🧠 AI Technology Analysis & Autonomous Search-and-Rescue Drone

📌 Overview

This project presents a critical, evidence-based analysis of cutting-edge multimodal AI systems — including CLIP, Flamingo, BLIP-2, LLaVA, and MiniGPT-4 — evaluating their architecture, technical evolution, societal implications, and ethical challenges.

Additionally, the project includes an AI system design analysis of an Autonomous Search-and-Rescue Drone, applying intelligent agent theory and the PEAS framework to model autonomous decision-making in disaster environments.

---

🔬 Part 1 – Multimodal AI Technology Analysis

📖 Technologies Reviewed

• CLIP (Contrastive Language–Image Pretraining)
• Flamingo (Few-shot Visual-Language Model)
• BLIP-2 (Frozen Vision Encoder + LLM Bridge Architecture)
• LLaVA (Visual Instruction Tuning)
• MiniGPT-4 (Lightweight Multimodal Alignment)

---

🔍 Key Focus Areas

• Architecture design patterns in multimodal AI
• Modular alignment of frozen LLMs and vision encoders
• Instruction tuning and synthetic data trends
• Ethical risks (hallucination, bias, privacy concerns)
• Computational and scalability trade-offs
• Open research gaps in visual reasoning and safety

---

📄 Project Resources

• 📑 Full Report: AI_RESEARCH_PAPERS.pdf

---

🚁 Part 2 – Autonomous Search-and-Rescue Drone (AI System Analysis)

This section analyzes an autonomous drone operating in disaster-affected, GPS-denied environments.

---

🧩 Covered Concepts

• PEAS Framework (Performance, Environment, Actuators, Sensors)
• Six AI Environment Classifications:
  • Partially Observable
  • Stochastic
  • Sequential
  • Dynamic
  • Continuous
  • Multi-Agent
• Hybrid Model-Based + Learning Agent design
• Planning under uncertainty (POMDP-style reasoning)
• Sensor fusion (RGB + Thermal integration)
• Human-in-the-loop fallback mechanisms

---

🎯 Core Focus

Designing a robust AI agent capable of safe and reliable autonomous decision-making in:

• Partially observable environments
• Unpredictable disaster conditions
• Real-time dynamic scenarios
• Multi-agent rescue coordination settings

---

🛠 Skills Demonstrated

• AI architecture analysis
• Critical evaluation of academic research
• Ethical AI reasoning and governance awareness
• Intelligent agent modeling
• AI environment classification
• Autonomous system design under uncertainty
• Research synthesis and comparative analysis

---

📚 References

All academic sources are cited in APA format within the full report.
Primary sources include recent publications from arXiv, ICML, NeurIPS, and leading AI research institutions.

---

👩‍💻 Author

Fathima Safva Ovinakath Kammukkakath
BSc Computer Science (Level 5)
University of West London