P. Marchal

1
Embryological ElectronicsFirst NASA/DoD
Workshop onEvolvable Hardware

• P. Marchal
• Centre Suisse d'Electronique et de Microtechnique
  SA
• Jaquet-Droz 1
• CH-2007 Neuchâtel
• pierre.marchal_at_csem.ch
• http//www.csem.ch

2
Summary

• Introduction to Bio-inspired Systems
• Embryological Electronics
• What is presently available ?
• Open Avenues for Evolvable Hardware
• Conclusion

3
Introduction to Bio-inspired Systems

• What is bio-inspiration?
• Building complex systems
• Genome-based design

4
Bio-inspiration?

• Nature has acquired a strong experience in
  complex system design
• 3-billion years of R D
• Powerful constructions (built and maintained)
• longer than hundreds years (animal life)
• longer than thousands years (plant life)
• Adapting and Evolving solutions
• personal modification is adaptation or learning
• inherited modification is evolution

5
3-billion years shrinked into 1 year
6
This is NOT bio-inspiration
7
Fields of Bio-inspiration
evolution
healing
perceptron
Neural nets

actuators
self- structuration
perception
sensors
Neural nets
Genetic algo
mechanics
VLSI
optics
Artificial life
algorithms
8
Building Complex Systems 1.- Engineers approach
???
9
Building Complex Systems2.- Natures Approach (1)
10
Building Complex Systems2.- Natures Approach (2)
MUSCLE CELL
LYMPHOCYTES
SPERMATOZOON
LEUCOCYTE
OSTEOCYTE
RED CELLS
10 hours, 30'000 cells
FIBROPLAST
NERVE CELL
11
Field Programmable Gate Arrays
12
Von Neumann Contribution

• He proposed that the production of an automaton
  by another one should be composed of two phases
• information is once read and copied
  (transcription)
• information is then read and interpreted
  (translation)
• He conceived a self-reproducing automaton

13
Self-structuring VLSI(genome-based design)
14
Biodule (biological-like module)
15
Embryological Electronics

• Reproduction
• Adaptation
• Evolution

16
No reproduction apparatus
17
A cell composed of proto-cells

• The silicon cell is composed of

• Genome memory

• Address computation

• Functional cell

• Failure handling

18
Nucleus-like proto-cell

• Its function is to
• store the genogram (set of bit-strings - genes
  - that describes the functionality of the silicon
  cell)
• transmit a copy of the genogram to neighbouring
  cells
• boot the address computation

19
Storing Process
.
20
Each Nucleus stores its own copy
.
?
3
0
6
0
0
0
3

0
2
0

0
2
5
2
3
0
0
1
2
0

3

2
1
0
0
4
0
0

0
1
3

2
1
0
0
0
0
1
0
0


2
2
0
0
0
21
Gradient-like control proto-cell

• Its function is to
• compute the local address (row column
  coordinates)
• transmit a copy of the local address to the
  neighbouring cells
• boot the differentiation process (gene expression)

22
Local Address Computation
23
Continuous Gradient
5,4
0
,
1

?
1,1
24
Repeating Structures
1,1
0
,
1
?
1,1
25
Cell Differentiation

• the local address is used to pick up, out of the
  genogram memory, the gene corresponding to that
  location
• the gradient like process enables cell
  differentiation

26
Differentiation Process
.
27
Functional Cell

• Its function is to
• load the programmable bit-string of the FPGA
  proto-cell with the local gene
• realise a part of the logical function
  (distributed among the circuit area)
• transmit convenient information with the
  appropriate neighbours

28
Family of Cells
INTER
CONNECTION
PART
LOCAL GENE
FIELD PROGRAMME
FUNCTIONAL
PART
29
Immune-like Proto-Cell

• Its function is to
• determine the faulty behaviour of a cell, if any,
  and the severity of the fault
• transmit the internal state (faulty or not) to
  the neighbours
• boot the healing phase (restart address
  computation) if a fault has occurred

30
Healing Process
3
3
2
0
6
0
0
0
0
0
3
2
2
0
2
3
3
3
2
0
0
0
0
0
3
2
2
0
2
3
3
3
3
1
3
2
3
1
3
2
3
2
0
2
3
4
0
0
2
3
0
0
0
2
2
0
0
0
2
3
4
0
0
2
3
0
0
0
2
2
0
0
1
2
2
2
3
2
1
2
2
2
3
2
1
0
1
0
0
1
0
0
1
0
1
0
1
0
5
0
0
1
0
0
1
0
0
1
0
1
0
1
0
5
0
2
1
3
1
2
1
3
1
1
1
1
1
Y
1
2
3
4
5
6
X
31
Healing Process
32
What is presently available ?

• A family of self-structuring circuits

33
A family of self-structuring circuits
34
DMUXTREE
35
GenomIC
36
Field Programmable Processor Array (FPPA)
37
Open Avenues for Evolvable Hardware

• Applications
• Adaptation
• Evolution

38
Applications

• Self-structuring and self-repairing VLSI should
  be considered in situations where changing and/or
  repairing is
• too difficult (under sea exploration)
• too dangerous (nuclear exposition)
• too expensive (deep space exploration)
• too risky (human life is in danger)
• and functionality should be conserved in presence
  of defects, radiations or wear out
• Emerging applications in automotive (WINS project)

39
Adaptation

• Reconfiguration is based on an event differing
  from the occurrence of a fault
• Physical event adaptation
• swing of power lines
• shift in temperature
• Informational event adaptation
• change of signals bandwidth
• object oriented processing

40
Evolution

• Development is based on a description of the
  structure stored in a genome
• Use the genetic algorithm and genetic programming
  techniques to evolve such systems
• Two levels of description may be considered
• high level description ? evolution for synthesis
• low level description ? evolution for adaptation

41
Conlusion
42
Parallelism, morphism and adaptation

• Massive parallelism
• Multicellular organization
• Morphism
• Configurable hardware
• Adaptation
• Upgradable software
• Reconfigurable hardware

43
To conclude

• We have investigated this research domain
• We have acquired the know-how to address a large
  amount of questions related to fault tolerance as
  well as evolvable hardware
• We have the mastery of the technology
• We have patents on it
• We are ready to answer any question regarding
  this field

44
(No Transcript)