Observe artificial digital neural architectures adaptive selection and evolutionary mechanics in real-time.
Overview
The Interactive Evolution Simulator is a simplified platform demonstrating selection pressures and adaptive optimization. Each digital creature is generated with an isolated connection genome forming a responsive neural brain. Over sequential lifespans of 300 cycles, these creatures sense environmental signals and produce motor directional responses. Generation structures filter non-performing combinations, utilizing crossover reproduction paired with custom-adjustable mutation structures to navigate challenges without pre-programmed movement code.
How To Use
Sequential operational parameters include:
- Execution Control: Use the Start/Stop controls to regulate live simulation cycles. Reset wipes internal generation arrays back to baseline configurations.
- Select Scenarios: Alternate environment modes such as "Go East", "Barrier Challenge", or "Corner Seeker" to test adaptability against obstacles.
- Dynamic Parameter Configuration: Adjust population counts, connection mutation frequencies, speed thresholds, internal nodes, and genes. Changing structural properties (Population, Genome length, or Node structures) automatically instantiates a fresh evolutionary seed.
- Gene Inspection: Interacting directly with active creatures displaying inside the Visualizer pane pulls their raw connected gene wiring into the interactive Genome Registry module.
Technical Details
The simulation processes environmental inputs to produce actions using basic neural mechanics:
- Sensory Receivers (Inputs): X-Coordinate, Y-Coordinate, Generation Age, Border Proximity, and a cyclic internal rhythmic oscillator.
- Motor Channels (Outputs): Horizontal thrust (Move X), Vertical thrust (Move Y), Directional Persistence (Move Forward), and Random Drift (Move Random).
- Neural Calculations: Connected pathways read and process input layers through optional internal hidden neurons, running mathematical tanh functions to restrict values before determining action thresholds.
- Selective Breeding: Surviving structures are evaluated based on their scenario positioning, utilizing crossing midpoints to mix successful genomes combined with randomized connection mutations.
Future Directions
Planned roadmap directions focus on broadening environmental complexity and neural flexibility:
- Competitive Interaction Matrix: Integrating a "Kill" output node allowing the emergence of co-evolving predatory behaviors.
- Chemical Scent Trails: Introducing dynamic grid pheromones to allow collective swarm patterns and resource coordination.
- User-Configurable Barriers: Enabling drag-to-draw custom layouts on the canvas to evaluate adaptation against arbitrary obstacle formations.
- Cyclical Resource Dynamics: Simulating varying nutrient drops that require energy budget planning to shift focus from simple spatial goals to complex feeding strategies.
Resource Directory
Further resources on evolutionary computation and genetic algorithms include: