ADVANCED TOPICS IN SCIENCE AND TECHNOLOGY IN CHINA ADVANCED TOPICS IN SCIENCE AND TECHNOLOGY IN CHINA Zhejiang University is one of the leading universities in China In Advanced Topics in Science and Technology in China, Zhejiang University Press and Springer jointly publish monographs by Chinese scholars and professors, as well as invited authors and editors from abroad who are outstanding e>q)erts and scholars in their fields This series will be of interest to researchers, lecturers, and graduate students alike Advanced Topics in Science and Technology in China aims to present the latest and most cutting-edge theories, techniques, and methodologies in various research areas in China It covers all disciplines in the fields of natural science and technology, including but not limited to, computer science, materials science, life sciences, engineering, environmental sciences, mathematics, and physics Feng Xia Youxian Sun Control and Scheduling Codesign Flexible Resource Management in Real-Time Control Systems With 118 figures ' ZHEJIANG UNIVERSITY PRESS springer AUTHORS: Dr Feng Xia State Key Laboratory of Industrial Control Technology Zhejiang University 310027, Hangzhou, China E-mail: f.xia@ieee.org Prof Youxian Sun State Key Laboratory of Industrial Control Technology Zhejiang University 310027, Hangzhou, China E-mail: yxsun@iipc.zju.edu.cn ISBN 978-7-308-05765-3 Zhejiang University Press, Hangzhou ISBN 978-3-540-78254-4 Springer Berlin Heidelberg New York e-ISBN 978-3-540-78255-1 Springer Berlin Heidelberg New York Series ISSN 1995-6819 Advanced topics in science and technology in China Series e-ISSN 1995-6827 Advanced topics in science and technology in China Library of Congress Control Number: 2008925538 This work is subject to copyri^t All ri^ts are reserved, whether the whole or p art of the material is concerned, specifically the ri^ts of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfihn or in any other way, and storage in data banks Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyri^t Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag Violations are liable to prosecution under the German Copyri^t Law © 2008 Zhejiang University Press, Hangzhou and Springer -Verlag GmbH Berlin Heidelberg Co-published by Zhejiang University Press, Han^hou and Springer -Verlag GmbH Berlin Heidelberg Springer is a part of Springer Science + Business Media springer.com The use of general descriptive names, registered names, trademarks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use Cover design: Joe Piliero, Springer Science + Business Media LLC, New York Printed on acid-free paper Preface Recent evolutionary advances in information and communication technologies give rise to a new environment for Real-Time Control Systems This is a new dynamic environment that features both resource limitation and workload variability As a consequence, the availability of the computing and/or communication resources becomes typically uncertain in modem Real-Time Control Systems In this context, the espQCtQd Quality of Control (QoC) of the systems cannot always be guaranteed by the traditional control systems design methodology that separates control from scheduling From a resource scheduling perspective, the prevalent open-loop scheduling schemes in real-time systems obviously lack flexibility when applied to Real-Time Control Systems operating in dynamic environments To make the best use of available resources, more holistic principles and methods need to be developed These requirements motivate the recent technological trend towards the convergence of computing, communication and control This book is a monograph that covers our recent and original results in this direction The main objectives of this work are: (1) To construct a unified framework of feedback scheduling that enables the integration of control with computing and communication This framework will encompass a set of concrete feedback scheduling methods and algorithms that are applicable to different systems With these methods and algorithms, solutions are provided for some key issues in feedback scheduling, thus promoting the emergence of this area {2)To enable flexible QoC management in dynamic environments with uncertainty in resource availability A number of new approaches to flexible management of the computing and/or communication resources in Real-Time Control Systems will be developed to maximize or improve the overall system performance With these objectives in mind, we focus on feedback scheduling strategies for flexible resource management in the context of real-time control The traditional control systems design methodology and the simple control task model based on the VI fixed timing constraints are discarded, closed loop dynamic resource managpment schemes are built by means of control and scheduling codesiga The major tool used in this book is feedback