README.md

metadata

title: Vehicle Diagnostics Agent
emoji: 🚗
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false
license: mit
short_description: Multi-Agent AI System for Predictive Vehicle Diagnostics
tags:
  - anomaly-detection
  - lstm
  - pytorch
  - langchain
  - langgraph
  - multi-agent
  - vehicle-diagnostics
  - time-series

🚗 Vehicle Diagnostics Agent

Multi-Agent AI System for Predictive Vehicle Diagnostics

This is a production-ready multi-agent AI system that analyzes vehicle sensor data to detect anomalies, identify root causes, and provide actionable maintenance recommendations.

🎯 Key Features

• 🔍 Anomaly Detection: LSTM-based neural network with 99.53% validation accuracy
• 🔬 Root Cause Analysis: Identifies underlying issues with OBD-II fault code mapping
• 🔧 Maintenance Recommendations: Provides cost estimates and prioritized action plans
• 📊 Interactive Visualizations: Real-time anomaly detection charts
• 📋 Natural Language Reports: Easy-to-understand summaries for vehicle owners

🏗️ System Architecture

The system employs a multi-agent architecture orchestrated by LangGraph:

1. Data Ingestion Agent - Loads and prepares vehicle sensor data
2. Anomaly Detection Agent - LSTM neural network for pattern detection
3. Root Cause Analysis Agent - Fault pattern matching and correlation analysis
4. Maintenance Recommendation Agent - Cost estimation and action planning
5. Report Generation Agent - Comprehensive diagnostic reports

🚀 Technology Stack

• ML Framework: PyTorch (LSTM-based time-series anomaly detection)
• Orchestration: LangGraph for multi-agent coordination
• Frontend: Gradio for interactive UI
• Data Processing: Pandas, NumPy, Scikit-learn
• Visualization: Plotly

📊 Model Performance

• Validation Accuracy: 99.53%
• Training Loss: 0.0003 (final epoch)
• Validation Loss: 0.0409 (best)
• Dataset: 50,000 records from 100 vehicles
• Features: 60+ engineered features from 14 sensor measurements

🎮 How to Use

1. Select a Vehicle: Choose from available vehicle IDs
2. Set Reading Count: Specify how many recent readings to analyze (default: 200)
3. Run Diagnostic: Click the diagnostic button to analyze
4. Review Results: View anomaly detection, root cause analysis, and maintenance recommendations

📈 Dataset

The system analyzes synthetic vehicle sensor data including:

• Engine temperature, RPM, speed
• Battery voltage and health
• Oil and fuel pressure
• Tire pressure (all four wheels)
• Vibration levels
• Coolant temperature
• And more...

🔬 Technical Details

Anomaly Detection Model:

• Architecture: 2-layer LSTM with 64 hidden units
• Input: Sequences of 10 timesteps with 60 features
• Output: Binary classification (normal/anomaly)
• Training: 31,570 sequences on GPU

Root Cause Analysis:

• 8 fault pattern definitions
• Sensor correlation analysis
• Confidence scoring
• OBD-II fault code mapping (P-codes, C-codes)

📝 License

MIT License - See LICENSE file for details

🔗 Links

• GitHub: VehicleDiagnosticsAgent
• Documentation: Full project documentation available in the repository

👨‍💻 Author

Built with ❤️ by Saad Mann

---

Note: This is a demonstration system using synthetic data. For production use with real vehicles, integration with actual OBD-II devices would be required.