Hybrid Control and Communication Karl Henrich Johansson KTH - - PDF document

▶
hybrid control and communication karl henrich johansson
SMART_READER_LITE
LIVE PREVIEW

Hybrid Control and Communication Karl Henrich Johansson KTH - - PDF document

Apr 14, 2023 134 likes •646 views

st 1 HYCON PhD School on Hybrid Systems www.ist-hycon.org www.unisi.it Hybrid Control and Communication Karl Henrich Johansson KTH Stockholm, Sweden kallej@s3.kth.se scimanyd suounitnoc enibmoc smetsys dirbyH lacipyt (snoitauqe ecnereffid

slide-1
SLIDE 1

14

HYSCOM

IEEE CSS Technical Committee on Hybrid Systems

scimanyd suounitnoc enibmoc smetsys dirbyH lacipyt (snoitauqe ecnereffid ro laitnereffid) scimanyd etercsid dna stnalp lacisyhp fo fo lacipyt (snoitidnoc lacigol dna atamotua) fo senilpicsid gninibmoc yB .cigol lortnoc ,yroeht lortnoc dna smetsys dna ecneics retupmoc dilos a edivorp smetsys dirbyh no hcraeser ,sisylana eht rof sloot lanoitatupmoc dna yroeht fo ngised lortnoc dna ,noitacifirev ,noitalumis egral a ni desu era dna ,''smetsys deddebme`` ria ,smetsys evitomotua) snoitacilppa fo yteirav ssecorp ,smetsys lacigoloib ,tnemeganam ciffart .(srehto ynam dna ,seirtsudni

www.ist-hycon.org www.unisi.it

1 HYCON PhD School on Hybrid Systems

st

Siena, July 1 9-22, 2005 - Rectorate of the University of Siena

Hybrid Control and Communication Karl Henrich Johansson

KTH Stockholm, Sweden

kallej@s3.kth.se

slide-2
SLIDE 2
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Hybrid Control and Hybrid Control and Communication Communication

Karl H. Johansson Department of Signals, Sensors and Systems Royal Institute of Technology Stockholm, Sweden www.s3.kth.se/~ kallej

slide-3
SLIDE 3
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Outline Outline

  • Lecture I: Control under communication constraints

– Motivating applications – Communication constraints – Compensation for delay and loss – Integrated design of control and communication – References

  • Lecture II: Hybrid control of communication systems

– Packet-switched networks – Hybrid model of congestion control – References

  • [Lecture III (by L. Palopoli): Stabilization of quantized systems]
slide-4
SLIDE 4
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Lecture I: Lecture I: Control under communication constraints Control under communication constraints

  • Classical control theory are

based on perfect exchange

  • f information
  • Modern control systems are
  • ften networked

+ added flexibility + cheaper implementation

  • Sensor and actuator data

are then transmitted over a shared network resource

– added uncertainty – higher complexity

Plant C S A Plant C S A

slide-5
SLIDE 5
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Motivating applications Motivating applications

  • Scania truck
  • Volvo XC90
  • SMART-1 spacecraft
  • Power control in wireless system
  • Congestion control in communication network
slide-6
SLIDE 6

Networked control architecture Networked control architecture

  • f a
  • f a Scania

Scania truck truck

  • Control units connected through 3 controller area

networks (CANs) coloured by criticality

  • CAN is a standard introduced by Bosch 1986

Trailer

7-pole 15-pole

AUS

Audio system

ACC

Automatic climate control

WTA

Auxiliary heater system water-to-air

CTS

Clock and timer system

CSS

Crash safety system

ACS 2

Articulation control system

BMS

Brake management system

GMS

Gearbox management system

EMS1

Engine management system

COO1

Coordinator system

BWS

Body work system

APS

Air prosessing system

VIS 1

Visibility system

TCO

Tachograph system

ICL1

Instrument cluster system

AWD

All wheel drive system

BCS2

Body chassis system

LAS

Locking and alarm system

SMS

Suspension management t

SMS

Suspension management system

RTG

Road transport informatics gateway

RTI

Road transport informatics system

EEC

Exhaust Emission Control

SMD

Suspension management dolly t

SMS

Suspension management system

ATA

Auxiliary heater system air-to-air

Green bus Red bus Yellow bus ISO11992/3 ISO11992/2 Diagnostic bus Body Builder Bus Body Builder Truck

slide-7
SLIDE 7

Networked control architecture of a Volvo XC90 Networked control architecture of a Volvo XC90