scheduling The introduction of feedback into dynamic resource management breaks the traditional open-loop mode of resource scheduling We are not interested in solutions that belong to any particular discipline, i.e control theory, computer science or communication technology Accordingly, we not attempt to: (l)Design new control algorithms No innovation is made in this book regarding controller design that could make the control loops robust against delay, data loss or jitter This is often difficult because it requires a solid foundation of mathematics Furthermore, there has been extensive interest in this direction for many years, with an abundance of theoretical results produced (2)Develop new real-time task scheduling policies Inplementing a new scheduling policy demands support from the underlying platforms, e.g the operating systems Therefore, it is generally hard for a new scheduling policy to become practical even if it indeed performs better than existing ones in some situations On the other hand, many mature scheduling policies are available in the area of real-time scheduling (3)Develop new communication protocols Despite rapid evolution, none of the wireless technologies in existence was designed particularly for control applications Intuitively, developing such a dedicated communication protocol (from scratch or based on some existing protocols) could better support wireless control However, this is beyond the scope of this book due to its emphasis on networking While the framework presented is applicable to a broad variety of computing and communication resources, special attention of this book is paid to three representative classes of resources, i.e CPU time, energy and network bandwidth By e^loiting the feedback scheduling methodology, we develop a set of resource management schemes Numerous examples and case studies are given to illustrate the applicability of these schemes The inherent multidisciplinary property of the codesign framework makes the intended audience for this book quite broad The first audience consists of researchers interested in the integration of computer science, communication technology and control theory This book presents a unified framework as an enabling technology for codesign of computing, communication and control Novel paradigms for Real-Time Control Systems research and development in the new technological environment also provide insist into new research directions in this emerging area The second audience is practitioners in control systems engineering as well as computer and communication engineering Careful balance between theoretical foundation and real-world applicability makes the book a useful reference not only for academic research and study but also for engineering practice Much effort has been devoted to make this book practical For instance, the problems addressed are of remarkably practical importance; all solutions are developed with the practicability in mind This book is also of value to graduate students in related fields, for whom the tutorial introduction to feedback scheduling and the extensive references to the literature will be particularly interesting The background of the reader assumed in the presentation encompasses a basic knowledge in feedback control theory, sampleddata control, and real-time systems Prior e^erience with intelligent control, poweraware computing, and network bandwidth allocation is also helpful, thou^ not necessary Outline of the Book This book is broken into four parts, each part containing two chapters The first part Background, covers Chapters and 2, in which the motivations, background information and basic framework for this work are given Part II (CPU Scheduling) is concerned with scheduling the shared processor in multitasking embedded control systems Chapters and belong to this part Their focuses are on developing intelligent feedback scheduling methods by e?q)loiting neural networks and fuzzy control respectively While CPU scheduling is also involved, the main concern of Part IE (Energy Management) that covers Chapters and is dynamic management of the energy consumption of embedded controllers The gpneral goal of this part is to reduce the CPU energy consumption while preserving QoC guarantees The last part Bandwidth Allocation, which covers Chapters and 8, studies how to dynamically allocate available communication resources among multiple loops in networked control systems and wireless control systems respectively In Chapter we give an overview of the field of control and scheduling codesigQ The motivations for codesigQ of control and scheduling are illustrated Recent trend towards the convergence of computing, communication and control is pointed out After related work in the literature is reviewed, a perspective on feedback scheduling of Real-Time Control Systems is given Chapter presents a tutorial introduction to feedback