This page intentionally left blank Principles of Embedded Networked Systems Design Embedded network systems (ENS) provide a set of technologies that can link the physical world to large scale networks in applications such as monitoring of borders, infrastructure, health, the environment, automated production, supply chains, homes, and places of business This book details the fundamentals for this interdisciplinary and fast-moving field The book begins with mathematical foundations and the relevant background topics in signal propagation, sensors, detection and estimation theory, and communications Key component technologies in ENS are discussed: synchronization and position localization, energy and data management, actuation, and node architecture Ethical, legal, and social implications are addressed The final chapter summarizes some of the lessons learned in producing multiple ENS generations A focus on fundamental principles together with extensive examples and problem sets make this text ideal for use in senior design and graduate courses in electrical engineering and computer science It will also appeal to engineers involved in the design of ENS G R E G O R Y P O T T I E has been a faculty member of the UCLA Electrical Engineering Department since 1991, serving in vice-chair roles from 1999 to 2003 Since 2003 he has served as Associate Dean for Research and Physical Resources of the Henry Samueli School of Engineering and Applied Science From 1997 to 1999 he was secretary to the board of governors for the IEEE Information Theory Society In 1998 he was named the faculty researcher of the year for the UCLA School of Engineering and Applied Science for his pioneering role in wireless sensor networks, and in 2005 was elected as a Fellow of the IEEE Professor Pottie is a deputy director of the NSF-sponsored science and technology Center for Embedded Networked Sensors and is a cofounder of Sensoria Corporation W I L L I A M K A I S E R joined the UCLA Electrical Engineering Department in 1994 and there, with Professor Pottie, initiated the first wireless networked microsensor programs with a vision of linking the Internet to the physical world through distributed monitoring Professor Kaiser served as Electrical Engineering Department Chairman from 1996 to 2000 He has received the Allied Signal Faculty Award, the Peter Mark Award of the Vacuum Society, the NASA Medal of Exceptional Scientific Achievement, the Arch Colwell Best Paper Award of the Society of Automotive Engineers and two R & D 100 Awards He is cofounder of Sensoria Corporation Principles of Embedded Networked Systems Design Gregory J Pottie and William J Kaiser University of California, Los Angeles Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo Cambridge University Press The Edinburgh Building, Cambridge , UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521840125 © Cambridge University Press 2005 This publication is in copyright Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press First published in print format 2005 - - ---- eBook (EBL) --- eBook (EBL) - - ---- hardback --- hardback Cambridge University Press has no responsibility for the persistence or accuracy of s for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate Dedication To Aldo, Cathy, Claire, and Laura for their patient love and support To the dedicated and creative students and colleagues of the LWIM, AWAIRS, and CENS programs at UCLA and to the Sensoria Corporation team Contents Preface page xiii Acknowledgments xv List ofAbbreviations xvi Embedded network systems 1.1 Introduction 1.2 ENS design heuristics 1.3 Remote monitoring 1.4 RFID 1.5 Enacted spaces 1.6 Historical context 1.7 Further reading 1 10 Representation of signals 2.1 Probability 2.2 Stochastic processes 2.3 Introduction to information theory 2.4 Summary 2.5 Further reading 2.6 Problems 12 12 17 24 30 30 31 Signal propagation 3.1 Basic wave propagation phenomena 3.2 Radio signals 3.3 Optical signals 3.4 Acoustic and seismic signals 3.5 Biochemical signals 3.6 Summary 36 36 44 50 52 56 57 vii viii Contents 3.7 3.8 Further reading Problems 57 58 Sensor principles 4.1 Sensor system ideal architecture 4.2 Sensor system non-ideal operation 4.3 Sensor system standard figures of merit 4.4 Environmental sensors 4.5 Motion and force sensors 4.6 Transducers for electromagnetic phenomena 4.7 Chemical and biochemical sensors 4.8 Electronic noise sources and noise reduction in sensor systems 4.9 Reducing sensor system errors by feedback control methods 4.10 Actuators for microsensor systems 4.11 Calibration 4.12 Summary 4.13 Further reading 4.14 Problems 61 61 63 64 72 77 88 93 96 100 103 104 105 106 107 Source detection and identification 5.1 Introduction to detection and estimation theory 5.2 Detection of signals in additive noise 5.3 Estimation of signals in additive noise 5.4 Hierarchical detection and identification systems 5.5 Summary 5.6 Further reading 5.7 Problems 109 109 111 127 139 146 147 148 Digital communications 6.1 Characterization of communication signals 6.2 Communication over the Gaussian channel 6.3 Synchronization 6.4 Communication over dispersive channels 6.5 Communication over dynamic channels 6.6 Summary 6.7 Further reading 6.8 Problems 152 152 155 162 165 176 186 187 187 Multiple source estimation and multiple access communications 7.1 Interference models 7.2 Source separation 7.3 Basic multiple access techniques 7.4 Multiple access in interference 7.5 Heterogeneous networks 7.6 Summary 195 195 198 201 207 218 219 16.3 ENS: information technology regulating the physical world 487 from many health clubs due to incidents of people taking pictures in changing rooms that were subsequently posted on the web; but cameras will get smaller still This raises the important question of whether mandates are needed that require an easily distinguished signal to be emitted by any camera so that people are warned when they are being monitored Example 16.12 Privacy and beacon chips With radios embedded throughout the environment, sensors may be commanded which information to report Sensors capable of providing high-resolution information at particular spatial locations (e.g., tracking a person) might be commanded not to continue tracking by the bearer of a chip which has proper authorization This might, e.g., prevent information being gathered for commercial purposes Alternatively, this information might not be collected unless the sensors detect a chip that