Genetic Algorithm — 2D Obstacle Course Navigator

Population Setup

Population size 80

Mutation rate 8%

Evolution speed 5

Live State

Generation

Reached goal

Best fitness

—

Avg fitness

—

Diversity

—

Crash rate

—

How it works

A genuine genetic algorithm, not a decorative animation. Each chromosome is a sequence of 70 genes, one steering value per step (a real number in [−1, 1] that bends the heading by up to 32° that step). Decoding a chromosome simulates a robot walking that steering sequence from the start marker across the obstacle course; the walk stops early if it hits an obstacle (a crash).

fitness = progress toward goal (−1..1)
        − 0.4 if it crashed
        + up to 0.2 bonus for reaching goal quickly

Selection — tournament of 5: pick 5 random chromosomes, keep the fittest.
Crossover — single random cut point splices two parents' gene arrays into one child.
Mutation — each gene independently has a chance (the mutation-rate slider) to be replaced by a fresh random steering value.
Elitism — the two fittest chromosomes survive unchanged into the next generation.

Faint grey lines are every path this generation; the bright green line is the fittest path found so far. Watch the fitness chart climb as selection pressure accumulates useful steering genes over generations.

Click: add obstacle · Right-click: remove

All paths this generation Best path so far Obstacle

⚠ Couldn't load the 2D engineCanvas 2D context failed to initialize. Try reloading.