Salvin George

• Master’s thesis on terrain traversability estimation using camera, LiDAR, and sensor fusion for drivable/non-drivable classification • Designed and trained deep learning models for multi-modal prediction, leading dataset curation and benchmarking • Built a custom ROS 2 pipeline for synchronized camera-LiDAR data collection, calibration, and preprocessing • Deployed optimized inference on NVIDIA Jetson AGX, engineering a real-time perception stack for autonomous ground robot navigation in rough terrain

• Developed a Python/C++ framework for autonomous radioactive source localization using UAVs, implementing multiple algorithms and evaluating performance systematically • Created a ROS2/Gazebo simulation environment with custom sensor and source plugins for reproducible experiments • Investigated reinforcement learning approaches for autonomous source localization and built a PyQt5 evaluation interface for comparative analysis