scheduling Background knowledgp about sampled-data control and scheduling in real-time systems is outlined, which makes this book more selfcontained The temporal attributes of control loops are described Motivating examples for applying feedback scheduling are presented Key concepts, basic framework, and design considerations related to feedback scheduling are also described Chapter concerns neural feedback scheduling The primary goal is to overcome the disadvantages of overly large computational overhead associated with mathematical optimization routines We present a fast feedback scheduling scheme to e}q)loit feedforward neural networks This scheme can dramatically reduce feedback scheduling overhead while delivering almost optimal overall control performance VI D Chapter presents a fuzzy feedback scheduling scheme based on fuzzy logic control This scheme is independent of task execution times, robust to measurement noise, while handling uncertainty in resource availability in an intelligent fashion Being easy to implement, the fuzzy feedback scheduler also incorporates quite a small runtime overhead Considering the unpredictability of task execution times as well as the variability of CPU workload, Chapter develops a feedback control real-time scheduling methodology called Energy-Aware Feedback Scheduling It integrates the management of both energy consumption and QoC After analytically modelling the Dynamic Voltage Scaling system, a control-theoretic design and analysis method for feedback schedulers is proposed Taking advantage of the basic framework of Energy-Aware Feedback Scheduling Chapter aims to achieve further energy consumption reduction while not jeopardizing the quality of control For this purpose, we present an Enhanced Energy-Aware Feedback Scheduling scheme to e?q)loit the methodology of graceful gradation Chapter pays attention to multi-loop networked control systems By e}q)loiting codesign of control and network scheduling, we develop an integrated feedback scheduling scheme This scheme can maximize resource utilization in the case of light workload and achieve graceful performance degradation under overload conditions To attack the uncertainty in available communication resource in wireless control systems Chapter presents a cross-layer Adaptive Feedback Scheduling scheme that takes advantage of cross-layer design It proves quite efficient to deal with channel capacity variations and noise interference We also suggest an eventdriven invocation mechanism for feedback schedulers to further improve the feedback scheduling performance We consider only linear time-invariant (LTI) systems in this book, thou^ the proposed approaches are not only applicable to this class of systems The control applications used as simulation examples are kept typical and illustrative to exhibit the wide applicability of the proposed approaches For instance, the controlled processes studied include inverted pendulum, DC motor, and many others The control algorithms cover PID (Proportional-Integral-Derivative), state feedback control with pole placement, LQG (Linear Quadratic Gaussian) controllers, etc The control theory involved comprises classical control, modem control, as well as intelligent control In addition, both of the two design methods for digital controllers i.e discrete-time design and discretization of continuous-time controllers have been adopted in different simulations Throu^out the book, all simulations are conducted using Matlab/Simulink^ along with the TrueTime^ toolbox Matlab and Simulink are registered trademarks of The MathWorks, Inc http ://www.control.lth.se/truetime/ K Acknowledgements This book summarizes our research results achieved in recent years Many people contributed to this book in various ways We would especially like to thank YuChu Tian at Queensland University of Technology, Moses O Tade at Curtin University of Technology, Chen Peng at Nanjing Normal University, Wenhong Zhao at Zhejiang University of Technology, Jinxiang Dong, Zhi Wang and Xiaohua Dai at Zhejiang University, Liping Liu at Tianjin University, for their collaboration We are grateful to Zon^ai Sun at South China University of Technology, Xiaodong Wang, Jianhui Zhang and Peng Cheng at Zhejiang University who reviewed earlier drafts Special thanks go to Li Yu at Zhejiang University of Technology and Russell Morgan at Imprimis Computers, Brisbane, who read the manuscript line by line and gave particularly helpful suggestions for improvements, as well as Qing Lin at Zhejiang University, who helped set up wonderful working environments We gratefully acknowledge all support from Springer and Zhejiang University Press We dedicate this book to our families Feng Xia Youxian Sun 232 PART IV BANDWIDTH ALLOCATION deadline miss ratio at a desired level Intuitively, when the deadline miss ratio is in or close to steady states, there is no need for executing the feedback scheduling algorithm On the contrary, if the deadline miss ratio has significantly deviated from the desired level, then it becomes mandatory to run the feedback scheduler to adjust system parameters In this chapter the following condition is used for issuing the execution-request event: \p(k) - p j ^ (8.9) According to Eq.