NEUROFORGE // Walking Test

EVOLUTIONEVOLVING

Algorithm Algorithm documentation ↗ Test Population 18 Elite count 4 Mutation / sigma 0.12 Max generation time 60s

Endless while leader progresses

HEIGHT LIMIT 1.20m

Core minimum height 1.20m

STALL LIMIT 3.0s

Leader stall time 3.0s

Progress reset step 0.03m

MINIMUM SPEED 0.08m/s

Eliminate walkers below minimum speed Minimum forward speed 0.08m/s

Low-speed grace 2.0s

Height, stall, and speed are independent test criteria. Height and speed can eliminate individual walkers; the stall rule ends the generation when the furthest surviving walker stops progressing.

COURSEPROGRESSIVE

Progressive terrain after editable path Difficulty growth 1.00×

Physical climb-over obstacles Obstacle spacing 14.0m

Obstacle scale 1.00×

Beyond your editable path, hill amplitude/grade and obstacle size increase with distance. All variants in a generation see the same deterministic course.

PHYSICSREV 1

Gravity 9.81 Body damping 0.04 Ground friction 0.90 Position iterations 8 Below-line grace 0.08s

DISPLAY / SPEEDBALANCED

Compute budget 10ms Visible variants 6 Simulation rate 1×

FIGUREBiped Walker

Saved figures can include the current leading gene and be loaded onto any terrain.

RESETDESTRUCTIVE

Keeps the figure and terrain. Clears neural weights, optimizer state, history, lineage and permanent variant numbering.

GENERATION 1EVOLVING

LEADER#—

REWARD—

CORE HEIGHT0.00m

SPEED0.00m/s

STALL0.0s

FORWARD0.00m

STEPS/S0

NEURAL POLICY · SENSOR / CORRECTION#—

SENSORCORRECTION IMPACT

DEVICE / RUNTIMEBALANCED

LOGICAL CPU—

DEVICE RAM—

JS HEAP—

FRAME LOAD*—

FPS—

GENOME—

POP MEMORY—

WEBGPU—

*Page simulation duty estimate, not OS-wide CPU utilization.

EVOLUTION HISTORY—

POPULATION0 variants

SEEDNEWCHILDCARRYES/CEM