My research at Voyant Photonics focuses on using LiDAR point cloud data to identify and assess traversable environments for self-navigating robots. I did this by generating and analyzing normal vectors for each point in the given point cloud to determine whether or not a robot could traverse there.
Light detection and ranging (LiDAR) is a powerful tool that is used to map environments by sending out a stream of laser light and reading information upon its return. This information can be displayed in the form of a point cloud, which maps the environment with the use of small dots or ‘points.’ Autonomous robots and vehicles can use LiDAR as an aid to their navigation. However, analysis must be performed on point clouds to ascertain which points are part of the ground and which are not, so the robot knows where it can drive. The goal of my project was to segment the ground plane from LiDAR point clouds through the use of normals. Normals are vectors that point perpendicular to the plane they are describing. I use Python in order to perform this analysis and create an algorithm that segments the ground in order for these robots to determine where they can traverse. The result expected from this work is a point cloud that contains only points traversable by the robot, determined by the robot’s wheel radius and the surface slope. This work will help to aid the use of LiDAR in autonomous robot navigation by quickly assessing point clouds and segmenting ground points.
Press the pop-out button to view:
ElevatorPitch2026.mp4