Calibration

Implemented a radar odometry based on Doppler Iterative Closest Point (DICP).

Outperforms LiDAR odometry, allowing robust radar integration to online calibration.

Novel view rendering via Gaussian Splatting.

Gaussian sampling based on Multi-View Stereo (MVS) cost volumes.

Worked on LiDAR localization fault detection, collaborating with a team of 20+ people through GitHub.

Designed an algorithm to detect if Normal Distribution Transformation (NDT) is trapped in local minimum.

Achieved real-time execution capability, ensuring seamless Unscented Kalman Filter (UKF) updates.

Led RGBD Localization project for URSrobot's lawnmower robots.

Enhanced frontend tracker to overcome limited feature points and scene vibration in lawn environments.

Improved performance of ORB-SLAM on NVIDIA Jetson Nano, increasing FPS from 7Hz to 15Hz.