indicates the bearer has authorized the sale of privacy (e.g., for a pay per view web-vision channel), or desires surveillance to continue for purposes of providing security or monitoring for medical reasons The chip might provide a positive signal for tracking, thus simplifying the tasks Discuss the implications of such systems How would information on bystanders within the sensor field be dealt with? How would prospective privacy rights for the individual be traded against the rights of owners of real property (e.g., shopping malls) in collecting information concerning persons who enter the premises (e.g., through networked security cameras and information systems)? Health-care systems may also be revolutionized by monitoring systems worn voluntarily by people A continual record of physiological response for individuals at high risk for particular ailments can be used for better diagnosis, automated alerts to ambulances when severe problems are detected, and adjustment of drug dosages More broadly, environmental factors that lead to disease could be better characterized, as the following example suggests Example 16.13 Pervasive environmental monitoring It is presently quite difficult to monitor the total exposures of individuals to various pollutants, and consequently to assess the combined risks resulting from exposures to combinations of different agents Improved chemical sensors will over time, however, enable widespread sampling at much lower cost This may enable far more detailed studies of the hazards in both urban and rural environments to the point where activist groups could both mount studies and monitor how well government agencies are doing their jobs Similar technology could enable citizens to test their municipal and bottled water, and to scan food for chemicals, pathogens, and unwanted genetic modifications Research what monitoring methods are used now for air, food, and water How likely is it that updated monitoring would find harmful substances in your food and water supply? Discuss how such monitoring would affect industry, farming, and your eating habits Security Today there are many systems that monitor public and private spaces The average person in a city is in view of many different security cameras over the course of a day in offices, stores, and traffic intersections These systems are so common that they have faded into the background of modern life Typically, the images collected are stored for a short period of time, and then erased The signal processing consists of people 488 Ethical, legal, and social implications of ENS looking at the tapes The information typically stays on the premises where it was collected, being retrieved only if some unusual incident is known to have occurred With ENS the monitoring systems can be networked, with increased possibilities for processing of images to track behaviors, determine identity, or link to other databases Companies could offer these specialized services off-site, and thus gather information from a broad set of clients, enabling profiles of many behaviors to be built up over time Therefore, instead of information being collected in one place that is largely ignored before being discarded, the data could be collected over a large environment, be processed in sophisticated ways, and persist Coordinated collection of information, even from the same number of sources, is thus completely different in its implications for privacy than first generation security camera systems Example 16.14 A robotic monitoring network Using articulated and mobile sensing systems can greatly increase the efficiency of monitoring systems Multiple cameras that can pan, tilt, and zoom can provide higher-resolution images from different angles of view, greatly increasing recognition accuracy With a system of wires between buildings or tracks on the ceilings of retail/office space, cameras and other sensors could quickly move through three-dimensional volumes, giving the same resolution that would be achieved with a larger number of cameras in a dense network This can also overcome the effect of obstructions that plague fixed installations Other more surreptitious monitoring will be possible, with tethered robotic elements that could snake through ducting systems to record images and sounds This could enable, e.g., a security retrofit, or be something that is used by law enforcement agencies Discuss the privacy and law enforcement implications of networked/robotic monitoring systems as compared to present day surveillance How would you feel about such a system being used to protect your home or place of work? Your local park? Science fiction and society Good science fiction allows the general public to imagine the consequences of particular technologies on individuals and societies, while also being entertained The next examples are about very unhappy societal outcomes for ENS technologies, inspired by popular science fiction Things start out well in the first example in each couplet, but then go horribly wrong in the second Example 16.15 Enacted ethical shopping Tag technology is on a path towards low-cost unique identification of every manufactured object of any significant size RFID tags are already being used to identify shipping pallets to reduce logistics costs, and can potentially lead to automated retail check-out Whatever the particular tag technology, consumers equipped with universal electronic scanners could link to databases which, in turn, could reveal price, country of origin, expiry date, and possibly a large body of information related to the entire chain of custody and the conditions of manufacture of the components (sweatshop labor?) Specially designed software agents could then enable consumers to pass by items that not meet ethics/price profiles In turn, merchants could assemble information on which items the consumer scans to build a profile and produce targeted pricing and advertising This can result in far greater efficiencies in retail, and thus lower prices This brings up several immediate issues First, if tags remain active after purchase, then a new means will have been established to identify individuals, since payment will link the item 16.3 ENS: information technology regulating the physical world 489 to an individual The consumer must therefore be able to turn off the tag in such a manner that no one else can reactivate it Second, merchants may attempt to control the flow of information