Neocs Jun 2026

We implement the NEAT algorithm to manage the evolution of the ANNs. This allows NEOCS to start with minimal structure (simple networks) and complexify only when necessary, keeping the computational overhead low for embedded systems.

At the core of NEOCS lies a population of Artificial Neural Networks (ANNs). Unlike standard DRL, which optimizes a single network, NEOCS maintains a population of $N$ distinct controllers. Each controller (genome) encodes: We implement the NEAT algorithm to manage the

Neuro-evolution, Autonomous Systems, Genetic Algorithms, Adaptive Control, NEAT. Unlike standard DRL, which optimizes a single network,

| Metric | Standard DRL (PPO) | NEOCS (Proposed) | | :--- | :--- | :--- | | Avg. Success Rate (Static) | 98% | 95% | | Avg. Success Rate (Dynamic) | 62% | | | Recovery Time after Sensor Failure | High (Retrain required) | Low (Instant switch) | Success Rate (Static) | 98% | 95% | | Avg