Map Perception. HD-Mapless self-driving system.

Localization and mapping for autonomous vehicles.

• Worked in the Laboratory for Perception, Robotics, and Grounded Reasoning Systems (4PROGRESS). Conducted research on a SLAM-based robust, accurate and real-time 6D object pose estimation and semantic world modeling pipeline for densely cluttered environments. • Developed a robust data association and false positives removal algorithm for semantic measurements by incorporating object-level geometric consistency. It achieved a 99% precision rate over multiple datasets.• Applied Lie-Algebra based graph optimization using Ceres nonlinear optimization library to improve the object poses accuracy.

Mapping for autonomous trucks.

• Localization : Mobile Robot Pose Estimation (URDF, Particle Filter, Sensor fusion, Hokuyo laser, Rviz, Gazebo) - Created a robot model utilizing Adaptive Monte-Carlo Localization to locate itself in a map and autonomously navigate to the desired position. - Developed a ROS package launching the robot model in the Gazebo/Rviz world and integrated it with ROS navigation stack to plan trajectories.• Mapping: 3D Environment Reconstruction (SLAM, RTAB-Map, Kinect RGB-D, Hokuyo laser, Sensor fusion, Rviz) - Built a simulated world in Gazebo with a robot model and implemented graph-based RTAB-Map SLAM to generate a map of the environment.

• Investigated the behavior of SH wave for fatigue damage detecting using Electro-Magnetic Acoustic Transducer.• Developed a thermal sensor for nuclear pressure vessel robotic inspection system and calibrated its accuracy