into the store to prevent unfavorable information about products from reaching the consumer What regulations should be put in place to safeguard privacy in a situation where most consumer products have electronically readable tags? Suggest means by which a free flow of information to consumers can be ensured; will market forces be sufficient? Example 16.16 The perfect profile In the movie Minority Report, a future is posited in which it is possible to predict the future criminal actions of people with enough certainty to have them arrested and convicted Things go badly wrong for the hero of the film, with the events turning on his attempts to escape the authorities The particular profiling means is a trio of mutants who can visualize future violent events While serving an important dramatic purpose in the film, there were other clever visualizations of the ways our actions will be monitored for commercial purposes such as selling particular consumer products As a consequence of terrorist events, renewed interest has gone into devising systems to search databases to mine and interpret data that may predict association with violent groups Given a very extensive sensor network designed to suppress criminal activity and support commercial interests (e.g., the automated retail system of the previous example), cooperation of government and private industry, and a set of incentives to consumers to share more of their life willingly, it is not a stretch to imagine technological attempts to predict future behavior These predictions may act as ‘‘probable cause’’ to enable even deeper invasions of privacy of which we would be unaware, since sensors would be everywhere in our environment What are some of the advantages to society of better profiling of individuals? What are some of the dangers? How is technological monitoring of this sort different in character from what has gone on for all of the history of civilization in small towns? What safeguards should be included in systems designed to profile for security or commercial purposes? From creature comforts we now turn to the more compelling question of medical monitoring, and consider what technologies may be pursued to improve quality of life Example 16.17 Medical ENS Advances in micro- and nano-technology have already enabled implants that connect to nerves in the brain, and over time the sophistication of these implants will improve We are on the technological path towards artificial eyes and ears with a high level of functionality, ultimately with links to the nervous system Such technology will also eventually allow precise monitoring of drug/hormone levels in particular organs, allowing localized control of drug levels within particular targeted systems (e.g., for chemotherapy) and thus far fewer side effects This could revolutionize cancer treatment or control of chronic diseases such as diabetes In the shorter term, much more extensive monitoring may greatly improve diagnostics Compare medical diagnostic techniques at the middle of the twentieth century with those at its end Given the still low level of penetration of information technology into the practice of medicine, what changes might take place over the next 50 years? What consequences might such increased automation have on the system of health-care delivery and the quality of life of people? Example 16.18 The rise of the collective In the television series, Star Trek: Next Generation, and Star Trek: Voyager, the most deadly enemy of humanity in the twenty-fourth century is a species known as the Borg Borg 490 Ethical, legal, and social implications of ENS technology includes artificial implants that enable direct mind-to-mind communication, resulting in a collective intelligence The hive mind is coordinated by a Queen, with the drones assigned to duties in the interest of maximizing efficiency One of these efficiencies lies in assimilating other humanoid species into the Borg Collective (the persons, the contents of their minds, and their technologies) Unhappy the space ship that receives the standard greeting: ‘‘We are the Borg resistance is futile.’’ Naturally, humanity resists assimilation, defending individuality against the values of collective purpose These episodes were the most popular of the series; the Borg are very creepy, the space battles spectacular, and the moral issues very clear What is never explained is how the Borg arose in the first place – what kind of society would choose such a fate? Part of what makes the Borg so scary is that it could be a society very much like our own Industrial economies already include an extensive high-speed communication infrastructure in support of decision-making and gathering of information, assimilate immigrants, and transform the cultures they touch into versions of themselves We would not survive at anywhere near current population levels without this vast interaction of man and machine Societies must either adapt to industrialism or be broken; resistance in the end is futile Further, there is huge economic value in being able to share information and cooperate at a distance There is consequently substantial ongoing research in making more natural interfaces between information technology and humans, and obvious economic advantages to making these connections ever more intimate, and dire economic consequences to groups that not adopt the latest information technology There are compelling miltary applications that push progress in human/machine interactions, as well as the medical applications mentioned in the preceding example, and a wide variety of interactive entertainment forms would be enabled While we are yet a very long way from directly monitoring the conscious thoughts of people, there are clearly compelling reasons to pursue technologies that move in this direction Unfortunately, the fundamental attributes of Borg society would appear well before thoughts are actually shared (assuming that is even possible in the far future) With more intimate man/machine connections that are designed to enable more efficient cooperation, little that is private would actually be kept from the public sphere One’s thoughts may be one’s own, but whatever is done would be monitored, and the opinions or orders of others within a collective enterprise would cascade upon the senses One might not have much choice but, in Borg parlance, to ‘‘comply.’’ What features must be built into interfaces, information collection engines, and the like to preserve individuality as the possibilities for direct communication to the sensory inputs become available? At what lower level of connectivity of sensors, computers, and human beings Borg-like features emerge? At what level are such features potentially helpful/ harmful to democratic society? Does the historical pattern of the spread of industrial society really predict that the technologies that would facilitate a Borg-like society would also become widely adopted? (Argument by analogy has a sorry history in philosophy; details of difference matter.) What factors would push these technologies forward or resist them? Embedded responsibility It is certain that ubiquitous deployment of ENS will have momentous societal impact; experiences with the Internet give but a small foretaste of what will come It is likewise clear that the design choices and regulatory decisions of the leading industrial nations will shape the kinds of outcomes that will arise in conjunction with market forces Such important decisions should not be left entirely to the market: it favors economic efficiency over all other values, which is why governments regulate markets and 16.3 ENS: information technology regulating the physical world 491 (theoretically) not the other way around On the other hand, these technologies present great temptations to governments and other institutions to cut corners on individual rights The design principle that therefore emerges is to make the architecture selection as open a process as possible, so that broad societal input is achieved at the beginning, and as the technology evolves There are a large set of design alternatives that yield similar intended results but vastly different incidental results For example, while the conventional way to use a camera as a sensor is to record the complete image either in raw or compressed form, it must be noted that this is a design choice If the objective is to determine light levels within a scene, the data can be recorded in such a way that the identities of all individuals are obscured Even should an image be required (e.g., for identification of individuals passing through a portal) the system need not record bystanders and, in fact, greater scalability results if only the relevant or triggered events are recorded and communicated A general principle of recording the minimal information to accomplish a task thus both protects privacy and enhances scalability More generally, information systems through the structure of their code give the possibility of embedded responsibility That is, certain societal values can be embedded in the system architecture and code that make it easier to use devices in one way than others judged to be harmful Possible values include assumptions of the right to individual privacy, facilitation of social interactions, symmetry of information flows between observed and observer, and so forth There is some cost associated with embedding such values, but in developing technology that can have huge societal impact there is no ethical choice but to consider such factors An integral part of product development is societal acceptance As discussed in Chapter 15, including the end-users throughout the development cycle results in increased design efficiency and also intellectual challenges Ignorance of societal concerns in the course of a design process can lead to catastrophic results such as legislative action (or consumer boycott) to ban or restrict a product after a great many resources have been spent Openness on the other hand carries little risk for information technologies In the end, the public is going to be aware of the product and so their concerns will need to be dealt with In the earlier stages of development concerns can be accommodated with far less expense than in a subsequent retrofit Further, large systems already require diverse design teams and the marginal expense of gaining broader inputs is low Seeking such inputs through an open process does not guarantee that what emerges in the end will necessarily move us towards a situation where liberty, security, and prosperity are simultaneously maximized as ENS become pervasive However, open processes are much better at discovering flaws than closed ones People are generally very good at finding what is wrong with things they oppose even if it is much more difficult to come up with a constructive solution The purpose of divided governance of society is not to promote efficiency, but to prevent disaster Similarly, an open process may not guarantee success but it will halt in their tracks many misconceived ideas, both technical (within the design team) and in application (with broader discussion) It is the job of the design team to come up with constructive solutions that meet these societal constraints, and thus advance rather than degrade civilization 492 Ethical, legal, and social implications of ENS Example 16.19 A happy ending Ubiquitous ENS systems monitor our health, homes, workplaces, and the environment Manufacturing systems and the delivery of services are automated to unprecedented levels freeing human beings for creative pursuits Privacy is available as requested through automated agents, which can also summon assistance as required Laws and design have come together to enable gathering of intelligence, according to principles arrived at through informed public debate Will open standards, open code, and open debate really happen, and can they lead to such a future (and why all research grant proposals implicitly assume that a happy ending is assured)? What hard barriers lie in the path to such a future, both technical and political? What are the responsibilities of engineers and other researchers on the one hand, and funding agencies on the other for seeing that these barriers are overcome? 