(8.9), the feedback scheduling algorithm will be executed if and only if the absolute difference between the actual deadline miss ratio and its desired level is no less than a specific threshold In this way, the disadvantages with timetriggered invocation with respect to response speed and overhead are avoided Furthermore, the negative effect of measurement noises on the deadline miss ratio may naturally be reduced There are two important parameters, 7"^^ and 5, to be set Generally speaking, choosing these parameters demands careful tradeoffs between quick response and low overhead Fortunately, this is not as difficult as usually e?q)ected Thanks to the small amount of computations associated with Eq.(8.9), it is possible to assign a small enou^ period r^p to the event detector to achieve quick response while keeping the feedback scheduling overhead small The magnitude of measurement noises should be taken into account when choosing the value of The system usually allows the deadline miss ratio to fluctuate around the setpoint with small deviations Therefore, a value that is sli^tly bigger than the magnitude of measurement noises could be used to reduce the number of executions of the feedback scheduler, which reduces runtime overheads 8.5 Simulation Experiments This section conducts simulation e>q)eriments on the case given in Section 8.2 to assess the performance of the proposed event-triggered cross-layer Adaptive Feedback Scheduling scheme Althou^ the results are still preliminary at this stage, some potential advantages of the proposed approach will be demonstrated Consider the following two scenarios: (1) Scenario I: the controller and the process are close to each other, WLAN operates at 11 Mbps, there is no interfering signals, =0.02, Ao =0.02; ^ (2) Scenario II: the transmission rate drops to 5.5 Mbps, the interfering transmitter sends a data packet of KB to the corresponding receiver every 10 ms, = 0.04, K, =0.006 One could notice that different values are used in these two scenarios This is because: (1) the deadline miss ratio setpoints for different transmission rates are different; (2) this makes it convenient to compare the event-triggered and timetriggered feedback scheduling, see Subsection 8.5.2 Chapter Cross-Layer Adaptive Feedback Scheduling 233 Table 8.1 summarizes some parameters used in the simulation e^eriments, where h^ denotes the nominal sampling period It is worth noting that completely identical results cannot be guaranteed for each run of the simulation even with the same system setup This is a natural consequence of the inherent stochastic feature of communications over WLANs All of the results given below are those representative ones among many others obtained from a variety of simulation runs Variable Value 8.5.1 Table 8.1 h,{ms) 15 Parameters for simulation e^qjeriments ^max(ms) ^ED(ms) Ap^ Ap- 50 500 0.1 0.08 Experiment I : Adaptive Feedback Scheduling vs Traditional Design Method In the first set of simulation e^eriments, the proposed AFS method and the traditional design method without any feedback schedulers (denoted Non-FS) are conpared Since the three control loops are identical and WLAN adopts a random medium access control mechanism without distinguishing between them, all loops are equivalent in principle Therefore, only the responses of one control loop will be given Fig.8.8 depicts the step responses of Loop (i.e y^) under different scenarios when the traditional method is used It can be seen that the system performs quite well when the transmission rate of WLAN is 11 Mbps However, under the second scenario, i.e when the transmission rate drops to 5.5 Mbps with interfering sigaals, the system finally soes unstable Fig 8.8 Control performance without feedback scheduling (a) Scenario I; (b) Scenario II 234 PART IV BANDWIDTH ALLOCATION Fig.8.9 shows the system performance when Adaptive Feedback Scheduling is adopted The system not only performs well under Scenario I, but also achieves good control performance under Scenario IT Fig 8.9 Control Performance with Adaptive Feedback Scheduling (a) Scenario I; (b) Scenario II Fig.8.10 plots the total control cost of the system, which is calculated as the sum of the lAE values of the three control loops Under Scenario I, the system performance under different schemes is almost identical Under Scenario II, Adaptive Feedback Scheduling achieves significant improvement of QoC After time t > s, the total control cost for the traditional method increase quickly, implying that the system goes unstable With