Real-Time Systems Hermann Härtig Introduction

Size: px

Start display at page:

Download "Real-Time Systems Hermann Härtig Introduction"

Poppy Richards
5 years ago
Views:

1 Real-Time Systems Hermann Härtig Introduction 08/10/10

2 Organisation Issues Web-Page Subscribe to the mailing list!!! Time 3 SWS: 2 lectures + 1 exercises Thursday, 5th/6th DS (whether or not 6th DS is needed is usually announced the day before on the Web-Site) WS 2010/11 Real-Time Systems, Introduction / Hermann Härtig 2

3 Real-Time Systems Defnition (strict) Systems, whose correctness depends (not only) on the correct logical results of computations (but also) on meeting all deadlines. Results must be produced before previously stated deadlines. Deadlines are dictated by the environment of the system. WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 3

4 Real-Time Systems Defnition (weaker) Systems, whose quality depend (not only) on the logical results of computations (but also) on the time these results are produced. Requested timing characteristics originate from the environment of the system. WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 4

5 Some Examples for Weaker Notions Occasional misses of deadlines are ok. The value of a result depends on the time it becomes available. An imperfect result early may be better than a perfect result (too) late. The more results can be obtained before a given deadline the better. There are more- and less-critical activities ( mixed criticality ) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 5

6 Some more Terminology hard vs. frm vs. soft and safety critical safety critical a malfunctioning system endangers human life hard real-time systems deadlines are strict, missing has fatal consequences for the controlled object or humans, must work under peak load frm real-time systems deadlines are strict soft real-time systems deadlines should be met, value of results decreases with time, graceful degradation under peak load is acceptable WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 6

7 A Saying Hard real-time systems are hard to build, soft real-time systems even harder. Doug Jensen (?) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 7

8 Embedded Systems (also see Lecture Prof. Hochberger) (Real-Time) Computer systems as integral part of a larger system... Common characteristics: large numbers (mass market) static structure (changing) minimize mechanics software very hard to change after deployment (ROM) cars, consumer electronics,... Well over 90% of microprocessor production in terms of units (in contrast to value) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 8

9 Periodic Processes Start-periodic-thread: While forever { Begin period (length of period) compute End period } important parameters: period length (worst) case execution time per period where do the times come from... WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 9

10 Principle Interfaces Person Real-Time Computer System Controlled Object times induced thru both interfaces WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 10

11 Layers of Control Pilot Pilot Interface Navigation Fly by Wire Aircraft Multiples stages may induce diferent times WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 11

12 Simple Control System external infuences actuators controlled system sensors y(t) u(t) control e(t) dif operator WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 12

13 Simple Digital Control System external infuences actuators controlled system sensors y(t) D / A u(t) control e(t) dif D / A periodic sampling operator WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 13

14 PID controller Continuous formula: t u=k p e k i =0 e d T d de dt Approximation by periodic sampling (rate T) Integral via Simpson's Rule: T 3 (e k 2 +4e k 1 +e k ) Diferential: e k e k 1 T Then: u k :=u k 2 a e k b e k 1 c e k 2 With a=k p + k T i 3 + T d T b= 4k T i 3 T d T c= k i T 3 WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 14

15 Digital Controllers at every xx time units do read r,y u k :=u k-2 +a*e k +b*e k-1 +c*e k-2 write u done sample period T depends on: reactivity of person (<100 ms) reactivity of controlled object (? ) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 15

16 Simple Control Loop (Kopetz) operator Controlling Computer System Controlled Object valve steam WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 16

17 Times Temperature rise time rise time (10% or other small neighborhood) object delay (inertia of contr. process) SOLL 90% sampling period rule of thumb < 1/10 to 1/20 rise time computation delay computation delay jitter (< sample period) IST 10% steam deadtime = object delay + comp delay shorter sampling periods result in: object delay real-time smoother operation less oscillation more resources used WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 17

18 Complications of simple model: state complete state of controlled object is not represented in sampled data (robot arm) state... at every xx timeunits do read... compute output and new state from samples and current state write... done WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 18

19 Digital Control System actuators controlled system sensors y(t) u(t) control state model dif trajectory generator WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 19

