1、_SAE Technical Standards Board Rules provide that: “This report is published by SAE to advance the state of technical and engineering sciences. The use of this report is entirely voluntary, and its applicability and suitability for any particular use, including any patent infringement arising theref
2、rom, is the sole responsibility of the user.” SAE reviews each technical report at least every five years at which time it may be revised, reaffirmed, stabilized, or cancelled. SAE invites your written comments and suggestions.Copyright 2011 SAE International All rights reserved. No part of this pub
3、lication may be reproduced, stored in a retrieval system or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. TO PLACE A DOCUMENT ORDER: Tel: 877-606-7323 (inside USA and Canada) Tel: +1 724-776-4970
4、(outside USA) Fax: 724-776-0790 Email: CustomerServicesae.org SAE WEB ADDRESS: http:/www.sae.orgSAE values your input. To provide feedback on this Technical Report, please visit http:/www.sae.org/technical/standards/J2958_201111SURFACEVEHICLEINFORMATIONREPORTJ2958 NOV2011 Issued 2011-11Report on Unm
5、anned Ground Vehicle Reliability RATIONALEThis information report was written to provide an overview of the state of the art in unmanned ground vehicle reliability and serve as a basis for potential later recommended practice and/or standards documents. INTRODUCTIONThis report provides information o
6、n the state of the art in the status and practice of unmanned ground vehicle (UGV) reliability. It decomposes UGV reliability into the component level (power systems, manipulator/end-effector, and sensors) and the system level (communications and operator control unit) and devotes a section to each
7、of the individual elements in these levels. The following conclusions are drawn: (1) Tools for analyzing the component-level reliability of hardware components subject to physical failure are available from the reliability engineering literature, and have been applied to mobile robots in a few cases
8、. (2) However, a more systematic and carefully collected set of field data is required in order to validate these models and make more confident statements about the primary areas of weakness and potential reliability improvement for UGVs. (3) System-level reliability is less amenable than the compo
9、nent level to quantitative analysis. Though some work has been done in system-level analysis, much more is needed, and its possible unification with component-level analysis should be explored in order to achieve something closer to full-system reliability analysis. TABLE OF CONTENTS 1. SCOPE 21.1 P
10、urpose . 22. REFERENCES 22.1 Applicable Documents 23. UNMANNED GROUND VEHICLE RELIABILITY . 63.1 Overview . 63.2 Power System . 113.3 Power Train and Running Gear 143.4 Manipulator and End Effectors 223.5 Sensors . 303.6 Communication Systems 353.7 Operator-Robot Interface and Interaction . 41APPEND
11、IX A COMPLEMENTARY PERSPECTIVES ON ROBOT RELIABILITY 46SAE J2958 Issued NOV2011 Page 2 of 49 1. SCOPE This report is applicable to the reliability characteristics of unmanned ground vehicles. 1.1 Purpose The purpose of this information report is to review the state of the art in unmanned ground vehi
12、cle reliability and to identifyneeded areas of increased information, analysis, and improvement. 2. REFERENCES 2.1 Applicable Documents The following publications form a part of this specification to the extent specified herein. Unless otherwise indicated, the latest issue of SAE publications shall
13、apply. 1 Adams J.A., “Critical considerations for human-robot interface development,” 2002 AAAI Fall Symposium: Human Robot Interaction Technical Report FS-02-03, pp. 1-8, 2002. 2 Arther J., FMEA Template, QIMacros http:/ Bertollio F., Jensfelt P. and Christensen H.I., “SLAM using Visual Scan-Matchi
14、ng with Distinguishable 3D Points,” IROS 2006, pp. 4042-4046, October 2006.4 Blackburn M.R., Laird R.T. and Everett H.R, “Unmanned Ground Vehicle (UGV) Lessons Learned,” TechnicalReport 1869, SPAWAR Systems Center, San Diego, November 2001. http:/www.spawar.navy.mil/sti/publications/pubs/tr/1869/tr1
15、869.pdf5 Brown W.R. and Ulsoy A.G., “Improving the Electromechanical Reliability of Unmanned Ground Vehicles,” University of Michigan Ground Robotics Research Center (GRRC) Technical Report 2009-02, 24 April 2009. 6 Burke J.L., Murphy R.R., Rogers E., Lumelsky V.J. and Scholtz J., “Final report for
16、the DARPA/NSF interdisciplinary study on human-robot interaction,” Proc. IEEE Trans. On Systems, Man, and Cybernetics Part C; Applications and Reviews, 34(2): 103-112, 2004. 7 Burschka D. and Hager G., “V-GPS (SLAM): Vision-based Inertial System for Mobile Robots,” Proc. 2004 IEEE Intl Conf on Robot
17、ics and (2) both component- and system-level elements.3.1.2.1 Component Level Section 3.2 Power Systems. This Section points out two facts that highlight the need for additional work and effort in this area. First, there are many potential manufacturing and assembly errors leading to power system fa