the adaptive feedback scheduler, in contrast, the total control cost remains small all the time The overall control performance is quite good From Fig 8.10 it can be found out that when the Adaptive Feedback Scheduling scheme is used, the system performs comparably under different scenarios This can also be seen from Fig.8.9 It can be seen that the proposed Adaptive Feedback Scheduling scheme is able to guarantee h i ^ QoC even if the transmission rate of WLAN drops from 11 Mbps to 5.5 Mbps and interfering signals become present The sampling periods and the deadline miss ratios under different schemes are shown in Fig.8.11 and Fig.8.12, respectively With the traditional method, the sampling periods of all control loops are fixed during runtime, i.e h = l5 ms When WLAN runs on top of 11Mbps (i.e under Scenario I), the deadline miss ratio is small with a mean of 2.3% Consequently, the control performance is good Under Scenario II, the deadline miss ratio remains near 100% after time ^ >4 s, implying that almost all data packets transmitted on the WLAN miss their deadlines This inevitably gives rise to system instability Chapter Cross-Layer Adaptive Feedback Scheduling 235 f I• I• • I Fig 8.10 Total control cost in E?q)eriment I (a) Scenario I; (b) Scenario II Fig 8.11 Sampling periods of control loops (a) Scenario I; (b) Scenario II As can be seen from Fig3.8.11 and 8.12, the adaptive feedback scheduler effectively controls the deadline miss ratio throu^ dynamically adjusting the sampling periods Under Scenario I, the sampling periods of control loops decrease from time 236 PART IV BANDWIDTH ALLOCATION Fig 8.12 Deadline miss ratio of control loops (a) Scenario I; (b) Scenario II ^ =0, and remain at a nearly steady level, i.e around ms after time ^ = 5.5 s During this term, the deadline miss ratio keeps at a low level, with a mean of 2.5% Finally, it approaches its setpoint 5% The levels of the deadline miss ratio under both schemes are close, but the sampling periods are smaller with Adaptive Feedback Scheduling According to sampled-data control theory, the smaller the sampling period the better the control performance Therefore, Adaptive Feedback Scheduling benefits QoC improvement Under Scenario II, the adaptive feedback scheduler successfully avoids overly h i ^ deadline miss ratios by increasing the sampling periods gradually After a transient process, the deadline miss ratio keeps around the setpoint 10% Thanks to the compensation method used, the control performance is not deteriorated by these deadline misses One could notice that both the sampling periods and the deadline miss ratio increase on average under Scenario II relative to Scenario I, which may have some negative effects on the control performance Consequently, the control performance is sli^tly worse in Scenario II than in Scenario I, as shown in Fig.8.9 The above simulation results show that for wireless control systems operating in dynamic environments, the Adaptive Feedback Scheduling scheme is able to effectively attack the problem of transmission rate changes and ambient noise interference, thus improving the QoC of the whole system 8.5.2 Experiment H: Event-Triggered vs Time-Triggered In the second set of e}q)eriments the performance of event-triggered and timetriggered Adaptive Feedback Scheduling methods is compared To facilitate com- Chapter Cross-Layer Adaptive Feedback Scheduling 237 parisons with the event-triggered scheme simulated in the first set of e>q)eriments, the invocation interval for the time-triggered feedback scheduler is set to Tp^ = T^^^ = 500 ms Now that the results for event-triggered Adaptive Feedback Scheduling have been presented above, the control performance under time-triggered Adaptive Feedback Scheduling will be studied below Fig 8.13 depicts the step response and the Integral of Absolute Error of Loop under Scenario I Obviously, the control performance in this case is pretty good Comparing the upper parts of Fig.8.13 with Fig.8.9 it can be seen that under Scenario I the time-triggered and event-triggered Adaptive Feedback Scheduling achieve almost identical control performance Fig 8.13 Control performance for time-triggered adaptive feedback scheduling under Scenario I (a) System Output; (b) lAE The sampling periods and the deadline miss ratio for the system using timetriggered Adaptive Feedback Scheduling are given in Fig.8.14 It is easy to find out that they vary in like manner as under event-triggered Adaptive Feedback Scheduling The main difference between them is that with event-triggered feedback scheduling the sampling periods remain unchanged at some consecutive sampling instants, which implies that the feedback scheduler does not actually execute (see the upper part of Fig 8.11), whereas the sampling periods are updated at every sampling instant when time-triggered feedback scheduling is used (see the upper part of