  • 3 CAN networks connect up to

40 control units

  • Example of control system using

CAN is vehicle dynamics control (electronic stability program)

slide-8
SLIDE 8

Networked control architecture Networked control architecture

  • f the SMART
  • f the SMART-
  • 1 spacecraft

1 spacecraft

  • First European lunar mission, launched Sep 2003
  • CAN networks for control system (“system”) and for

scientific experiments (“payload”)

  • Node and communication redundancies

Sun sensors (3 in total) Star tra Hydr (4 in Reaction wheels (4 in total) EP thruster and orien mechanism Sun sensors (3 in total) Star tra Hydr (4 in Reaction wheels (4 in total) EP thruster and orien mechanism

slide-9
SLIDE 9

Distributed power control in cellular systems Distributed power control in cellular systems

  • Power control in each mobile station tries to keep signal-to-interference

ratio (SIR) at a threshold value

  • Disturbances from channel fluctuations and interfering traffic
  • Control in mobile station based on quantized estimate of SIR

communicated from base station

slide-10
SLIDE 10

Sender Receiver Sender Receiver

Congestion control in packet Congestion control in packet-

  • switched

switched data communication network data communication network

  • Each sender regulates sending

rate based on congestion information from receiver

  • Variations in available

bandwidth and traffic load

  • Congestion indicated implicitly

through missing acknowledgement packets

  • Quantized control command

[Discussed in detail next lecture]

slide-11
SLIDE 11
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Examples of Examples of networked control architectures networked control architectures

Plant C S A Plant C S A S A C Plant Plant C S A Plant C S A

slide-12
SLIDE 12
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

What is hybrid in networked control? What is hybrid in networked control?

  • Networked control systems are inherently hybrid, not only

because interaction of physical plant and computer control, but also because they have

– mixture of event- and time-triggered communication protocols – asynchronous network nodes (no global clock) – quantized sensor data to limit network traffic – symbolic control commands to simplify design and operation

  • Now on we mainly discuss

modelling and compensating some communication constraints, cf., Mitter’s lecture

slide-13
SLIDE 13
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Control with constrained communication Control with constrained communication

  • Limitations in the communication of sensor and actuator data

impose constraints on the control system

  • Communication imperfections include

– Delay and jitter – Quantization [Lecture by Palopoli] – Packet loss – Bit error – Outage (lost connection)

slide-14
SLIDE 14
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Time delays Time delays

  • Delays in communication due to buffering and propagation delays
  • Delays are bad for control loops (avoid if possible)
  • Delays can be fixed or varying, known (measurable) or unknown
  • Data loss can be interpreted as infinity delay
slide-15
SLIDE 15
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Unknown and fixed time delay Unknown and fixed time delay

slide-16
SLIDE 16
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Remark: Closed-loop gain needs to be sufficiently small.

Easy to check through Bode plot.

Unknown and varying time delay Unknown and varying time delay

slide-17
SLIDE 17
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Proof Proof

slide-18
SLIDE 18
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Relation to Relation to Nyquist Nyquist Criterion Criterion

slide-19
SLIDE 19
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Unknown and varying delay jitter Unknown and varying delay jitter

slide-20
SLIDE 20
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Known time delay Known time delay

slide-21
SLIDE 21
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Time delay estimation Time delay estimation

  • Major improvement in control

performance if delays are known/measurable (or can be accurately estimated)

Example

  • CAN protocol (discussed earlier) is event-triggered and does not give

timing guarantees in general

  • TTCAN (Time-triggered communication on CAN) is an extension to the

CAN standard targeting the need from sampled-data control

slide-22
SLIDE 22
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Compensating known delays: Compensating known delays: State State feedback controller feedback controller

slide-23
SLIDE 23
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Compensating known delays: Compensating known delays: Output Output feedback controller feedback controller

slide-24
SLIDE 24
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Large delays and out Large delays and out-

  • of
  • f-
  • order delivery
  • rder delivery

Large known delays can be treated as before by extending the estimator state (one dim per extra sampling period delay)

  • Buffers can handle out-of-order delivery,

but may also increase delays

  • Don’t wait for late data, but when they

arrive use them to adjust old estimates

slide-25
SLIDE 25
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Packet loss Packet loss

Modify traditional observer; simplest case:

  • Can be hard to handle packet loss: when decide that ?

