Goal/Rule Synthesis Pipeline for Autonomous Driving

This repository demonstrates a Goal/Rule Synthesis Pipeline applied to autonomous driving scenarios. The pipeline decomposes high-level goals into logically consistent causes, consolidates them semantically, translates them into formal symbolic representations, and evaluates necessary and sufficient conditions for achieving the goal.

---

Pipeline Overview

The pipeline consists of four main stages:

1. Decomposition to Causes

• Objective: Generate candidate causes that explain how a high-level goal (effect) may occur.

2. Semantic Consolidation

• Objective: Merge overlapping causes that refer to the same underlying concept.
• Benefit: Produces a unique set of causes for further analysis.

3. Symbolic Translation

• Objective: Convert natural language causes into formal symbolic (FOL) form for structured reasoning.

4. Evaluation of Necessary & Sufficient Causes

• Objective: Identify which causes are necessary, sufficient, or both for achieving the effect.

Steps:

1. Individual Necessity Evaluation

   • Assesses each cause independently.
   • Provides justification referencing safety or legal constraints.

2. Subset Necessity Evaluation

   • Identifies causes that are necessary only in combination.

3. Minimal Sufficient Set Evaluation

   • Identifies subsets of causes that are sufficient to achieve the effect.
   • Minimal sufficient sets are combined into a necessary-and-sufficient set, capturing all pathways to the effect.

---

Quick Start

Setup

git clone <repository-url>
# Create a Python virtual environment named "venv"
python3 -m venv venv
# Activate the virtual environment on macOS / Linux
source venv/bin/activate
# Activate the virtual environment on Windows (PowerShell)
.\venv\Scripts\Activate.ps1
# Install required packages
pip install -r requirements.txt
# Create a .env file in the project root and add your OpenApi Key:
OPENAI_API_KEY=your_openai_api_key_here
# Rub the script by running:
python3 src/pipeline.py