Fig.8.14) The response and the lAE of Loop under Scenario II are depicted in Fig.8.15 It is clear that time-triggered Adaptive Feedback Scheduling also delivers good control 238 PART IV BANDWIDTH ALLOCATION Fig 8.14 Sampling periods and deadline miss ratio for time-triggered Adaptive Feedback Scheduling under Scenario I (a) Sampling Period(s); (b) Deadline Miss Ratio performance Comparing Fig.8.15 with Fig.8.13 indicates that the control performance becomes sli^tly worse under Scenario II than Scenario I As shown in Fig.8.16, the deadline miss ratio is kept around the setpoint 10% by the feedback scheduler adjusting the sampling periods However, the sampling periods and the deadline miss ratio are bigger (on average) than in Scenario I, see Figs.8.14 and 8.16 That is why the system performance is better in Scenario I, just as the system performs with event-triggered Adaptive Feedback Scheduling Fig.8.17 shows the total control costs of the system with time-triggered and event-triggpred Adaptive Feedback Scheduling An interesting result is that the event-triggered invocation mechanism mi^t possibly result in better control performance than the time-triggered invocation mechanism, thou^ the improvement is not so significant In Scenario II, for example, the total control cost for eventtriggered Adaptive Feedback Scheduling decreases 2.4% in comparison with that for time-triggered Adaptive Feedback Scheduling To this point only the control performance is examined, which is actually the ideal system performance without taking into account the effect of feedback scheduler executions That is, in the above simulation e>q)eriments the overhead of feedback scheduling is neglected For simplicity, the times of execution of the feedback scheduler is used as a simple metric for comparing the feedback scheduling overheads Table 8.2 summarizes the total control costs and the times of execution of the feedback scheduler with different invocation mechanisms It is obvious that for different invocation mechanisms the overall QoC remains fairly close The relative Chapter Cross-Layer Adaptive Feedback Scheduling Fig 8.15 Fig 8.16 Control performance for time-triggpred adaptive feedback scheduling under Scenario II (a) System Output; (b) lAE Sampling periods and deadline miss ratio for time-triggered Adaptive Feedback ScheduUng under Scenario II (a) Sampling Period(s); (b) Deadline Miss Ratio 239 240 PART IV BANDWIDTH ALLOCATION Fig 8.17 Total control cost in Ejqjeriment II (a) Scenario I; (b) Scenario II difference of total control costs is less than % In Scenario I, the times of execution of feedback scheduler decreases 31.3% with event-triggered Adaptive Feedback Scheduling as compared to time-triggered scheme In Scenario II it reduces from 16 to 7, with a relative reduction of 56.3% The overhead of feedback scheduling is significantly reduced Table 8.2 Comparison of event-and time-triggered invocations Scenario I Scenario II TT ET TT ET SlAE 1.148 1.127 1.285 1.254 Times of Execution 16 11 16 The above results show that the proposed event-triggered invocation mechanism yields sigaificant reduction in feedback scheduling overhead while providing good feedback scheduling performance, thus improving the efficiency of the feedback scheduling scheme Furthermore, the event-triggered invocation mechanism can also be used to achieve quicker responses associated with the feedback scheduler without introducing excessively large overheads by simply selecting a smaller T^j^ value In this way, the practical performance of feedback schedulers could be further improved Chapter Cross-Layer Adaptive Feedback Scheduling 8.6 241 Summary This chapter deals with dynamic management of the communication resource in wireless control systems Using cross-layer design, an Adaptive Feedback Scheduling scheme has been presented to attack the effects of factors such as noise interference and node movement on control performance This scheme makes use of related information at the physical layer to adjust the sampling periods of the control systems at the application layer The control performance is optimized by means of deadline miss ratio control To meet the requirements deriving from the adverse properties of wireless communications and to avoid the disadvantages of time-triggered invocation in making tradeoffs between response speed and overhead, an event-triggered invocation mechanism has also been suggested, which improves the practical performance of feedback scheduling The emerging area of wireless control is currently in its infancy The method proposed in this chapter offers an effective solution for constructing control systems over wireless networks However, the present work accomplished by this time is still preliminary More systematic analysis, both theoretical and e^erimental, is left for our future work The proposed method could be extended in many aspects One