| --- |
| language: en |
| license: apache-2.0 |
| library_name: jax |
| tags: |
| - deep-reinforcement-learning |
| - neuroevolution |
| - tpu |
| - autonomous-agents |
| - artificial-life |
| datasets: |
| - synthetic-navigation-sim |
| metrics: |
| - foraging-efficiency |
| - pathfinding-success |
| --- |
| |
| # DeepRL (Standard Edition) |
|
|
| **DeepRL** is a robust, neuroevolutionary navigation agent. Unlike the "Micro" tier, the **Standard Edition** is designed for higher-resolution environments (64x64) and possesses enough neural capacity for active pathfinding and "hunting" behaviors. |
|
|
| ## 📊 Model Profile |
| | Feature | Specification | |
| | :--- | :--- | |
| | **Model Architecture** | 3-Layer Deep CNN (64-Filter Width) | |
| | **Parameters** | ~45,000 (~180 KiB) | |
| | **Grid Resolution** | 64x64 Cellular Automata | |
| | **Input Channels** | 6 (Life, Food, Lava, 3x Signaling) | |
| | **Training Steps** | 48,000+ Generations | |
| | **Compute** | 16x Google Cloud TPU v5e (TRC Program) | |
|
|
| ## 🧬 Key Behavioral Upgrades |
| The Standard Edition introduces **Metabolism Decay**. Unlike previous iterations that could "camp" on a single food source, this model must actively navigate to new food patches to maintain its mass. |
|
|
| - **Advanced Pathfinding:** Capable of detecting food sources up to 32 pixels away. |
| - **Lava kiting:** Uses a wider 64-filter "vision" to weave through complex obstacle patterns. |
| - **Mass Regulation:** Emergent behavior allows the agent to stabilize its size relative to available resources. |
|
|
| ## 🚀 Deployment (Inference) |
| This model is optimized for modern consumer CPUs. It is designed to run smoothly at **30-60 FPS** on mid-tier hardware. |
|
|
| ### Hardware Target: "Desktop Tier" |
| - **CPU:** Intel Core i5/i7 (8th Gen+) or AMD Ryzen 5/7 |
| - **RAM:** 8GB - 16GB |
| - **Platform:** Works with any NumPy-compatible environment (Python 3.8+). |
|
|
| ## 🛠️ Loading the DNA |
| The weights are stored as a ragged NumPy object array to preserve the specific convolutional shapes required by the JAX/Keras training pipeline. |
|
|
| ```python |
| import numpy as np |
| |
| # Load the Standard DNA payload (Apache 2.0 Licensed) |
| dna = np.load("DeepRL_Final_Gen48k.npy", allow_pickle=True) |
| |
| # Weight Indices: |
| # [0, 1] - Conv Block 1 (64 Filters, 3x3) |
| # [2, 3] - Conv Block 2 (64 Filters, 1x1) |
| # [4, 5] - Decision Output (6 Channels, 1x1) |