20 Complications of simple model: Multirate Controllers multiple sensors, actuators, and state variables with diferent sampling rates often the larger are integer multiples of smaller rates harmonic rates example: rotation, temperature (engine control) method (successive loop closure): start with highest rate sensor integrate it in system and consider it part of the controlled object determine next rates (as multiples of fastest) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 20

21 Example: Engine Control Controls amount of fuel precise time of injection (0.1 degree of crankshaft) depending on rpm, 10 ms per 360 degree ca. 3micro-sec precision WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 21

22 Example: Aircraft Control Fly by wire Navigation Pilot interface Observations: hierarchical control very high safety required WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 22

23 Example: Air Trafc Control Signal processing Tracks and displays position of aircrafts. Detects danger of collision situations. Observation: very dynamic, planes come and leave hierarchical structure less stringent timing requirements at higher levels WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 23

24 Example: Measurement Systems Signal Conditioning use physical, possibly non linear efect and produce conditioned data for human perception Data Sampling sample data at precise points in time Observation: often it is more important to collect data with precise timestamps than collecting data at precise times WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 24

25 Example: Protection/Alarm Systems Continuous Monitoring whether a physical process meets certain criteria Alarm otherwise Interpretation of raw data and deduction of causality for human operator (reduction of alarm showers) Observation: Timestamps are used for detection of causal dependency. WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 25

26 Example: Signal Processing Filters Compression Radar Tracking Observation: required computing can vary widely WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 26

27 Example: Multimedia Systems 25 Frames per second max 20 ms audio jitter includes disk/network dynamic: additional applications come and go animation in real-time WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 27

28 Example: AWACS 10 Hz radar signals Any number of objects to track... WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 28

29 Example: Long Term deadlines Intel s fabs coordinations... WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 29

30 Example: Assembly line remove object from belt if not possible before object passed the point of removal, stop the belt WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 30

31 Events Task actions are at certain times or by events Periodic sporadic Stochastically non-deterministic Non-stochastically non-deterministic More in chapter on models! WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 31

32 Necessary, Desirable Properties Predictability / Sustainability establish behavior under extreme cases/situations then be sure that RT properties are ensured in all cases (more later) Composability if for a subsystem some property has been established, it should hold after integration in subsystem (Kopetz) Maintainability slight changes of parameters, repair,... WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 32

33 Necessary, Desirable Properties Safety no desasters Fault Tolerance tolerate failures how to construct reliable systems from unreliable components Security maintain properties even under attack WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 33

34 Computer Science Areas Parallel programming Low Level I/O Computer architecture Communication Modeling techniques (Scheduling) Programming languages Fault Tolerance SW design Operations Research Computer science theory WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 34

35 Lecture Overview (not a temporal order) Foundations: Introduction (Org, Defnitions, Examples, CS-Areas) Time and Order: Time, Clocks (logical and physical) Modeling Real-Time Systems: Tasks, Jobs, Periods, Resources Time Driven Systems: Mars Video, Principles, cyclic executives Event Driven Scheduling: RMS, EDF, Priority Inversion... Other areas: Hardware: Busses, Caches, Catch and Compare, Operating Systems: general,case-studies: QNX, Posix, may be: AUTOSAR Communication: near (CAN), far(atm), alarm showers (Mars-Example) HLL: Ada, Synchronous HLL, evt: RT-Java real-time literature recherche (exercises) May be s: Multiprocessor scheduling Mixed criticality (per video from Sanjoy Baruah) Fault Tolerance in Real-Time, Replica Determinism Statistical Scheduling: SRMS, QRMS Non Periodic RT-Systems (per video from Doug Jensen) WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 35

36 Material Textbooks (available in library): Hermann Kopetz Real-Time Systems (Kluwer) Jane Liu Real-Time Systems (Prentice Hall) Additional papers: provided in lectures WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 36

37 Distributed Real-Time Systems... Why The controlled object often is distributed Technology... microcontrollers and busses Isolated treatment of subsystems Fault tolerance through replication Problems: (global) time and order communication in real-time WS 2010/11 Real-Time-Systems, Introduction / Hermann Härtig 37

Introduction to Real-Time Systems

Introduction to Real-Time Systems Real-Time Systems, Lecture 1 Martina Maggio and Karl-Erik Årzén 16 January 2018 Lund University, Department of Automatic Control Content [Real-Time Control System: Chapter