This paper introduces a Graph Slam framework for improving the accuracy of elevation-maps. The core-idea is to utilize LIDAR images in determining the relationships between the road nodes and calculate the relevant elevation constraints. Phase correlation is applied to estimate the common areas between the nodes and confirm the loop-closure events. The dead-reckoning altitudinal measurements are used to initialize the graph in order to avoid the low GPS quality and guarantee a smooth change of the map consistency. The generated elevation-maps have been compared with the GPS-RTK maps using a reliable evaluation scheme. The results have verified that the proposed method can produce elevation-maps at the same accuracy level of GPS-RTK maps in open-sky area. Moreover, it significantly recovers the low accuracy of GPS-RTK maps in critical environments. Accordingly, the generated maps can safely be used during autonomous driving with better estimation of the vehicle altitudinal position.
Accurate Elevation Maps based Graph-Slam Framework for Autonomous Driving*
2019 IEEE Intelligent Vehicles Symposium (IV) ; 1254-1261
2019-06-01
1299604 byte
Conference paper
Electronic Resource
English
ACCURATE ELEVATION MAPS BASED GRAPH-SLAM FRAMEWORK FOR AUTONOMOUS DRIVING*
British Library Conference Proceedings | 2019
|Springer Verlag | 2023
|LIDAR Graph SLAM based Autonomous Vehicle Maps Using XY and Yaw Dead-Reckoning Measurements
British Library Conference Proceedings | 2021
|DL-SLAM: DIRECT 2.5D LIDAR SLAM FOR AUTONOMOUS DRIVING
British Library Conference Proceedings | 2019
|