研究目的
Investigating the use of cadastral plans together with 2D laser information and odometry in a graph-based approach to realize real-time global localization for autonomous driving.
研究成果
The paper proposed a real-time graph-based localization algorithm that uses cadastral plans, jointly considering odometry, laser scan matching, and the matching of laser scans with a cadastral map as constraints. The approach demonstrated that freely accessible geo-referenced sources of information can be used as a first step to build localization algorithms that do not require a first passage from an equipped vehicle, reaching an average positioning accuracy of 55cm at best.
研究不足
The cadastral plan itself might be the most important source of errors as it has been noticed several times to be inaccurate (misalignments for instance). Moreover, the fact that the laser is observing buildings at a certain height and not at its basis might also add noise to the estimation, especially with long balconies or irregular facades. Many buildings are missing from the cadastral plan due to a lack of tag in OSM.
1：Experimental Design and Method Selection:
The approach uses a graph-based formalism to represent the map-aided localization problem, combining odometry, laser scan matching, and matching of laser scans with a cadastral map as constraints.
2：Sample Selection and Data Sources:
The experiments were carried out in the streets of Versailles, France, using an experimental vehicle equipped with a Velodyne VLP-16 and an IxBlue ATLANS-C for ground truth.
3：List of Experimental Equipment and Materials:
Velodyne VLP-16 (simulating a 2D laser scanner), IxBlue ATLANS-C (RTK-GNSS with a high-end IMU), and a laptop with a processor running 4 cores at a maximum frequency of
4：5GHz. Experimental Procedures and Operational Workflow:
The approach involves creating a virtual point cloud from cadastral maps, filtering non-building objects from real laser observations, aligning both scans using Generalized ICP, and adding the resulting observation to the graph. Corridor-like environments are detected and corrected.
5：Data Analysis Methods:
The accuracy of the approach is evaluated by comparing the computed trajectory with the ground truth, providing Mean Error and Root-Mean-Square Error metrics.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