16.4 Summary In this chapter the basic regulatory structure for technology in industrial liberal democracies has been described The manner in which design shapes societal outcome was discussed in Section 16.2 in the context of the Internet It was observed that the basic desire of the designers (who were also the first users) for openness was embedded in the design, with subsequent efforts to change that structure being costly although possible Section 16.3 presented a variety of examples of the ways embedded networked systems may change our society, and suggested that for reasons of design efficiency and to guard against unintended outcomes it is best to proceed with the endusers (the general public) engaged in an open process That is, the responsible course is to embed societal values as design constraints 16.5 Further reading The question of efficiency in government has a long historical record Aristotle categorized absolute monarchy, oligarchy, and democracy in terms of their efficiency when led well and when led badly Democracy was observed to be the least efficient but also the least bad in its degenerate form of mob rule, and thus on the whole preferred to the others, (clearly, an early example of minimax optimization) The formulation of the separation of the powers of government owes much to the essay of John Locke, Second Treatise of Civil Government, first published in 1689 to justify the Glorious Revolution (effectively, the ascendancy of Parliament in a constitutional monarchy) Montesquieu subsequently articulated the need for an independent judiciary The essay also posits basic human rights, and as such is echoed in the Bill of Rights (1689, England), the Declaration of the Rights of Man (1789, France), the Declaration of Independence (1776, the United States), and the Bill of Rights (1791, the first ten amendments to the Constitution of the USA) These documents collectively embody many of the principles at the heart of liberal democracy One can easily find them on the web A broad discussion of the Enlightenment, including both the evolution of ideas of societal organization and the political/economic background is found in 16.5 Further reading 493 W Durant and A Durant, The Story of Civilization, Vols VII–X Simon and Shuster, 1954 The governance implications of 51 of the largest 100 economies in the world being corporations rather than nation-states are explored in N Hertz, The Silent Takeover: Global Capitalism and the Death of Democracy Harper Business, 2001 Section 16.2 is largely based on the highly readable book Laurence Lessig, Code and Other Laws of Cyberspace Basic Books, 1999 Detailed discussion is given on much of the material presented as exercises Examples of specific social interactions (including political ones) enabled by portable communications technology are described in H Rhiengold, Smart Mobs: The Next Social Revolution Perseus Publishing, 2002 We anxiously await the study on ELSI implications of ENS recommended in D Estrin, Embedded Everywhere: A Research Agenda for Networked Systems of Embedded Computers National Academy Press, 2001 A new National Research Council study is due in 2005 Much of Section 16.3 has relied upon interesting transdisciplinary discussions held at UCLA among faculty Some of these are memorialized in the website of the Institute of Pervasive Computing and Society www.ipercs.ucla.edu A discussion of RFID technology and its consequences from several points-of-view can be found on the web site of the California senate subcommittee on new technologies One of the hearings in the series is documented at www.senate.ca.gov/ftp/SEN/COMMITTEE/STANDING/ENERGY/_home/ 08-1803agenda.htm The focus of this chapter has been on the societal consequences of a particular set of technologies, rather than on ethical behavior in a professional engineering setting For a discussion of such issues, see, e.g., C.B Fleddermann, Engineering Ethics Prentice Hall, 1999 Chapter 17 Design principles for ENS This brief chapter summarizes a number of the design themes that are threaded through the book, expanding upon the design heuristics introduced in Chapter The section headings denote the basic principles 17.1 The physical world may not be abstracted away While philosophers have doubted the existence of an objective reality, engineers will not get very far in the development of embedded systems without at least modeling the physical world as such The purpose of an embedded system is to gather some information about the physical world and then take some action, if only to report that information to some other entity As such an understanding of the physical process being monitored or controlled is essential Chapter provided some of the mathematical tools used in modeling such processes, while Chapter discussed some of the basic propagation laws for natural and man-made signals There are fundamental limits to how well it is possible to observe phenomena with a given set of sensors, due to the presence of noise, obstructions, and propagation losses There are similarly fundamental limits on communications capacity, while for any real apparatus computational, storage, and energy resources are also limited System design must respect these limits and explore how resources can best be added to meet performance requirements while living within cost constraints There are occasions when questions such as ‘‘what would happen if resource x were free?’’ are appropriate (e.g., in imagining some limiting case as a useful approximation) and others where they are not (e.g., when it comes to implementing the practical system) Thus a useful heuristic in some situations is that communications costs are much more than signal processing costs, and therefore one should design to minimize communications However, it does not follow that signal processing will in the end consume less energy in the final design than communications; it is more likely that the energy minimum will occur with the same order of magnitude of energy spent on both A set of physical devices will have some particular costs associated with these basic operations, whatever the long-term technology trends 494 17.3 Hierarchy is usually unavoidable 495 As discussed in Chapter 15, the level of abstraction with which reality is approached depends in part on what part of the design cycle is being worked on A lesser level of detail is needed at some times, and indeed reality is modeled by a set of abstractions of varying levels of verisimilitude This enables analysis and testing of which factors are most important in the design In the end, however, a system will not be fully trusted until it is actually tested in the field, since reality provides the ultimate model of itself (at least in the minds of engineers) Experimentation not only verifies system performance but also advances theory 17.2 Play the probability game Much of academic training leaves the impression that the highest goal is to find an equation that exactly describes what is going on This is misguided on several fronts as far as embedded systems are concerned