Generating Vector Map Data for Autonomous Driving
Name
Krister Looga
Abstract
Self-driving cars use detail-rich maps called High Definition maps (HD-map) to enable self-driving. These maps contain much information, but there is no standardized way of making them. Furthermore, creating such maps by hand takes much time.
One of the properties on the HD-map used by the Autonomous Driving Lab at the University of Tartu describes a trajectory that a self-driving vehicle uses as virtual rails.
For the bachelor's thesis, a program was created to aggregate data collected during manual drives in Tartu. That data is used to create a new reference trajectory for the HD-map. The reason for generating this data is to decrease the amount of manual labor required to create the HD-map. Furthermore, since data is generated based on actual driving data, it should better reflect vehicles' speed and position on the road.
The first two chapters talk about HD-maps and trajectories and how they can be created. After that, an overview is given of the solution created during the thesis. The results are introduced, and its accuracy is compared to the manually created trajectory. Finally, the limitations and possibilities for future development are proposed.
One of the properties on the HD-map used by the Autonomous Driving Lab at the University of Tartu describes a trajectory that a self-driving vehicle uses as virtual rails.
For the bachelor's thesis, a program was created to aggregate data collected during manual drives in Tartu. That data is used to create a new reference trajectory for the HD-map. The reason for generating this data is to decrease the amount of manual labor required to create the HD-map. Furthermore, since data is generated based on actual driving data, it should better reflect vehicles' speed and position on the road.
The first two chapters talk about HD-maps and trajectories and how they can be created. After that, an overview is given of the solution created during the thesis. The results are introduced, and its accuracy is compared to the manually created trajectory. Finally, the limitations and possibilities for future development are proposed.
Graduation Thesis language
Estonian
Graduation Thesis type
Bachelor - Computer Science
Supervisor(s)
Tambet Matiisen, Edgar Sepp
Defence year
2021