18、ilure that can be eliminated or greatly reduced by well-designed and -implemented quality control, but the U.S. Army Test Operating Procedure for UGVs 72 does not address power systems and only cursorily mentions batteries and fuel. Second, although one study of 5,000 currently fielded UGVs in Iraq
19、and Afghanistan reports a 4 to 10 hour MTBF vs. a desired 96 hours, the relative contribution of power system reliability to this aggregate MTBF is not clear, and few data on specifically power system reliability of fielded UGVs are available. If the needed manufacturer and field reliability data ca
20、n be acquired and analyzed, then power systems is an area that lends itself to established reliability engineering analysis. Section 3.3 Power Train and Running Gear. This section summarizes the results from the relatively few studies on UGV reliability in several contexts including search-and-rescu
21、e robotics, museum robots, and the DARPA PerceptOR program. In all but one of the studies (CRASAR), the number of robots considered was too small to give confidence in the statistical significance of the results. In the CRASAR study, the most common mode of failure (36%) was the effector category, w
22、hich included thrown and slipped tracks, mobility mechanism failures, and sheared pins and gears, all of which belong under the “power train and running gear” heading. The section also summarizes the principal methods of quantitative reliability analysis and closes by observing that the relative pau
23、city of data on UGV power train and drive system reliability suggests the need for a much more purposeful and comprehensive data acquisition effort in this area. SAE J2958 Issued NOV2011 Page 9 of 49 Section 3.4 Manipulators and End-Effectors. This section provides a good example of the application
24、of component-level reliability analysis to determine the reliability of aggregate assemblies, in this case the manipulators and end-effectors that are used on the widely employed PackBot and Talon robots. MTBF values from typical components making up such assemblies are combined using fault trees, w
25、hich are a graphical method for capturing the serial or parallel combination of component reliabilities using the OR and AND gates of Boolean logic, to derive overall reliabilities (i.e., probability of continued operation) for a specified period of time. The authors note that the greatest challenge
26、 (at the component level) is the acquisition of accurate and current reliability data, rather than the analysis itself. Section 3.5 Sensors. This section describes both hardware and software reliability. It cites two causes for the hardware failure of computers, sensors, and other related electronic
27、 devices: destructive voltage variations and heat. Each of these can be ameliorated through improved design and modeled from a component-level reliability engineering standpoint. Electronic device components typically undergo rigorous testing, so component-level MTBF data, from which aggregate devic
28、e MTBF may be calculated should be available. The modeling of software reliability (e.g., for sensor interpretation and fusion algorithms), which belongs at the system level in the classification proposed here and is therefore further discussed below, is more difficult and less frequently attempted,
29、 though some studies exist 31. 3.1.2.2 System Level Section 3.6 Communications Systems. This section deals with multiple aspects of communications reliability: hardware, software settings, transmit power, environment, and network topology. Basic communications hardware is amenable to straightforward
30、 reliability engineering analysis, while overall communications reliability analysis is difficult due to the complexity of RF modeling in 3D environments and the dynamic nature of radio networks. Nevertheless, the section argues that statistical characterization based on simulation and testing can g
31、ive system performance bounds with quantifiable confidence levels. Section 3.7 Operator-Robot Interface and Interaction OCU. This section makes the following three main points. First, for all currently fielded UGVs, the human operator is part of the feedback loop representing the UGVs control, impor
32、tant components of which are communications and degree of autonomy. Second, key potential failure points are lack of operator situational awareness, response delays, and wrong/inappropriate operator responses. Third, human-in-the-loop reliability analysis is relatively immature, and may profit from
33、augmentation with techniques from the field of Cognitive Task Analytics 15. From the reliability standpoint, the presence of a human operator in the loop is a two-edged sword: on the one hand, the humans ability to react intelligently and resourcefully to an immense range of situations increases sys