All measurements are subject to uncertainty such as noise, and in digital systems the A/D conversion introduces quantization error and other distortions from filtering and under-sampling There is thus no such thing as an exact representation of reality inside a digital computer; all engineering solutions are approximations The objective of design is not to provide an exact solution, but rather a solution that meets the goals in some probabilistic and approximate sense Just as natural processes are usefully represented as random processes, so may be the operations of large-scale systems A focus upon the events or features which are most likely to have an impact on performance is a basic component of the design process Various classical criteria for making decisions were presented in Chapter Bayes optimization, with probabilities weighted by some cost or revenue function, is perhaps the most frequently invoked design principle even if our knowledge of the probabilities or the true costs is often approximate It comes down to a focus on the essential: most design effort should be expended on those elements that will most affect the outcome over the range of possible scenarios The complete list of factors that affect performance might be infinitely large, but an approach that focuses on a few of the most important ones can still be extremely effective The art consists of finding which ones should be considered and how Intimate knowledge of the underlying processes and prior successful engineering approaches in similar situations is an essential guidepost for progress 17.3 Hierarchy is usually unavoidable The preceding chapters have described numerous instances of hierarchy: processing within nodes, multiple classes of sensors within networks, multiple classes of radios, processors, and storage elements, logical hierarchies, and multiple layers of abstraction in the various phases of the design process The last instance provides a clue as to the utility of hierarchy in general Large problems are difficult to solve using a single set of abstractions, and devices that are supposed to many things usually none of them as well as devices designed for particular purposes Therefore large systems are composed of hardware and software components that are matched to problems admitting some level of abstraction that interface to components matched to the 496 Design principles for ENS problems at some different level of abstraction A specialized integrated circuit can, e.g., be more efficient in energy than a general-purpose processor implementing the same function, but it is then limited to a particular model of the physical world and limited set of applications A system that includes a variety of specialized devices at one layer and an abstraction that allows fusion or decision-making among the outputs of these devices can yield both flexibility and high performance Moreover, when provided with well-defined interfaces a large team can work on optimizing the layers and devices independently with a greatly simplified task in debugging the overall system How the layers are constructed is intimately connected with the probability that particular events occur Specialized devices may be constructed for the events/functions that are expected to have the highest resource burden, with more general-purpose or higher-performance devices used for the rest, which while not probable may be large in number Hierarchy is a natural way to play the probability game Scalability is also enhanced with hierarchy It is easier to design a layered network than one in which every node must potentially take on all possible functions, as these typically multiply with the number of nodes Heterogeneity should therefore be embraced both within nodes and in the network as a whole This does not mean that it is not a worthwhile goal to expand the capabilities of components or nodes used at each of the levels of the hierarchy However, the broader setting of inherent heterogeneity should be kept in mind 17.4 Innovate only as much as necessary Engineering requires individual creativity, but at its heart is a collective process Conventional designs generally exist for sound reasons, and standards exist because a community has found them to be useful for a variety of applications They are seldom the best solution to a given problem, but often good solutions to a variety of problems A huge amount of design effort has typically gone into them, with most of it dealing with nasty bugs that came up during the validation process that were seldom apparent in the early design stages (Put another way, respect your engineering elders It is very hard to get anything new to work.) Thus, an important resource consideration at the beginning of a project is how much design time would be consumed in doing a custom design for some component in place of using a commercially available device or software suite (including development tools) When searching through manuals of available electrical components or searching for software it is very often surprising how much is available Small redesigns of a system in favor of the use of commercially available software and parts can result in large cost and time savings Here the Rule of Twos comes in handy If only one major function is to be supported, a custom design that crashes through layers can result in large efficiencies in embedded systems If, however, two or more different applications are to be supported, then caution is in order First, the efficiency gain compared with conventional design is likely to be far less than is possible for a single application Second, once it is decided that two applications must be supported, it becomes very likely that three, four, or more applications will eventually be decided to be necessary Magic numbers of the universe are zero (does not happen), one (unique), and infinity 17.5 Scale matters 497 (pervasive) There being exactly two objects, incidents, or applications of one kind is unnatural Another version of the Rule of Twos applies to optimizations Probability distributions are usefully characterized as a body and a tail One set of techniques will subdue the main part If the outliers in the tails are not important, a simple custom design will the job If the tail part matters, a second and far more complicated set of techniques will generally be required for the incremental performance gains These may be poorly modeled, and thus difficult to capture