possible extension is improving the cross-layer design framework For example, to take into account the effect of different communication protocols (i.e MAC protocols), the MAC sub-layer may be included in the framework of cross-layer feedback scheduling, thus e}q)loring the interaction between the physical layer, the MAC layer, and the application layer In cases where the energy consumption of the nodes is a concern, physical-layer parameters such as the transmit power may be made available for upper layers Another possibility is improving the Adaptive Feedback Scheduling algorithm Suppose the behaviour of the wireless network in terms of deadline miss ratio could be modelled with sufficient accuracy, for example, an optimized Adaptive Feedback Scheduling algorithm would be easily obtained using the presented design methodology In this context, even the optimal sampling periods with respect to different system states could possibly be derived analytically using related control theory and technology Our future work in this direction also includes the development of an e^eriment system for wireless control systems, which will be used to evaluate the performance of Adaptive Feedback Scheduling with extensive results 242 PART IV BANDWIDTH ALLOCATION References [APN03] [ARZ99] [ARZ06] [ICC03] [IEEE99] [KAW05] [LIN03] [LIN06] [LIU04] [MAT05] [NIK05] [PLO04] [RAM05] C Apneseth, D Dzung, S Kjesbu, G Scheible, W Zimmermann Wireless-Introducing Wireless Proximity Switches Sensor Review, Vol 23, No 2, pp.116 - 122, 2003 K.E Arzen A simple Event Based FID Controller Proc 14th IFAC World Congress, Beijing, China, Vol Q, pp.423 -428, 1999 K.E Arzen, A Bicchi, S Hailes, K.H Johansson, J Lygeros On the Design and Control of Wireless Networked Embedded Systems Proc of IEEE Computer Aided Control Systems Design Symposium, pp.440-445, 2006 ICC panel Defining Cross-Layer DesigQ in Wireless Networking, http ://www.eas.asu.edu/ ~ junshan/ICC03panel.html, 2003 International Standard ISO/IEC 8802 - 8811, ANSI/IEEE Std 802.11, Part 11 Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications The IEEE, Inc., 1999 P A Kawka, A.G Alleyne Stability and Feedback Control of Wireless Network Systems Proc of the American Control Conf (ACC05), Portland, OR, 2005 Q Ling, M.D Lemmon Optimal Dropout Compensation in Networked Control Systems Proc of the IEEE Conf on Decision and Control, 2003 X Lin, N.B Shroff, R Srikant A Tutorial on Cross-Layer Optimization in Wireless Networks IEEE Journal on Selected Areas in Communications, Vol 24, No 8, pp.1452 - 1463, 2006 X Liu, A.J Goldsmith Wireless Network Design for Distributed Control Proc IEEE Conf on Decision and Control, 2004 M Mathiesen, G Thonet, N Aakwaag Wireless Ad-hoc Networks for Industrial Automation: Current Trends and Future Prospects Proc of the IF AC World Congress, Prague, Czech Republic, 2005 G Nikolakopoulos, A Panousopoulou, A Tzes, J Lygeros Multihoping Induced Gain Scheduling for Wireless Networked Controlled Systems Proc of the 44th IEEE Conf on Decision and Control, European Control Conf., Seville, Spain, 2005 N Ploplys, P Kawka, A Alleyne Closed Loop Control over Wireless Networks IEEE Control Systems Magazine, pp.58 - , 2004 H Ramamurthy, D Lai, B.S Prabhu, R Gadh ReWINS: A Distributed Multi-RF Sensor Control Network for Industrial Automation IEEE Wireless Telecommunication Symposium WTS, Pomona, California, pp.24 - 33, 2005 Chapter Cross-Layer Adaptive Feedback Scheduling [SCH06] [TIA06] [TIP03] [TUC03] [VAR03] [WIL05] [XIA05] [XIA07] [YEOl] 243 L Schenato Optimal estimation in Networked Control Systems Subject to Random Delay and Packet Loss IEEE Conf on Decision and Control (CDC 06), San Diego, CA, USA, 2006 Y.C Tian, D Levy, M.O Tade, T Gu, C Fidge Queuing Packets in Communication Networks for Networked Control Systems The 6th World Congress on Intelligent Control and Automation^ Dalian, China, pp.205-209, 2006 Y Tipsuwan, M.Y Chow Control Methodologies in Networked Control Systems Control Engineering Practice^ Vol.11, pp.1099 1111,2003 A Tuck, C Burgess Mesh Meets the Need-Wireless for Industrial Control ISA Technical Information and Communities, www.isa.org, 2003 A Varma, B Ganesh, M Sen, S.R Choudhury, L Srinivasan, J Bruce A control-theoretic approach to dynamic voltage scheduling Proc ACM CASES, Georgia, USA, pp.255 -266, 2003 A Willig, K Matheus, A Wolisz Wireless Technology in Industrial Networks Proceedings of the IEEE, Vol 93, No 6, pp.1130 - 1151, 2005 F Xia, Y.X Sun Event-Triggered Feedback Scheduling of Embedded Control Systems Proc 1st National Conf on Pervasive Computing, Kunming, China, pp.660 - 666, 2005 F Xia, Y.C Tian, C Peng, Y.X Sun Cross-Layer Adaptive Feedback Scheduling of Wireless Control Systems over WLAN Submitted, 2007 H Ye, G.C Walsh, L Bushnell Real-Time Mixed Traffic Wireless Networks IEEE Transactions on Industrial Electronics, Vol 48, No 5, pp.883 -890,2001 