ReArchitecting System Science in Engineering Education

1
Re-Architecting System Sciencein Engineering
Education

• Presented by
• Edward A. Lee
• Chess, UC Berkeley
• Curriculum Council Kickoff Meeting
• UC Berkeley, March 1, 2003

NSF
UC Berkeley Chess Vanderbilt University ISIS U
niversity of Memphis MSI
Foundations of Hybrid and Embedded Software
Systems
2
Our Goals

• Curriculum Council
• Mold the new curriculum
• Spread the new curriculum
• Summer Institute
• How to facilitate change
• Recruit participants
• Curriculum Change
• Advocates of the new curriculum
• Critics of the new curriculum
• Draft Mission Statement
• The Curriculum Council of the Chess Center
  provides strategic leadership and advocacy within
  the community for curriculum modernization and
  reform in engineering and computer science. It
  serves as the principal liaison between the Chess
  Center and education professionals, and will
  strive to facilitate and promote improvements in
  the way systems science is taught.

3
Chess Center for Hybrid and Embedded Software
Systems

• Berkeley Component in the NSF/ITR project
• College of Engineering, University of California
  at Berkeley

Board of Directors Tom Henzinger, tah_at_eecs.berkel
ey.edu Edward A. Lee, eal_at_eecs.berkeley.edu Albe
rto Sangiovanni-Vincentelli, alberto_at_eecs.berkeley
.edu Shankar Sastry, sastry_at_eecs.berkeley.edu
Other key faculty Alex Aiken, aiken_at_eecs.berkeley
.edu Dave Auslander, dma_at_me.berkeley.edu Karl He
drick, khedrick_at_me.berkeley.edu
Kurt Keutzer, keutzer_at_eecs.berkeley.edu
George Necula, necula_at_eecs.berkeley.edu
Masayoshi Tomizuka, tomizuka_at_me.berkeley.edu
Pravin Varaiya, varaiya_at_eecs.berkeley.edu
4
Hybrid Embedded Software Systems

• Computational systems
• but not first-and-foremost a computer
• Integral with physical processes
• sensors, actuators
• Reactive
• at the speed of the environment
• Heterogeneous
• hardware/software, mixed architectures
• Networked
• adaptive software, shared data, resource discovery

5
Mission of Chess

• To provide an environment for graduate research
  on the design issues necessary for supporting
  next-generation embedded software systems.
  
• Model-based design
• Tool-supported methodologies
• For
• Real-time
• Fault-tolerant
• Robust
• Secure
• Heterogeneous
• Distributed
• Software

The fate of computers lacking interaction with
physical processes.
6
Current Software Engineering Methods are
Inadequate

• In software today
• Accidental complexity
• Unpredictability
• Uncomposability
• Brittleness
• Not good at interacting with the real world
• We are stuck in a lines of code view of SW
• Scaling this by putting more people onto it
• Lose understanding of the system as a whole
• Separation of specification from construction

7
French Guyana, June 4, 1996
800 million embedded software failure
8
Mars, December 3, 1999
Crashed due to uninitialized variable
9
4 billion development effort
40-50 system integration validation cost
10
Embedded Software Architecture Today
11
Embedded Software Architecture Tomorrow
12
The Goal

• To create a modern computational systems theory
  and systems design practice with
  
• Concurrency
• Composability
• Time
• Hierarchy
• Heterogeneity
• Resource constraints
• Verifiability
• Understandability

13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
How Safe is Our Real-Time Software?
18
Another Traditional Systems Science -
Computation, Languages, and Semantics
Everything computable can be given by a
terminating sequential program.
Functions on bit patterns Time is irrelevant
Non-terminating programs are defective
sequence
f States ? States
States Bits
results state out
19
Current fashion Pay Attention to
Non-functional properties

• Time
• Security
• Fault tolerance
• Power consumption
• Memory management
• But the formulation of the question is very
  telling

