[CAE x AI] Building a Fully Autonomous Healing AI (Part 1): The Concept of the 'ATL' Neural System

• Part 1: Concept and Environment Setup This article
• Part 2: Core Logic Implementation
• Part 3: Self-Evolution and Metacognition

Introduction: Crossing the "Wall" of CAE Automation

CAE pre-processing, especially geometry correction (healing), is a task that even skilled engineers struggle with. When attempting to automate this, conventional scripts inevitably stop at "unexpected shapes."

Therefore, we are launching a project to build a "fully autonomous AI (God AI) that turns failure into assets and grows on its own" from scratch on a low-spec Windows PC.

In Part 1, we will establish the concept of the "ATL (Autonomous Trace Logger)," which forms the core of this AI, and set up the development environment.

The Architecture of "God AI"

The difference between a mere automation tool and the "autonomous AI" we are creating lies in whether it possesses a central nervous system (ATL).

Conventional Automation

• State → Rules → Execution
• Stops if it fails. It does not record why it failed.

The "Fully Autonomous ANSA Healing AI" We Are Building

This structure allows the AI itself to record "options it thought about but didn't choose" and "reasons for failure" as logs, using them to rewrite its own skills.

┌─────────────────────────┐
│        God Core         │
│  (Orchestrator Brain)   │
└─────────┬───────────────┘
          ↓
┌─────────────────────────┐
│   Healing Planner       │  ← Generates multiple plans (deliberation)
└─────────┬───────────────┘
          ↓
┌─────────────────────────┐
│   Skill Executor        │  ← ANSA operations (hands and feet)
└─────────┬───────────────┘
          ↓
┌─────────────────────────┐
│   Geometry Evaluator    │  ← Success/failure and quality assessment (eyes)
└─────────┬───────────────┘
          ↓
┌─────────────────────────┐
│ ★ ATL (Self Trace Log)  │  ← Central nervous system (accumulation of experience)
└─────────┬───────────────┘
          ↓
┌─────────────────────────┐
│ Skill Confidence DB     │  ← Growing database
└─────────┬───────────────┘

This ATL (Autonomous Trace Logger) is the prototype of "consciousness" in our AI.

Technical Stack and Requirements
You do not need a high-spec machine. The design is intended to run in the following environment:

OS: Windows 10 / 11

Language: Python 3.10

LLM: OpenAI API (or Azure OpenAI) *Can run with a mock since the focus is on logic

DB: SQLite (lightweight, no configuration required)

UI: Streamlit (God's perspective tool)

Hands-on: Environment Setup
Now, let's start the actual construction. Open Windows PowerShell and execute (copy and paste) the following commands in order. This will complete the "vessel" for your project.

Step 1: Preparing the Python Environment
First, install Python 3.10 and create a virtual environment.

# Installing Python (if not installed)
winget install Python.Python.3.10

# Creating and activating the virtual environment
python -m venv venv
.\venv\Scripts\activate

Step 2: Installing Libraries
Install the minimum necessary libraries.

pip install pyyaml streamlit requests sqlite-utils

Step 3: Creating the Directory Structure
Create the folder structure that will serve as the "God's vessel" all at once.

mkdir ansa_god_ai
cd ansa_god_ai
mkdir bootstrap core atl skills skill_db ui mcp

Everything is now ready. Your folder structure should look like this:

ansa_god_ai/
├─ bootstrap/   (For DB initialization)
├─ core/        (Brain functions: Plan, Eval)
├─ atl/         (Central nervous system: Log)
├─ skills/      (Skill definition files)
├─ skill_db/    (Memory DB)
├─ ui/          (For visualization)
└─ mcp/         (For external tool manipulation)

Next Time
In the next article, we will finally start implementing the Python code. We will implement the "ATL," the central nervous system, and proceed to the point where the AI can autonomously create plans, execute them, evaluate them, and close the "autonomous loop" by saving logs.

Stay tuned.