Artificial Intelligence Engineer drawn to projects that involve experimentation, research, and discovering new approaches to difficult problems. Whether it's Computer Vision, Data Science, or emerging developments in AI, I enjoy learning how ideas evolve and where the field is heading next. I'm passionate about building efficient AI systems that bridge research and real-world deployment.
🚁 GNSS Spoofing Detection for UAVs
Developing a lightweight multi-modal navigation framework for autonomous UAVs that detects GNSS spoofing and maintains accurate localization using vision and inertial sensing.
The proposed architecture consists of two ConvNeXt models optimized for deployment on an NVIDIA Jetson companion computer:
- 2D ConvNeXt performs cross-view geo-localization by matching the UAV camera feed against cached satellite imagery to verify GNSS positions.
- 1D ConvNeXt processes IMU time-series data and continuously learns sensor drift using the vision-verified GNSS position as supervision.
When GNSS spoofing or jamming is detected, the vision branch rejects the compromised navigation solution and seamlessly transfers localization to the IMU branch. The 1D ConvNeXt compensates for accumulated inertial drift in real time, enabling reliable navigation until trusted GNSS signals become available again.
This research builds upon recent advances in vision-based localization and learning-based inertial navigation while focusing on an efficient, GPU-accelerated architecture suitable for edge deployment.