in a fully custom design or in a short time If a solution already exists, it is very tempting to use it Custom design should generally be focused only upon that part of the problem most critical for performance Beyond that, it may turn out to be intellectually interesting but will result in a very high cost of development in terms of time and resources Now what is critical to performance depends upon the technology generation The number of features that become heavily optimized increases in each generation, as older features become part of the baseline design Thus, innovation is constantly and obviously required, and results in revolutionary changes over the multiple generations While it is good to be on the look-out for disruptive technologies that can lead to a whole different path of development (and run with them if such are found), such episodes are rare, and must in any case be pursued with a similar eye to what can be reused to make the development cycle as fast as possible 17.5 Scale matters Scale is one of the most critical issues in embedded systems development It is still possible to sell personal computers or cell-phones that demand some degree of configuration by the user, including dealing with recharging It is not reasonable to expect people to expend similar effort when there are tens or more of such devices per person Scale demands autonomy, including network self-organization, calibration, automatic download of new software, and long-term energy management Logistical items such as replacement of batteries that are merely annoying for 10 nodes become absolute barriers to deployment at 1000 nodes Centralized methods that work well on the scale of 10–20 nodes fail for larger numbers, owing to the impossibly large number of combinations Most decision-making of necessity becomes local However, the network must also meet global goals So far the only practical way to this has been to create a hierarchy in which the lower layers deal with abstractions closely connected to their physical situations, and upper layers have abstractions that more and more resemble symbolic reasoning The whole translates the desires of the user(s) into actions As observed in Chapter 8, for a system to be scalable given finite resources most of the action takes place at the lowest layers in local neighborhoods, and thus great attention needs to be paid to optimizations here This allows more complicated solutions to be applied to the less frequent events that rise beyond the capabilities of local control Moore’s Law and Gene’s Law together dictate that what happens at these local levels will grow increasingly sophisticated over time: more advanced collaborations, higher levels of reasoning and decision-making, support for higher-end sensors, and 498 Design principles for ENS greater communications capabilities This reduces the need or desire for tight centralized control even while allowing its reach to possibly expand The net effect is increased intelligence embedded in the physical world As noted in Chapter 16, the social implications are almost surely momentous There is a vast difference between a few expensive and unconnected video monitors and web cameras in every room Large-scale deployment demands urgent consideration of ethical questions that would be irrelevant for deployments of small numbers of the same basic technology Once such deployments become plausible, the end-users (the larger society) must become involved in the dialogue that leads to the creation of the design constraints This is work for the designers and other stake-holders in the uses of the technology but in the end results in a more interesting design process and a greater likelihood of widespread acceptance of the technology 17.6 Teamwork In Chapter 15 explicit attention was given to the usefulness of having end-users (such as scientists) or their proxies (such as a marketing department on behalf of consumers) involved in the formulation of objectives through the different phases of a design cycle While this partnership is helpful for the efficient development of essentially any product, for embedded systems the cost of large-scale deployments and the huge range of possible design choices make it a near-necessity The large number of possible design choices is a great part of the attraction of building new ENS, but also demands some discipline in the design team not to constantly revisit technical decisions It is often possible to better, but there must be trust of other team members’ technical judgments that the alternatives present speculative gains not likely to be achievable with the time/resources available This does not mean that there should not be heated technical arguments, nor that an issue which was settled erroneously should not be acknowledged and rectified This is simply a caution not to give in to the siren song of perfection at the expense of the unglamorous goal of meeting requirements with adequate margin Teamwork can be extremely rewarding when all contribute in an atmosphere of trust (earned through accomplishments), with the group achieving far more collectively than the individual efforts would seem to imply It is the very essence of engineering Appendix A Gaussian Q function For a zero-mean Gaussian distribution PðX > yx Þ ¼ QðyÞ ¼ Z1 y pffiffiffiffiffiffi eÀz =2 dz; 2p an upper bound is QðyÞ pffiffiffiffiffiffi eÀy =2 : y 2p The bound is tight for y > The Q function is related to the complementary error function by erfcðyÞ ¼ pffiffiffi p Z1 eÀz =2 pffiffiffi dz ¼ 2Qð 2yÞ; y > 0: y Some values of the Q function are given in Table A.1 Note Q(0) = 0.5 499 Appendix A Gaussian Q function 500 Table A.1 Values of the Q function y 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 Q(y) 0.4801 0.4602 0.4405 0.4207 0.4013 0.3821 0.3632 0.3446 0.3264 0.3085 0.2912 0.2743 0.2578 0.2420 0.2266 0.2119 0.1977 0.1841 0.1711 0.1587 y 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 Q(y) 0.1469 0.1357 0.1251 0.1151 0.1056 0.0968 0.0885 0.0808 0.0735 0.0668 0.0606 0.0548 0.0495 0.0446 0.0401 0.0359 0.0322 0.0287 0.0256 0.0228 y 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 Q(y) 0.0179 0.0139 0.0107 0.0082 0.0062 0.0047 0.0035 0.0026 0.0019 0.0013 0.0010 0.00069 0.00048 0.00034 0.00023 0.00016 0.00010 0.00007 0.00005 0.00003 Q(y) À1 10 10À2 10À3 10À4 10À5 10À6 y 1.28 2.33 3.10 3.70 4.27 4.78 Appendix B Optimization Optimization problems arise in many different applications They include