Appl. Sci. 2020, 10, 347 4 of 18
the Cartographer’s local SLAM integrated with the AMDS to control the sensor distances, and the scan
matcher’s search window during map generation and pose generation.
3.1. Multi-Distance Scheduler in LiDAR Scans
A multi-distance scheduler (MDS) is inspired from the idea of multiple resolution maps in Google’s
Cartographer utilizing a multi-stage scan matcher from low to high resolution. An MDS can be used to
reduce computational cost as compared with using full high-resolution scans. It uses a scan matcher
with low-resolution map data for the initial pose estimation, which then is used to bypass scan matcher
iteration of every possible translation and rotation in a high-resolution map. Thus, it minimizes the
performance penalty while maintaining accuracy.
The multi-distance approach is similar to the multi-resolution map. Instead of controlling
resolution, we propose controlling sensor distance that generates point cloud frame (
F
), whereby
contains an array of object point (
F
) detected from the LiDAR in local sensor coordinates. Limiting
sensor distance results in different details of the scan, as shown in Figure 2.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 4 of 18
to the Cartographer’s local SLAM integrated with the AMDS to control the sensor distances, and the
scan matcher’s search window during map generation and pose generation.
3.1. Multi-Distance Scheduler in LiDAR Scans
A multi-distance scheduler (MDS) is inspired from the idea of multiple resolution maps in
Google’s Cartographer utilizing a multi-stage scan matcher from low to high resolution. An MDS can
be used to reduce computational cost as compared with using full high-resolution scans. It uses a
scan matcher with low-resolution map data for the initial pose estimation, which then is used to
bypass scan matcher iteration of every possible translation and rotation in a high-resolution map.
Thus, it minimizes the performance penalty while maintaining accuracy.
The multi-distance approach is similar to the multi-resolution map. Instead of controlling
resolution, we propose controlling sensor distance that generates point cloud frame (ℱ), whereby ℱ
contains an array of object point (𝑝_𝑝𝑜𝑖𝑛𝑡
,,
) detected from the LiDAR in local sensor coordinates.
Limiting sensor distance results in different details of the scan, as shown in Figure 2.
(a) (b)
Figure 2. Sensor scan state with different distances: (a) LiDAR scan 25 m distances and (b) LiDAR
scan 60 m distances.
The controlling method is done via a scheduler, which swaps between two sensor states, namely
a short distance state and a long distance state at a fixed frequency. It is known in scan matching
algorithms that the larger the image data or map, the greater the amount of processing time required
as compared with small images [23]. However, because the large images from the sensor scan capture
more features, the computation results using these features generate less rotational and translational
errors. Thus, it is hoped that the combination of these two scan states, i.e., short and long distance,
achieves a similar result of increasing the processing speed performance while maintaining accuracy.
Some additional modifications are needed to accommodate the proposed method. Table 1 shows the
scheduler algorithm to control the distance of the scan. It uses a time variable which is based on the
period for trigger. The period ratio is configured to, for example, 1:1, 1:2, 1:3, etc. The weight of scan
matcher also needs to be weighted differently. A more extensive distance sensor scan requires higher
weight on the SLAM pose estimation as compared with a shorter scan distance because a larger map
contains more feature data, and thus is likely to have low covariance as more features are used to
compute the pose.
Figure 2.
Sensor scan state with different distances: (
a
) LiDAR scan 25 m distances and (
b
) LiDAR scan
60 m distances.
The controlling method is done via a scheduler, which swaps between two sensor states, namely
a short distance state and a long distance state at a fixed frequency. It is known in scan matching
algorithms that the larger the image data or map, the greater the amount of processing time required
as compared with small images [
23
]. However, because the large images from the sensor scan capture
more features, the computation results using these features generate less rotational and translational
errors. Thus, it is hoped that the combination of these two scan states, i.e., short and long distance,
achieves a similar result of increasing the processing speed performance while maintaining accuracy.
Some additional modifications are needed to accommodate the proposed method. Table 1 shows the
scheduler algorithm to control the distance of the scan. It uses a time variable which is based on the
period for trigger. The period ratio is configured to, for example, 1:1, 1:2, 1:3, etc. The weight of scan
matcher also needs to be weighted differently. A more extensive distance sensor scan requires higher
weight on the SLAM pose estimation as compared with a shorter scan distance because a larger map
contains more feature data, and thus is likely to have low covariance as more features are used to
compute the pose.
剩余17页未读，继续阅读
laocui1
- 粉丝: 494
- 资源: 60
我的内容管理
收起
- 我的资源
快来上传第一个资源
我的收益 登录查看自己的收益
我的积分
登录查看自己的积分
我的C币
登录后查看C币余额
我的收藏 
我的下载 
下载帮助
会员权益专享
最新资源
- zigbee-cluster-library-specification
- JSBSim Reference Manual
- c++校园超市商品信息管理系统课程设计说明书(含源代码) (2).pdf
- 建筑供配电系统相关课件.pptx
- 企业管理规章制度及管理模式.doc
- vb打开摄像头.doc
- 云计算-可信计算中认证协议改进方案.pdf
- [详细完整版]单片机编程4.ppt
- c语言常用算法.pdf
- c++经典程序代码大全.pdf
- 单片机数字时钟资料.doc
- 11项目管理前沿1.0.pptx
- 基于ssm的“魅力”繁峙宣传网站的设计与实现论文.doc
- 智慧交通综合解决方案.pptx
- 建筑防潮设计-PowerPointPresentati.pptx
- SPC统计过程控制程序.pptx
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
评论0