As intelligent vehicles (IVs) applications expand to large-scale and open road scenarios, vectorized high-definition maps (VHD maps) demonstrate greater potential than traditional metric maps. Despite substantial progress, the low-cost automated construction of VHD maps remains a serious challenge, which has slowed their practical application.

In this dissertation, we tackle the issue of low-cost automated VHD map construction by enhancing Inverse Perspective Mapping (IPM) technology. We begin with a straightforward scenario, proposing a framework that improves IPM through pose-guided optimization, specifically addressing the automated creation of high-definition angular marking maps (HDAM maps). This framework not only enhances the accuracy of IPM-based mapping results but also refin...[ Read more ]

As intelligent vehicles (IVs) applications expand to large-scale and open road scenarios, vectorized high-definition maps (VHD maps) demonstrate greater potential than traditional metric maps. Despite substantial progress, the low-cost automated construction of VHD maps remains a serious challenge, which has slowed their practical application.

In this dissertation, we tackle the issue of low-cost automated VHD map construction by enhancing Inverse Perspective Mapping (IPM) technology. We begin with a straightforward scenario, proposing a framework that improves IPM through pose-guided optimization, specifically addressing the automated creation of high-definition angular marking maps (HDAM maps). This framework not only enhances the accuracy of IPM-based mapping results but also refines the initial estimates of the IPM matrix. We then extend this framework to encompass more complex transportation scenarios, incorporating additional traffic elements such as common ground markings and lane lines.

To illustrate the practical value of VHD maps in IV applications, we focus on the state estimation task of autonomous vehicles (AVs). We introduce a universal state correction system that relies solely on VHD maps, enabling AVs to achieve higher state estimation accuracy compared to traditional simultaneous localization and mapping (SLAM) methods. To optimize the use of VHD maps, we propose a query-based data exchange protocol, allowing multiple vehicles to share the same map data, aligning with the future trend of crowdsourced mapping.

In summary, this dissertation advances the low-cost automated construction of VHD maps and demonstrates their extensive potential in IV applications, particularly through the example of AV state estimation. The development of low-cost automated VHD maps paves the way for their broader adoption. The practical application of VHD map in the field of AV proves that this map form is not merely theoretical but meets the current needs of IVs and is poised to replace traditional metric-based maps.