the following elements: a mathematical model that describes the problem of interest over some set of variables This may be discrete or continuous; a cost or revenue function of these variables that must be optimized according to some measure or norm; a set of constraints on the variables that defines their allowed range For example, the problem might be to determine the position of a source observed by several sensors The model may include a stochastic description of the sources, noise, and propagation conditions The optimization may be cast as a least squares problem, in which the expected variance of the position estimate is minimized The constraints may include involvement of some maximum number of sensors or some maximum number of bits exchanged among the sensor nodes to conserve energy Optimization is a very broad and deep subject In this appendix, a brief exposition of the basic tools of numerical analysis is presented, followed by a characterization of some classes of optimization problems and an outline of some classic approaches B.1 Basic tools of numerical analysis A basic fact of numerical methods is that linear problems are much easier to solve than non-linear ones Consider, e.g., the problem of finding the roots (zeros) of the equation f(x) ¼ Now if the function were a line one could readily compute the point of intersection with the x-axis Otherwise, the problem is typically approached by linearizing it and proceeding in a sequence of iterations For example, Newton’s method begins by guessing the root as x0, and forming (using Taylor’s Theorem) the linear approximation at x0: lðx; x0 Þ ¼ fðx0 Þ þ f ðx0 Þðx À x0 Þ; (B:1) where f (x) denotes the first derivative The unique root to the linear equation l(x;x0) ¼ is, assuming the derivative is non-zero at x0, 501 [...]... compositions of embedded systems that may function as the connection of the Internet to the physical world Embedded network systems (ENS) are poised to become pervasive in the environment with the potential for far-reaching societal changes that have hitherto been the subject of science fiction It is the purpose of this book to lay out the foundations of this technology, the emerging design principles. .. Agenda for Networked Systems of Embedded Computers National Academy Press, 2001 It contains a survey of technology, potential applications, and future directions One of the recommendations is a study of the ethical, legal, and social implications of ENS Some of the sensor deployments at the James San Jacinto Mountains Reserve can be monitored on-line at: www.jamesreserve.edu The interplay of ideas in... creation of dense networks of larger nodes that can be used to learn more about the types of networking, sensing, and signal processing that will be needed in future systems It is apparent even from the two application examples that a broad range of skills is required to design systems that network the physical world Later chapters of this book will therefore provide introductory treatments of topics... activated to take a picture of the source location so that a positive identification can be made This hierarchy of signal processing and communications can be orders of magnitude more efficient in terms of energy and bandwidth than sending images of the entire region to the gateway Further, the interaction of diverse types of nodes can more simply lead to automation of most of the monitoring work, with... behooves the designer of large-scale systems to consider both Homogeneity is in any case impractical in long-lived systems composed of integrated circuit components; as for the Internet, the architecture must accommodate the addition of successive generations of more powerful components 1.3 Remote monitoring To make the discussion more concrete, consider an application requiring identification of particular... Appendix A consists of tables of the Gaussian Q function, while Appendix B provides an introduction to formal optimization techniques Clearly not all of these issues can be covered in a single quarter or semester, and in a design team not all of this expertise need reside in every individual However, in total this is the set of topics that must be mastered in order to design efficient systems It is expected... Discussions at the NSF Center for Embedded Networked Sensing, under the leadership of Professor Deborah Estrin, have been a constant source of inspiration The patience of the editorial team at Cambridge University Press for our perpetually pushed-back deadlines is also appreciated Finally, we also thank our respective spouses, Aldo Cos and Cathy Kaiser for their cheerful tolerance of our many weekend and evening... vast majority of processors now being manufactured are used in embedded applications (i.e., having connection to physical processes) rather than in what would ordinarily be thought of as a computer Many are networked within the confines of a local control system, typically in master/slave configurations However, advances in wireless technology and in the understanding of distributed systems are now... desirable from the point of view of software development initially to provide a platform with considerable flexibility Example 1.1 Evolution of a habitat monitoring system An overview of one set of sensor deployments in the James Reserve in the San Jacinto mountains near Palm Springs CA is provided in Figure 1.3 There are very large elevation changes and consequently a broad set of species, some endangered,... complete system consists of a set of tags, one or more interrogators, and a backend data management system Even with purely passive tags this system is a sensor network, with the interrogators exciting responses and receiving as information the identity of the tags in range Unlike a system of bar codes and readers, RFID systems provide the capability of identifying not only the category of an item but also ...This page intentionally left blank Principles of Embedded Networked Systems Design Embedded network systems (ENS) provide a set of technologies that can link the physical world... two R & D 100 Awards He is cofounder of Sensoria Corporation Principles of Embedded Networked Systems Design Gregory J Pottie and William J Kaiser University of California, Los Angeles ... distributed systems are now making possible far more elaborate compositions of embedded systems that may function as the connection of the Internet to the physical world Embedded network systems