Evolution in materio: Initial Experiments In Liquid Crystal

1
Evolution in materioInitial Experiments In
Liquid Crystal

• Simon Harding Julian Miller
• Department Of Electronics
• University Of York

2
What is evolution in materio?

• Exploiting the physical properties of a material
  to perform a task.
• Material must be configurable.
• Material should be able to modify signals.
• Evolutionary algorithm provides the configuration
  information.
• It is not optimisation its discovery!

3
Why do evolution in materio?

• Who says silicon is the right medium?
• Are clean/predictable responses ideal?
• Can we exploit a richer physics?
• What happens if we go beyond humanized models?
• What benefits do we get?
• Smaller? Cheaper? More fault tolerant?
• Do we overcome existing complexity ceilings?
• Is this a strategy for future system design?

4
An example Adrian Thompsons FPGA
5
(No Transcript)
6
Evolution with liquid crystal
7
Tone Discriminator Response
LOW
HIGH
LOW
HIGH
8
Tone Discriminator Response
LOW
HIGH
HIGH
LOW
9
Other frequencies
Easier
1st Input Frequency
Harder
2nd Input Frequency
10
LC Evolvable Motherboard
LC OUTPUT
INPUT SIGNAL
GND
CONFIGURATION VOLTAGES
11
Genetic Algorithm
0 1 2 6 3 7 2 4 8 3 0 1 7 3 5 1 3 5 .
0.2 -0.4 1.4 3.4 3.9
12
Liquid Crystal Display
13
Liquid Crystal
14
Simulation
15
Tone discriminator in simulation
LOW
HIGH
HIGH
LOW
16
Conclusions

• We have evolved the worlds first information
  processor in liquid crystal.
• It is relatively easy to evolve a tone
  discriminator and other functions.

17
The End
In conventional design the vast majority of
interactions that could possibly contribute to
the problem are deliberately excluded Michael
Conrad, 1988
18
Liquid Crystal
19
(No Transcript)