20
What about real time?
21
Processes and Process Calculi
Infinite sequences of state transformations are
called processes or threads
Various messaging protocols lead to various
formalisms.
In prevailing software practice, processes are
sequences of external interactions (total
orders). And messaging protocols are combined i
n ad hoc ways.
incoming message
outgoing message
22
Interacting Processes Concurrency as
Afterthought
Software realizing these interactions is written
at a very low level (semaphores and mutexes).
Very hard to get it right.
stalled by precedence
timing dependence
stalled for rendezvous
23
Interacting Processes Not Compositional
An aggregation of processes is not a process (a
total order of external interactions). What is
it? Many software failures are due to this ill-
defined composition.
24
Compositionality
Non-compositional formalisms lead to very awkward
architectures.
25
Real-Time Multitasking?
26
Promising Alternatives

• Synchronous languages (e.g. Esterel)
• Time-driven languages (e.g. Giotto)
• Hybrid systems
• Timed process networks
• Discrete-event formalisms
• Timed CSP
• We are working on interface theories and meta
  models that express dynamic properties of
  components, including timing.

27
(No Transcript)
28
Education Changes The Starting Point
Berkeley has a required sophomore course that
addresses mathematical modeling of signals and
systems from a computational perspective.
The web page at the right illustrates a broad
view of feedback, where the behavior is a fixed
point solution to a set of equations. This view
covers both traditional continuous feedback and
discrete-event systems.
29
The Roots of Signals and Systems

• Circuit theory
• Continuous-time
• Calculus-based

Major models Frequency domain Linear time-invari
ant systems
Feedback
30
Changes in Content

• Signal
• used to be voltage over time
• now may be discrete messages
• State
• used to be the variables of a differential
  equation
  
• now may be a process continuation in a
  transition system
  
• System
• used to be linear time invariant transfer
  function
  
• now may be Turing-complete computation engine

31
Changes in Intellectual Scaffolding

• Fundamental limits
• used to be thermal noise, the speed of light
• now may be chaos, computability, complexity
• Mathematics
• used to be calculus, differential equations
• now may be mathematical logic, topology, set
  theory, partial orders
  
• Building blocks
• used to be capacitors, resistors, transistors,
  gates, op amps
  
• now may be microcontrollers, DSP cores,
  algorithms, software components

32
Themes of the Course

• The connection between imperative and declarative
  descriptions of signals and systems.
  
• The use of sets and functions as a universal
  language for declarative descriptions of signals
  and systems.
  
• State machines and frequency domain analysis as
  complementary tools for designing and analyzing
  signals and systems.
  
• Early and often discussion of applications.

Brain response when seeing a discrete Fourier
series.
33
Outline

• 1. Signals
• 2. Systems
• 3. State
• 4. Determinism
• 5. Composition
• 6. Linearity
• 7. Hybrid Systems
• 8. Freq Domain
• 9. Freq Response
• 10. LTI Systems
• 11. Filtering
• 12. Transforms
• 13. Sampling
• 14. Review
• 15. Examples

34
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
35
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
36
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
37
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
38
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
39
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
40
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
41
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
42
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
43
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
44
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
45
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
46
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
47
Outline
1. Signals 2. Systems 3. State 4. Determinism
5. Composition 6. Linearity 7. Hybrid Systems 8
. Freq Domain 9. Freq Response 10. LTI Systems
11. Filtering 12. Transforms 13. Sampling 14.
Review 15. Examples
48
Lab Highlights

• Use Matlab and Simulink
• Tightly integrated
• Relating declarative imperative models

49
Differences from Tradition

• No circuits
• More discrete-time, some continuous-time
• Broader than LTI systems
• Unifying sets-and-functions framework
• Emphasis on applications
• Many applets and demos
• Tightly integrated software lab

Text (Lee Varaiya) and website available.
50
Bottom-Up or Top-Down?
Top-down - applications first - derive the foun
dations
Bottom-up - foundations first - derive the appl
ications
51
Notation

• Sets and functions
• Sound Reals ? Reals
• DigitalSound Ints ? Reals
• Sampler Reals ? Reals ? Ints ? Reals
• Our notation unifies
• discrete and continuous time
• event sequences
• images and video, digital and analog
• spatiotemporal models

52
Problems with Standard Notation

• The form of the argument defines the domain
• x(n) is discrete-time, x(t) is continuous-time.
• x(n) x(nT)? Yes, but
• X(jw) X(s) when jw s
• X(ejw) X(z) when z ejw
• X(ejw) X(jw) when ejw jw? No.
• x(n) is a function
• y(n) x(n) h(n)
• y(n-N) x(n-N)h(n-N)? No.