– E.g., TCP uses TimeOut variable to decide when a packet is lost

  • It can be better to drop data, than use old information for control
slide-26
SLIDE 26
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Bit errors and outages Bit errors and outages

  • Bit errors are due to rapid variations in the physical channel
  • Unlikely in wired systems, but important in wireless systems
  • Compensation through forward error correction (coding)
  • Outages are sudden events when connection is lost
  • A severe problem in most wireless systems
slide-27
SLIDE 27
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Integrated design of control and Integrated design of control and communication systems communication systems

  • Up to here, communication has been considered as a

disturbance or model imperfection affecting control

  • What about jointly design control and communication,

along the proposal in Mitter’s lecture?

  • Let’s illustrate with an example
slide-28
SLIDE 28
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Example: Example: Stabilization of networked systems Stabilization of networked systems

  • Consider joint state feedback stabilization of a set plants,

when only one plant can utilize the bus at a time:

  • What control and communication policy should be adopted?

Communication bus

slide-29
SLIDE 29

Hybrid system representation Hybrid system representation

  • How choose the guard conditions of the hybrid automaton to

stabilize the system?

Communication bus

slide-30
SLIDE 30

“ “Largest state first Largest state first” ”-

  • policy

policy

Theorem [Hristu-V. and Kumar] For scalar unstable systems:

slide-31
SLIDE 31
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Summary Summary

  • Communication and networking

important in growing number of control applications

  • Communication delays, quantizers, losses, errors and outages can be

viewed as constraints imposed on the control system

  • Can in many cases be suitably modelled as hybrid systems
  • Design methodologies exist for certain classes of constraints, but much

more remains to be done

  • Desirable to jointly design control and communication system

– Control algorithms need to adapt to changing network conditions – Communication protocols should be aware of control needs – But, other network applications set competing restrictions

slide-32
SLIDE 32
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

References References

  • D. Hristu-Varsakelis and W. S. Levine, Ed.’s, Handbook of

Networked and Embedded Control Systems, Birkhäuser, 2005

  • L. G. Bushnell, Ed., Special Issue on Networks and Control,

IEEE Control Systems Magazine, Feb, 2001

  • B. Lincoln, "Dynamic programming and time-varying delay

systems,“ PhD thesis, Lund University, 2003

  • K. J. Åström and B. Wittenmark, Computer-Controlled

Systems, 3rd edition, Prentice Hall, 1997

  • 2E1245 Hybrid and Embedded Control Systems,

http://www.s3.kth.se/control/kurser/2E1245 http://www.s3.kth.se/~ kallej

slide-33
SLIDE 33
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Outline Outline

  • Lecture I: Control under communication constraints

– Motivating applications – Communication constraints – Compensation for delay and loss – Integrated design of control and communication – References

  • Lecture I I : Hybrid control of communication systems

– Packet-switched networks – Hybrid model of congestion control – References

slide-34
SLIDE 34
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Traffic Traffic control control in in packet packet-

  • switched

switched communication communication network network

Objective is to

  • Give each user suitable service
  • Utilize network resources efficiently

Obtained through two control mechanisms:

Spatial control

  • route traffic short way through the network
  • receiver address in header of each packet
  • shortest-distance matrix in each router
  • updated on a slow time scale

Temporal control

  • adjust sending rate to available bandwidth
  • base on info available in sender (end-to-end)
  • implicit bandwidth estimate through ack’s
  • updated on a faster time scale

Sender Receiver Sender Receiver

slide-35
SLIDE 35
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Sender Receiver Sender Receiver

Temporal Temporal congestion congestion control control

  • Each network link has a limited capacity
  • Variations in traffic is primarily handled

by temporary storage in router buffers

  • If a buffer gets full, it simply throws

away incoming packets

  • Hence, congestion leads to lost packet
  • Acknowledgements (ack’s) indicate

sender, who can take action

Sender Receiver

slide-36
SLIDE 36
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Transmission control protocol (TCP) Transmission control protocol (TCP)

  • TCP implements a congestion controller that regulates the sending rate
  • Control variable is the congestion window w, which represents

number of outstanding (not-yet-acknowledged) packets

  • Control is based on implicit feedback information from ack’s
  • TCP follows additive increase multiplicative decrease (AIMD) strategy

Sender Receiver time Sender Receiver time

drop

slide-37
SLIDE 37
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

TCP TCP congestion congestion avoidance avoidance

Typical windows evolution:

slide-38
SLIDE 38
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Hybrid dynamics of a queue Hybrid dynamics of a queue

