I have recently submitted my master’s thesis “Failure-resilient Graph-based SLAM for Autonomous Robotic Exploration in GNSS-Denied Environments”, on which I worked at Czech Technical University in Prague, COMROB lab.
I created a robust Lidar-Inertial odometry based on an LIO-SAM system, that detects the failure of the LiDAR odometry using IMU measurements, and incorporates additional odometry in case the LiDAR-based fails, changing the structure of the graph to account for degenerated laserscans that caused the failure. Custom datasets with a HUSKY ClearPath mobile platform. ECMR conference paper on this topic is submitted. For the thesis, the localization system is incorporated into a custom underground exploration stack, which includes an efficient scalable elevation mapping module, and frontier-based exploration.
Developed localization system showcase and explanations.
Developed localization system showcase and explanations.
Summarizing video of the thesis: developed autonomous exploration of the campus backyard using the developed localization system.
Setup of the outdoor experiment to collect the data for localization. An equipped robot and a packed total station.
Exploration result: elevation map of the backyard and exploration trajectory.
Just a nice photo:) Me through the filter of the robot.
Autonomous exploration trial in the campus corridor.