53
Using the New Notation

• Discrete-time Convolution
• Shorthand
• Definition

54
The Lab Component

• Introduce applications before the theory fully
  supports them.
  
• Connection between a mathematical (declarative)
  and a computational (imper-ative) view of
  systems.
  
• Use of software to perform operations that could
  not possibly be done by hand, operations on real
  signals such as sounds and images.

55
Technology

• Matlab
• imperative programming language
• finite signals (matrices and vectors)
• discrete signals
• Simulink
• block diagram language
• infinite signals
• continuous-time semantics
• We view these as complementary.

56
Organization

• 3 hour scheduled sessions, once a week
• 11 labs in 15 week session
• 1 organizational, 1 technological, 2 review
  sessions
  
• In-lab section
• takes about 1 hour, completed with signoff
• Independent section
• takes 1-6 hours, completed with a report
• Tightly synchronized with the lectures.

57
Lab 1 (audio)

• Arrays and vectorization in Matlab
• Construct finite sound signals

58
Lab 2 (images)

• Colormaps

59
Lab 2 (images)

• Sinusoidal images

60
Lab 2 (images)

• Images
• blurring (averaging)
• edge detection (first-order differences)

61
Lab 3 (state)

• State machines
• Tamagotchi virtual pet

62
Lab4 (feedback control)

• Closed loop control of the virtual pet

63
Lab 5 (difference equations)
64
Lab 6 (differential equations)
65
Lab 7 (spectrum)
66
Lab 7 (spectrum)
67
Lab 8 (comb filters)
68
Lab 9 (plucked string model)
69
Lab 10 (modulation/demodulation)
70
Lab 10 (modulation/demodulation)
71
Lab 10 (modulation/demodulation)
72
Lab 10 (modulation/demodulation)
73
Lab 10 (modulation/demodulation)
Note that the entire exercise remains in the
audio band.
74
Lab 11 (sampling aliasing)
75
Areas for Improvement

• Extensive capabilities of the tools can be
  intimidating.
  
• Requires some programming background.
• On-line help for Matlab is much better than for
  Simulink.
  
• Simulinks discrete-time models are
  continuous-time models in disguise.
  

76
Analysis of Spring 2000 Offering

• Class standing had little effect on performance.
• On average, the GPA of students was neither
  lowered nor raised by this class.
  
• Students who attend lecture do better than those
  that dont.
  
• Taking at least one of Math 53, 54, or 55 helps
  by about ½ grade level.
  
• Taking Math 54 (linear algebra differential
  eqs.) helps by about 1 grade level (e.g. B to
  B).
  
• Computing classes have little effect on
  performance.

77
Distribution by Class Standing
78
Effect of Class Standing
80 and above As 63 and above Bs 62 and below
Cs 176 of the 227 students responded (the b
etter ones).
79
Effect of Showing Up

• Students who answered the survey were those that
  showed up for the second to last lab.
  
• The mean for those who responded was 78, vs. 65
  for those who did not respond (two grades, e.g. B
  to A-).
  
• The standard deviation is much higher for those
  who did not respond.
  
• A t-test on the means shows the data are
  statistically very significant.
  
• We conclude that the respondents to the survey
  do not represent a random sample from the class,
  but rather represent the diligent subset.

80
Attendance in Class vs. Score
Attendance is measured by presence for pop
quizzes, of which there were five.
81
Effect on GPA
On average, students GPA was not affected by
this class.
82
Student Opinion on Prerequisites
series
linear algebra
83
Fundamental Change Requires Your Help

• Curriculum Council
• Mold the new curriculum
• Spread the new curriculum
• Summer Institute
• How to facilitate change
• Recruit participants
• Curriculum Change
• Advocates of the new curriculum
• Critics of the new curriculum

84
Conclusion

• A new system science that is at once physical
  and computational is emerging.
  
• It will form the foundation for our
  understanding of computational systems that
  engage the physical world.
  
• And it will change how we teach, research and
  engineer systems.
  
• The engineering curriculum has to change