be somewhat sequential in time. The illustrates this concept Coat Human operator, or first-of-a-kind automated. Fully automated, complex, multi-operation The development of RIM-deliverable processes will evolve over the timeframe covered by the RIM Roadmap. The table below illustrates the relation of RIM-deliverable processes to the RIM basis science and technology: action; reasoning; perception; and the human interface to computer and machine. The trajectories of each are keyed to the timeframe of development of the RIM-deliverable processes. To bridge the gap between the Basis Science and Technology Areas identified as underpinning RIM, and DOE S view of RIM as a provider of processes and operations; a table summarizing necessary RIM technology development is provided below. 8 8/1/98

11 Reasoning S&T for RIM Perception S&T for RIM Action S&T for RIM Novel Interfaces and Systems for Integration Incremental mapping Motion planning: geometry-based Motion planning: -Geometry-based -Process modelbased -Sensor-based Geometry mapping Survey for A, B, C Plug and play sensors Safety Process quality assurance Plug and play sensors Safety Mobile Arm-type Integration of mobile and motion Integration of mobile and process control Interface for safe mapping operations Interface for human manipulation of virtual objects I Fine-motion planning Sensor-based control Plug and play sensors Process quality assurance Safety System-level planning Planningfor integration of operations Plug and play sensors Safety Dextrous manipulators Heavy and large object manipulation Cooperative control of multiple devices Lnterface for human manipulation of remote real objects Interface for system-level interaction ~ 4. CONCLUSION: THE LONG TERM VISION FOR RIM AND ITS APPLICATIONS IN DOE DOE S Motivations for RIM 4.1 Today, the Department of Energy is poised to simultaneously improve its operations and significantly accelerate the evolution of robotics and intelligent machines. The reasons for this are many and include: The DOE has ample motivation for deployment of RIM: - Improved health and safety of its employees - Improved design and production infrastructure - Improved productivity and cost of its environmental management operations. No other agency-public or private-has the breadth of operations, processes and material forms. This permits a sweeping view of the possibilities for application of RIM as well as the requirement for new technologies. The motivations and the breadth will endure into the foreseeable hture. The Department can take the long view, deploying mature RIM technologies while developing new

12 .. c While the Department s materials and operations are unique in many cases, many of the RIM technologies it needs will be useful to industry and to other agencies. 4.2 The Long Term Vision for RIM Building upon and complementing the continued rapid evolution of computing, communication and micro-engineered technologies, the research and development discussed in this roadmap will drive the basic science and technology of RTM within DOE-and beyond. By the year 22, advanced RIM technologies will fundamentally change the manner in which people use machines. RIM will perform virtually all hazardous operations. The S&T basis developed for DOE S hazardous operations will be further developed and used in the hazardous operations of other agencies such as the DoD-in a host of ways. It is expected that: Microscale robots with the ability to crawl, fly, and swim will be able to work together to perform monitoring, surveillance, and intelligence operations, as well as in-service inspection of critical national assets; Environmental cleanup, monitoring and inspection operations, and resource exploration will be performed with high efficiency through fully autonomous teams of robots; Automated methods for design and manufacturing will allow designs to be proposed in a simple manner and all other follow-on operations for production of both large and small lot manufacturing products will be totally automated. Teams of cooperating robots will work 24 hours a day to accomplish difficult tasks with high quality and safety. The paradigms that presently define the relationship of humans and RIM will fundamentally change. People will work directly with RIM in complete safety and will interact with multiple intelligent machines through sensory, immersive interfaces that intelligently adapt to human and supervisor desires. RIM health monitoring and maintenance will be fully automated. Power sources and communications will no longer inhibit missions, and all RIM systems will be constructed and configured through fully automatic plug-and-play approaches. By the year 22, RIM will both duplicate and extend human dexterity, perception, and work efficiencies in broad ranges of tasks across the U.S. economy. Acknowledgement: The author wishes to acknowledge the team of individuals from the Department of Energy and its national laboratories and plants who contributed to the development of the Robotics and Intelligent Machines Roadmaps. The document is currently in draft and will be published later in Sandia is a multiprogram laboratory 8/1/98 operated by Safidia Corporation, a Lockheed Martin Company, for the United States Department of Energy under contract DE-AC4-94ALS5. 1