Driven to better understand autonomous navigation under real-world constraints, I designed and implemented a full 3D LiDAR-based SLAM pipeline for GPS-denied environments. I translated theory into practice by building a physics-aware simulation and estimation framework using MATLAB and Simulink. The system fuses LiDAR point clouds with inertial navigation data to simultaneously estimate vehicle pose and incrementally construct a global map, mirroring perception algorithms used in autonomous vehicles and robotics platforms.

To validate the pipeline, I synthesized sensor data: using Unreal Engine to model a vehicle following a smoothed waypoint trajectory while collecting time-synchronized LiDAR and INS measurements. Point clouds were processed into a temporal stream, registered into a growing map, and monitored for loop closures using adaptive distance thresholds.