Index absolute control error, 23, 161 adaptation, 16, 68, 120 Adaptive Feedback Scheduling, 216, 218 backward approximation, 46 battery, 129, 130 BP, 78, 83 CAN, 22, 55, 187 cascaded feedback scheduling, 186, 188, 194 centralized feedback scheduling, 105 circuit delay, 131 closed-loop, 15,43,47 CMOS, 130, 134 codesign, 3, 10 coefficient, 138, 140, 195, 226 Commercial Off-The-Shelf, communication delay, 6, 57, 187 complexity analysis, 85 computing systems, 11, 48 congestion control, 20, 102 control cost, 17, 48, 77 control engineering, 21 control error, 23, 47 control input, 43, 60, 137 control network, 4, 22, 186 control task, 4, 18 communication protocol, 22, 53, 189 controlled variable, 12, 43 convergence, 3, CPU resource, 12, 16 CPU speed, 20, 68, 129 cross-layer desigQ, 24, 212 CSMA/BA, 55 CSMA/CA, 24, 54, 222 CSMA/CD, 54, 55 data loss, 5, 56 DC servo, 61 deadline, 6, 48 deadline miss, 48, 61, 95, 111 deadline miss ratio, 64, 67, 135, 186 decentralized feedback scheduling, 105 defuzzification, 108, 112 delay, 5, 16,22 design considerations, 226 design methods, 46 DeviceNet, 55, 189 direct feedback scheduling, 19, 186, 198 discrete-time design, 45, 59 discretization, 45, 59, 108 Dynamic Voltage Scaling, 15, 68, 129 dynamic-priority scheduling, 50, 191 EDF, 6, 12, 19,51 embedded control systems, 3, 11, 26, 101 Index energy consumption, 129, 134, 142 energy management, 15, 150 Ethernet, 4, 22, 53 event-triggered, 59, 161 event detector, 230 execution time factor, 133, 137 e?q)onential, 163 feedback, 3, 12, 165, 175 feedback control scheduling, 12, 150 feedback scheduling, 6, 129, 192 fixed-priority scheduling, 50, 191 forward approximation, 46 framework, 12, 70 fundamentals, 42 fuzzification, 108 fuzzy control, 13, 102 fuzzy feedback scheduling, 102, 123 gain scheduling, 24, 25 graceful degradation, 16, 154 half-duplex, 54, 217 hard real-time, 6, inference, 108, 111 input scaling factor, 109, 114 Integral of Absolute Error, 47, 65 integral of time-wei^ted absolute error, 47 integrated control and communication, 3, 10 integrated control and computing, 3, 10 integrated feedback scheduling, 187, 189 integration, 8, intelligent feedback scheduling, 78, 97 inverted pendulum, 43, 88 invocation interval, 69, 85 jitter, 6, 17 Kuhn-Tucker condition, 81 Levengerg-Marquardt, 84 linear, 13, 14 linguistic rules, 109, 111 linguistic values, 109, 110 linguistic variable, 109 link capacity, 217, 219 look-up table, 69, 109 loss, 22, 42 LQG, 88, 141 MAC protocols, 52, 53 manipulated variable, 12, 19 measured output, 43 measurement noise, 26, 100 membership function, 110 mobile robot, 100, 102 model, 6, 11 modelling, 13, 27 multiple sampling, 220 network scheduling, 23, 187 networked control, 6, networked control systems, 6, 20 neural feedback scheduling, 77, 78 neural networks, 77, 78 non-control task, 88, 108 normalized energy saving, 143 open loop, 61 open-loop scheduling, 64, 91 optimal feedback scheduling, 77, 78 optimization, 12, 14 OSI reference model, 223 output scaling factor, 109, 114 overhead, 18, 49 overload, 7, 14 packet, 22, 23 performance index, 16, 17 period rescaling factor, 106, 107 perspective, 25, 41 PID control, 13, 18 245 246 Index preemptivity, 192 priority, 50, 51 pseudo code, 114, 140 proportional control, 219, 227 quality of control, 5, quantization, 114, 115 real-time communication, 5, 26 real-time computing, 26, 40 Real-Time Control Systems, 3, real-time scheduling, 6, 12 real-time systems, 3, reference input, 43, 161 resource constraints, 16, 18 resource reclaiming, 168, 181 resource scheduling, 6, 10 response time, 15, 50 RM,6, 50 rule-base, 108, 109 sampled-data control systems, 45, 59 sampling, 6, 14 sampling period, 14, 16 schedulability, ,16 separation of concerns, sequential quadratic programming, 81 server systems, 14, 102 setpoint, 43, 47 Signal-to-Noise Ratio, 222 soft real-time, 6, 12 stability, 5, switch threshold, 199, 201 target tracking, 100, 102 task, 3, TDMA, 52, 53 temporal non-determinism, 5, 57 time-triggpred 18, 106, 136 timing, 6, training, 83, 84 Justin's approximation, 46 underload, 202 universe of discourse, 112, 114 utilization, 12, 14 vacant sampling, 60, 220 variability, 5, 26 WCET, 6, 48 wireless control, 20, 24 wireless control systems, 20, 24 wireless network, 24, 61 WLAN, 5, 24 workload, 3, ... life sciences, engineering, environmental sciences, mathematics, and physics Feng Xia Youxian Sun Control and Scheduling Codesign Flexible Resource Management in Real- Time Control Systems With 118... Real- Time Control Systems Since resource scheduling becomes the main concern in this context, this area is often referred to as control and scheduling codesign or integrated control and scheduling, ... disciplines in terms of both theory and technology In the context of real- time control, there are intuitively two kinds of convergences: integrated control and computing, and integrated control and