AI-HMT: Enhancing FPV Drone Targeting Precision

AI-Enabled Human-Machine Teaming for Enhancing FPV Drone Targeting Precision

Dissertation Presentation | Prepared by: Simranjeet Singh Raina

• Introduction & Context: Background on FPV drones and the Ukraine conflict.
• Problem Definition & Objectives: Core research premise and objectives.
• Technical Analysis: Technical analysis of FPV capabilities and limitations.
• Proposed Methodology: Proposed AI architectures and methodologies.
• Results & Conclusion: Expected outcomes and future directions.

• Rise of First-Person View (FPV) drones as tactical asymmetric weapons.
• Extensive combat use in the Ukraine conflict showing low-cost high-impact capability.
• Evolution from civilian hobby equipment to critical battlefield asset.
• Current reliance on pilot skill for high-precision targeting.

---

Photo by Simon Fitall on Unsplash

• Problem Statement: Manual pilot operation in high-interference environments reduces targeting precision.
• Research Gap: Lack of robust, real-time, low-latency AI-assisted targeting in low-cost FPV hardware.
• Aim: Develop a human-machine teaming (HMT) architecture for enhanced precision.
• Hypothesis: Integrating lightweight AI models on edge hardware will significantly improve target acquisition accuracy compared to traditional manual control.

• FPV Drone: Agile, low cost, high mobility.
• Conventional Limitations: Manual control is highly susceptible to electromagnetic interference (EMI) and fatigue.
• The Need for AI: AI-enabled HMT enables target tracking, path prediction, and autonomous stabilization in degraded visual environments.

---

Photo by Sergey Koznov on Unsplash

• YOLO (You Only Look Once): Real-time, fast object detection on edge devices.
• CNN-LSTM: Temporal sequence processing for tracking moving targets.
• Vision Transformers (ViT): Feature extraction and attention mechanisms for complex environments.
• Advantages: Increased hit rate, reduced pilot fatigue, mission success in high-threat zones.
• Challenges: Power consumption, latency in inference, model size vs. memory constraints.

---

Photo by Conny Schneider on Unsplash

• Operation Sindoor: Strategic focus on rapid, indigenous drone capabilities.
• Goal: Scaling to 100,000+ tactical drone platforms.
• Imperative: Development of autonomous, resilient AI-HMT systems for large-scale operations.
• Literature Insights: Studies from the Ukraine theater indicate that AI-assisted target acquisition can increase effective hit rates by up to 35% in high-interference scenarios.

---

Photo by I'M ZION on Unsplash

• Dataset: AFT-HMT 2025 - 50,000+ annotated tactical drone flight frames.
• Architecture: Multi-stage HMT - Edge Processing Unit (EPU) + Pilot HMI (Human-Machine Interface).
• Performance Metrics: Latency (ms), Precision/Recall (IOU), Mean Average Precision (mAP), and Power Efficiency (mWh).
• Expected Results: > 90% target detection under 20ms inference time.

---

Photo by Sergey Koznov on Unsplash

Advancing FPV Precision through Human-Machine Teaming: A Strategic Leap for Future Operations.

Summary of AI-Enabled HMT Dissertation Research

---

Photo by Joash Viriah on Unsplash

• Future Work: Transitioning from edge-inference to federated learning models.
• Refining robustness against adversarial electronic warfare (EW) countermeasures.
• Scaling HMT modules for multi-agent autonomous drone swarms.
• Expanding AFT-HMT 2025 dataset with heterogeneous terrain scenarios.

---

Photo by Winston Chen on Unsplash

> Strategic superiority in modern combat is defined by our ability to seamlessly integrate artificial intelligence with human operational intuition.

— Simranjeet Singh Raina

---

Photo by Joash Viriah on Unsplash

Thank you for your attention. Open to Q&A.

Final Remarks