2023 Fiscal Year Annual Research Report
On Multilevel Road Mapping for Autonomous Vehicles: A Study to Generate Accurate 2.5D LIDAR Maps Using Graph SLAM in Challenging Environments
| Project/Area Number |
22K17974
|
|---|
| Research Institution | Kanazawa University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2022-04-01 – 2024-03-31
|
| Keywords | SLAM in Multilevel Roads / LIDAR Mapping Systems / Autonomous Vehicles / Largescal Map Generation / 2.5D Maps |
|---|
| Outline of Annual Research Achievements |
We dedicated the last year to validate against the accurate global positions of layers in multilevel road structures. GNSS/INS-RTK system was not able to align layers in the XY plane in challenging traffic nodes such as Ohashi and Bejogi junctions. Therefore, we modified the proposed Graph SLAM system by integrating the driving directions into the road segmentation strategy based on the Yaw angle IDs. This tactics helps to group the road layered images in the same directions at loop closures and create virtual constraints between layers to optimize the global positions and align the road context. This contrbution was reported in in a Journal paper in Sensors. Pleasantly, we achieved impactful results to outperform accurate and expensive GNSS/INS-RTK systems and generate accurate maps in the XY plane in complex multilevel environments. Furthermore, we are glad to discover a critical issue to be addressed in the future proposals that the layers may intersect in the Z direction.
|
