Each organism is a sphere body with small appendage cubes; an 8-bit genome (shown as a strip of colored cubes: lit = 1, dark = 0) encodes leg length and stride pattern that determine how fast it crawls toward the central food source. Each generation, organisms are scored by distance covered, then selected to breed with bit-flip mutation.
P(select genome i) = fitness_i / Σ fitness_j (fitness-proportionate selection)
P(bit flips) = mutation_rate (per bit, independent)
- Mutation rate — probability each genome bit flips when copied to an offspring.
- Selection strength — how strongly fitness (distance-to-food traveled) biases which parents reproduce; 0 = random, 1 = only the fittest breed.
- Population size — number of organisms simulated per generation in the dish.
- Play / Step gen — auto-run generations continuously or advance exactly one at a time.
This mirrors real genetic algorithms used in evolutionary robotics, where locomotion controllers are bred rather than hand-programmed.