II Characteristics & Constrints of Es
Uploaded by
meseret sisayII Characteristics & Constrints of Es
Uploaded by
meseret sisayEmbedded Systems:
Characteristics and Constraints
Pascal Raymond
Verimag-CNRS
MOSIG - Embedded Systems
What is an Embedded System?
Minimal definition
A computer system dedicated to a particular function
• system: mix of software/middleware/hardware
• particular function: not really/fully programmable, does 1 thing
Related/similar notions
More or less “synonyms” emphasize particular characteristic(s):
• Reactive systems: perform “everlasting” interaction with the environment
• Real-Time systems: must react “instantaneously” with respect of a particular
environment
• Cyber-physical: emphasize the difference between the system (digital, discrete
time) and its environment (physical, continuous time)
• ... and historically, strongly related to control engineering systems.
What is an Embedded System? 1/14
Computer Systems in Everyday-Life Objects
• Trains, subways, cars ...
• Avionics and space
• Consumer electronics (phones, digital cam-
eras, ...)
• Smart cards
• Household appliances
• Telecom equipments
• Computer Assisted Surgery
• Smart buildings and Energy
What is an Embedded System? 2/14
Characteristics
Various problems/difficulties ...
• Real time: system must be fast enough to react to prhysics
• Criticity: safety-critical and/or business critical
• Limited resources: memory, processor, energy, space
... whose importance depends on the particular domain. What are the main
problems for these domains ?
• Embedded control (train, cars, planes, power plants) ?
• Consumer Electronics ?
• Sensor Networks ?
Let’s focus on embedded control ...
What is an Embedded System? 3/14
Detailed Example: Embedded Control
• In trains, cars, aircraft, space objects ...
• and also power plants, elevators etc.
Characteristics
• The environment is mainly physical, with more or less human intervention
• Submitted to strong real-time constraints (often called: hard-real time)
• They are safety-critical systems
• The computer system is the implementation of a control engineering solution
• The computer system is reactive
Detailed Example: Embedded Control 4/14
A (tiny) example: heater control
The environment
AIR
Actuator
Heater
Sensor
Reactive
ON, OFF
System
Temperature
initializations
while (true) { • Real-time: time to execute the code from (1)
--- point (1) to (2) must be short enough
get inputs
• Reactive: output to the environment influ-
from the sensors
compute outputs ence future inputs
and update memory • Criticity/Safety: badly controlled outputs may
write outputs have dramatic consequences e.g., this a
on the actuators
(small) part of a nuclear power-plant con-
--- point (2)
} troller
Detailed Example: Embedded Control 5/14
Real-Time Programming Problems
• Write code that is sufficiently fast (not always possible to “try a faster machine”)
• Be able to tell how fast your program is, in advance (Worst-Case-Execution-Time
static evaluation)
• It’s not always possible to write single-loop code, because of the intrinsic
parallelism of a reactive system.
e.g., multiple sensor-computing-actuator lines, like temperature and pressure
Detailed Example: Embedded Control 6/14
Safety Problems
• Criticity: faults may be irreparable (lives, environment), or just very expensive
(e.g., launcher, Ariane 5 flight 501 1996)
• HW failures:
,→ HW Fault-tolerance: a whole domain, mainly based on redundancy (e.g.,
several sensors + voter, several processors running the same code)
necessary since HW may break down.
• SW “failures”:
,→ a SW does not “break down”, it is buggy
,→ (run-time) SW Fault-tolerance ?
e.g, several codes developed independently from the same specification...
,→ (off-line) classical methods to track bugs: programming methodology;
intensive testing; when possible: formal verification
Detailed Example: Embedded Control 7/14
Safety and certification
• Critical ES are submitted to Design norms, defined by certification authorities.
• Example in civil avionics: DO178B
,→ a bundle of definitions and rules
,→ classify risks form “none” to “catastrophic”
,→ recommends/imposes design/validation methods depending on the level
,→ basically: an aircraft whose ES is not DO178B certified is not allowed to fly
• Other examples:
,→ Railways: IEC 62279
,→ Automotive: ISO 26262 (concerns safety in general, including SW)
Detailed Example: Embedded Control 8/14
Centralized or Distributed Systems?
• Fact: centralized systems are far simpler than distributed ones
• However, distribution is required:
,→ HW fault-tolerance
,→ topology of sensors/actuator lines (several dozens in an aircraft)
,→ sometimes, for efficiency purpose.
• Other fact: distributed real-time programming is very hard
Detailed Example: Embedded Control 9/14
Designing Embedded System
Main Difficulties for the Design of Embedded Systems
• Real-time parallel and distributed programming (choice of a programming
language?)
• Relation with control engineering
• Intricate dependency between HW, application SW, and OS or middleware
• Certification authorities
• Several degrees of dynamicity (from simple reconfigurations to mobile code...)
Designing Embedded System 10/14
Industrial Practice
• “Hand-craft”: use general purpose tools and languages (Java/C/assembly...)
• Domain Specific Languages: real-time features (multi-tasks, timers, synchro),
specific device operation (sensor/actuator libraries)
• Model-based design: more radical, continuity between high level design (control
engineering problem) and implementation
Designing Embedded System 11/14
Example: Simulink
• Originally: design/simulation tool for
control engineers
• Allows to simulate conjointly:
,→ a continuous-time model of the
environment
,→ a discrete-time solution of the con-
troller
• Implementing the controller:
,→ classically manual encoding (Simulink = specification language)
,→ more and more: automat-ic/ized code generation (Simulink = programming
language)
• Widely used in automotive, transportation, and all domains where control engineering
culture is strong.
Designing Embedded System 12/14
Example: Scade
• Close to control engineering concepts
(block-diagram)
• Formal language, deterministic speci-
fication
• Programming language/environment:
,→ Software engineering features (modularity, libraries)
,→ Automatic code generation:
KCG compiler is DO178B qualified
in particular, eliminates the need of low-level code testing
,→ Thanks to formal semantics, high-level validation possible (and sufficient cf. qualif.):
automated testing, formal verification
• Widely used (imposed) for high-critical systems (avionics, helicopters, power plants)
Designing Embedded System 13/14
Summary
Important points/problems
• General purpose vs dedicated languages
• Validity/correctness: functional (no bugs), extra-functional (time)
This course
• Focus on functionality
• How to design/validate safe ES:
,→ programming languages (features? styles?)
,→ code generation
,→ functional validation (formal methods?)
,→ timing validation (Worst Case Exec time ?)
• Based on the so-called “Synchronous approach”
Summary 14/14
