engleski

Accurate traffic cone detection in racing environments based on a LIDAR sensor

Feature detection and localization play an important role in vehicle autonomy because it is a prerequisite for correct inference about the vehicle's environment. Man-made environments often contain fiducial distinctive markers which autonomous systems use to determine their position. The "Formula Student Driverless" (FSD) competition is an example of such an environment, containing traffic cones as markers used to determine the track edges. In the competition, a FSD race car needs to autonomously reach the finish line. Due to accuracy requirements, a light detection and ranging sensor is used to detect traffic cones and determine their position relative to the race car. This thesis investigates and applies traffic cone extraction from point cloud data, considering the racing environment structure.

lidar ; point cloud ; ground removal ; ground segmentation ; RANSAC ; autonomous driving ; clustering ; formula student ; driverless ; FS ; FSD

nije evidentirano