34、tem reliability; on the other hand, that same ability to respond creatively makes the humans role difficult to quantify and link to the better-understood component-level analysis. APPENDIX A Complementary Perspectives on Robot Reliability. This appendix provides a different view and food for thought
35、 on achieving UGV reliability. 3.1.2.3 Software Software pervades the system level (controls, communications, human-machine interaction) and renders useful the component-level devices (power system, manipulator/end-effector, computers, sensors, mobility platform) whose hardware aspects can be straig
36、htforwardly modeled via standard reliability engineering methods. However, software reliability is different in nature from hardware reliability; the latter involves the effects of wear over time, while the former is compromised primarily by a combination of human design error with system complexity
37、. The appropriateness of using statistical methods to model software reliability without severe restrictions is unclear, though there is a literature on this topic which may be consulted 314474. Though this report does not include a separate section on software, future reports or updates to this rep
38、ort should consider the subject of software as an important cross-cutting system aspect with high potential impact on reliability. 3.1.3 Approaches to Improving UGV Reliability The individual sections mention various approaches to improving UGV reliability. These and some additional approaches are s
39、ummarized here under component-level and system-level headings. SAE J2958 Issued NOV2011 Page 10 of 49 3.1.3.1 Component Level Improved manufacturing processes. This refers to at least two distinct phenomena. On the one hand, improved manufacturing, particularly quality control, processes can reduce
40、 the number of components that suffer infant mortality. On the other hand, innovations in design, materials, etc. can result in producing higher-reliability components at lower cost. Economies of scale may play a beneficial role in this respect as the military has a demand for increasing numbers of
41、robots and the components from which they are assembled. Use higher-reliability components. All other things being equal, this increases cost, but it is an effective way to increase reliability if the money is available and one has sufficient data or insight to determine which components reliability
42、 increase will have the biggest impact on overall system reliability. Redundancy. Like using higher-reliability components, this increases cost, both design and component. If funds are available, the question of which approach is better, higher-reliability components or redundancy, can be answered u
43、sing standard reliability engineering techniques. Redundancy may also increase weight, volume, and power requirements, and it may reduce performance. Mission-based team design for reliability. Given a mission specification and the need to assemble a team of robots to perform it, how does one maximiz
44、e probability of mission completion given choices of number of robots, reliability level of components, cost of the mission, etc.? Techniques for this are described in 6869. 3.1.3.2 System Level Improved system engineering. The phrase “system engineering” can apply to many things, but the particular
45、 emphasis here, given the heavy human operator participation in currently fielded UGVs, is on the design of the human-UGV interface. Among other things, it should provide intuitive situational awareness and user input. Improved operator training. The need for operator training complements an intuiti
46、ve user interface by working on the other end of the human-machine interaction problem. Ideally, the profitable use of fielded UGV systems should be quickly learnable by users, with further training enhancing, rather than enabling, system usefulness. The rapidly increasing pace of UGV deployment off
47、ers an excellent opportunity for data collection and systematic improvement of UGV reliability. Carefully collected data from the field can be used to identify the components with highest reliability impact, refine reliability prediction models, and generate best human-machine interaction practices.
48、 These improvements can be fed into next-generation UGVs, and the resultant iterative cycle can significantly improve UGV reliability.3.1.4 Challenges Facing UGV Reliability Improvement and Analysis There are various challenges facing the systematic improvement of UGV reliability: Scarcity and peris
49、hability of manufacturer reliability data. Component-level reliability analysis depends for its accuracy on accurate and current reliability data on manufactured components and suffers to the extent that these data are unavailable or out of date. Dependence on loading and environmental conditions. Manufacturer MTBF
