The Amorphous Computing Project

1
The Amorphous Computing Project

• Daniel Coore, PhD
• Dept. Mathematics and Computer Science
• University of the West Indies, Mona

2
The Amorphous Computing Concept

• Inspiration A collection of cells self-organize
  to form a single organism

What are the organising principles at work within
each of the individuals?
3
Motivation

• To control massively parallel systems
  (biological, chemical, quantum) that are likely
  to be built in the near future to do useful work
  e.g.
• Build molecular scale circuits,
• Act as smart drug dispensers
• Build 'smart' materials
• To surpass the physical limits of traditional
  computers by using low-cost parallelism

4
The Amorphous Computing Model

• Particle Constraints
• Asynchronous
• Irregularly located
• Local broadcast only
• Limited Resources
• Single program for all
• Only "reasonable" initial conditions

5
The Challenge

• How do we write programs to reliably produce
  coherent global behaviours under these
  constraints?

6
Outline

• Background
• ECOLI
• Simulator
• Current Directions

7
Background

• 1996 Simple Algorithms (Clubs, Gradients)
• 1997 Coordinate Systems (Coore, Nagpal)
• 1997 Tube Formation (Weiss)
• 1997 Amorphous Inverter (Coore)
• 1998 GPL Growing Point Language (Coore)
• 2000 OSL Origami Shape Language (Nagpal)
• 2003 Growth and Self-Assembly (Kondacs)
• 2003 Persistent Nodes (Beal)

8
Gradients

• Gradients Assign values that decrease with the
  distance from some specified processor.
• Solution Measure hop count from processor.
  Smooth values by averaging, if necessary.

9
Coordinate Systems
b
a

• x k
• y k
• r k
• ? k
• error

c
d
e
10
The Amorphous Inverter

• A CMOS Inverter Processors apply labels to
  themselves in the pattern of the CMOS layout of
  an inverter.

11
The Growing Point Language

• Language for describing topological relations
• In theory, any planar graph can be described
• Pattern descriptions are automatically converted
  into individual processor activities.

12
The Origami Shape Language

• OSL described patterns of folds as if they were
  origami commands.
• Compiled to agent level where local level
  actuations were effected to produce folds in
  correct sequence.

13
ECOLI

• Extensible Calculus of Local Interactions
• Idea cooking since 1997.
• Programming discipline used to develop GPL
• Formalised into a language at UWI, in 2000 with
  implementation of interpreter
• Event-Driven Language
• Events caused by messages arriving at processors,
  sent from neighbours

14
Example Finding Hop Counts

• Problem Have each agent discover its shortest
  hop count to a given set of agents.

(define-behaviour find-dist ((current-dist
10000)) (define-dict default (DIST (src-id
n) (guard (lt n current-dist)) (set
current-dist n) (send DIST src-id ( n
1))) (ACTIVATE () (set current-dist
0) (send DIST my-id 1))))
(init (set my-id (random 2000)) (if (lt
my-id10) (run find-dist)))
15
Simulator Screen Shot
16
Current Efforts

• Amorphous Protocols (Nation)
• Reliable Communication Through Redundant Paths
  (R. Anderson)
• ECOLI enhancement. (Holness)
• A Higher Level Pattern Description Language
  (Coore)
• Persistent Nodes (Beal)

17
Current Efforts

• Reliable Communication Through Redundant Paths
• Reimplementation of ECOLI
• Generalized Pattern Description Language
• Amorphous Medium Language (Beal)
• Generalised Software Abstractions

18
Things To Be Done

• Language for programming mobile agents.
• Language for self-organizing function (vs.
  pattern formation self-organizing form)
• Cooperative Computing agents self-organize
  resources (computation space) to solve a big
  problem.
• Powerful generalizable simulator for Amorphous
  Computing model to allow experiments with
  languages