Sender Receiver

Queue Full Queue Active Queue Empty

slide-39
SLIDE 39

Hybrid model of TCP over single link Hybrid model of TCP over single link

Sender Receiver

slide-40
SLIDE 40
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

The The discrete discrete states states of TCP

  • f TCP

Congestion Avoidance Timeout Fast Retransmit Slow-Start

  • Congestion avoidance is the main state
  • Timeout handles severe congestion
  • Slow-Start gives faster growth initially
slide-41
SLIDE 41
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Hybrid model of TCP Hybrid model of TCP

slide-42
SLIDE 42
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Alternative models of network traffic: Alternative models of network traffic: Packet models and fluid models Packet models and fluid models

Packet models

  • Model each individual packet

(event-driven)

  • Accurate but computationally

heavy

Fluid models

  • Averaged fluid quantities

(time-driven)

  • Capture only steady-state

and slow behaviors

  • Hybrid model combines features of these traditional network models
slide-43
SLIDE 43
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

TCP over wireless links TCP over wireless links

  • Integration of Internet and cellular networks hard due to radio link variations
  • When used over wireless links, TCP cannot ensure a high link utilization
  • Packet drops, bandwidth and delay variations in radio link erroneously

indicate network congestion to TCP

  • How do radio links affect TCP throughput?
  • Can we make the radio link and the cellular system “TCP friendly”?

Mobile Receiver Sender Base Station

Internet

Radio Link : TCP protocol : RLC protocol

slide-44
SLIDE 44
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

A hybrid cascade control problem A hybrid cascade control problem

  • Radio link transforms losses into random

delays

  • Key dynamics from cascaded feedback

control loops – Inner and outer power controls – Link-layer retransmission – TCP

  • Increased probability of spurious timeout

gives reduced TCP throughput

  • Adjust link layer properties to optimize TCP

throughput

slide-45
SLIDE 45
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

New feedback protocols for wireless Internet New feedback protocols for wireless Internet

  • Improved TCP throughput through new radio network feedback protocol
  • Proxy between cellular system and Internet adapt sending rate to radio

bandwidth variations obtained from radio network controller (RNC)

3G-GGSN

3G CN

Terminal BTS PROXY 3G-SGSN RNC BTS App Serv

Internet

slide-46
SLIDE 46
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

New feedback protocols for wireless Internet New feedback protocols for wireless Internet

  • Hybrid controller in proxy regulates sending rate based on

– Events generated by radio bandwidth changes obtained from RNC – Sampled measurements of queue length in RNC

  • Improved time-to-serve-user and utilization compared to

traditional end-to-end TCP

slide-47
SLIDE 47
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Challenges in network traffic control Challenges in network traffic control

  • Deployment

– Implementation in end computers

  • Distributed control

– Communication constraints – Implicit state information

  • Complex interacting dynamics

– Network, protocol and user dynamics – Large and varying time delays – Wired and wireless links – Packet loss

  • No clear optimality objective

– Network throughput – User throughput – Response time – Fairness

slide-48
SLIDE 48
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

Summary Summary

  • Hybrid model of congestion control in packet-switched networks
  • Combine event-driven packet models with time-driven fluid models
  • Accurate on time-scale of the round-trip time
  • Enables analysis and efficient simulations of congestion control
  • Interactions between wireless links and TCP lead to performance loss
  • Hybrid controller gives improved user experience and network utilization
slide-49
SLIDE 49
  • K. H. Johansson, 1st HYCON PhD School on Hybrid Systems, Siena, 2005

References References

  • V. Jacobson, “Congestion avoidance and control,” in Proc. of SIGCOMM,

1988, pp. 314–329

  • J. Walrand and P. Varaiya, High-performance communication networks,

Morgan Kaufmann, 2nd edition, 2000

  • J. Hespanha, S. Bohacek, K. Obraczka, and J. Lee, “Hybrid Modeling of

TCP Congestion Control”. In M. D. Di Benedetto, A. Sangiovanni- Vincentelli, Hybrid Systems: Computation and Control, number 2034 in

  • Lect. Notes in Comput. Science, 2001
  • N. Möller, C. Fischione, K. H. Johansson, F. Santucci, and F. Graziosi,

“Modelling and control of IP transport in cellular radio links”, IFAC World Congress, 2005

  • N. Möller, I. Cabrera Molero, K. H. Johansson, J. Petersson, R. Skog, and

Å. Arvidsson, “Using radio network feedback to improve TCP performance

  • ver cellular networks”, IEEE CDC-ECC, 2005

http://www.s3.kth.se/~ kallej