Utveckling samt evaluering av lokalisering och kooperativa styrsystem

Abstract

This report covers a bachelor’s thesis at Chalmers University of Technology. The purpose of this project was to further develop systems for localization and driving of autonomous vehicles inside of a lab environment building upon previous projects. Improvements were made in three distinct areas. The first being the indoor localization system, called GulliView, which consists of four ceiling-mounted cameras using Apriltags as calibration and detection of vehicles. GulliView was improved by implementing an efficient undistortion algorithm to counteract the camera’s distortion while maintaining accuracy. As well as creating a unified global coordination system using the world position in meters. Secondly, additional general maneuvers were implemented to autonomous robots using GulliView for positioning. The added maneuvers handle common traffic situations such as intersections, highway merging and roundabouts. Lastly, an advancement was made on implementing Sensor Fusion between a vehicle, which has a previously developed internal positioning system, and GulliView. The vehicle integrates the internal position and the external position given from GulliView. GulliView attained a median delay decrease of 86%, going from 102 ms to 14.0 ms in time per execution cycle. Meanwhile, the median frequency increased from 10.5 Hz to 16.6 Hz. Improvements on GulliView’s positioning accuracy were also observed, going from discrepancies of 8-18% to 1-2%. For the autonomous vehicles, the added maneuvers added only an average of 24.32% increased waiting time in traffic scenarios while maintaing safety. Finally, the fusion of internal and external values resulted in a positioning discrepancy of 5%. These results prove promising and may greatly help further development of all three systems in future projects.