Phenotype:
body
type, x, y, z, width, height, depth of block 1 type, x, y, z, width, height, depth of block 3 |
type, x, y, z, width, height, depth of block 2 .... type, x, y, z, width, height, depth of block 9 |
brain
sensor1 excites neuron 1, weight 5 neuron1 excites neuron 6, weight 6 sensor2 excites neuron 2, weight 4 neuron2 excites neuron 4, weight 3 |
neuron2 inhibits neuron 5, weight 0 sensor3 excites neuron 3, weight 8 .... neuron 3 excites neuron 9, weight 4, output of neuron 9 to muscle4 |
Genotype:
Chromosome 1
11 |
11010110 |
10101101 |
10101110 |
10011010 |
01101010 |
01101010 |
... |
10001010 |
10001010 |
10001010 |
type1 |
xpos 1 |
ypos 1 |
zpos 1 |
width 1 |
height 1 |
depth 1 |
... |
width 9 |
height 9 |
depth 9 |
Chromosome 2
0000 |
10 |
1001 |
1 |
0110 |
1 |
... |
1110 |
neuron1 marker |
neuron1 type (in/intl/out) |
neuron1 link1 |
neuron1 link1 type (ex/inhib) |
neuron1 link2 |
neuron1 link1 type (ex/inhib) |
... |
neuron9 link2 |
Figure. Evolutionary Artificial Life Form.
To illustrate this type of Evolutionary Design, we return to the 'table' example one final time. However, instead of a static table, we now require a robot / virtual creature / animat capable of carrying objects around - a robot waitor, perhaps. The figure shows the dual nature of these designs: Evolutionary AL-Forms typically involves the evolution of the form (or some aspect of the form) and the brain. In this example, the form, or 'body' is defined by a collection of variable-sized blocks (which may be 'sensors', 'body' or 'muscles'), the 'brain' is defined by a network of neurons which receive input from 'sensory blocks' and produce output to the 'muscle blocks'. Each part of the phenotype is encoded as a variable-length chromosome in the genotype. The fitness function judges phenotypes on their ability to move whilst keeping the flat upper surface level. Over time, evolution will co-evolve both chromosomes in individuals to generate a virtual creature capable of supporting objects and movement in a virtual world.