Cartographer Overview

• Cartographer Overview
  • Ch0 Basic Concepts
    • history
    • lib
    • advantage
  • Ch1 Grid Maps
    • Assumption 1
    • Assumption 2
    • Assumption 3
    • Estimating a Map From Data
      • Static State Binary Bayes Filter
      • From Ratio to Probability
      • Log Odds Notation
      • Occupancy Mapping Algorithm
  • Ch2 Local SLAM
  • Ch3 Closing Loops
  • Contact
  • License

Cartographer Overview

Created 2021.07.20 by William Yu; Last modified: 2021.07.20-V1.0.0

Contact: windmillyucong@163.com

Copyleft! 2021 William Yu. Some rights reserved.

References

papers

• Real-Time Loop Closure in 2D LIDAR SLAM https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/45466.pdf
• Cartographer Ros https://google-cartographer-ros.readthedocs.io/en/latest/
• Algorithm walkthrough https://google-cartographer-ros.readthedocs.io/en/latest/algo_walkthrough.html
• Ros Cartographer 调参 https://google-cartographer-ros.readthedocs.io/en/latest/tuning.html
• Github https://github.com/cartographer-project/cartographer
• code note https://zhuanlan.zhihu.com/p/47997092
• A Review of Point Cloud Registration Algorithms for Mobile Robotics https://hal.archives-ouvertes.fr/hal-01178661/file/2015_Pomerleau_FnTRo_Review.pdf ICP及其各种变形，2015比较详细的综述
• grid map http://ais.informatik.uni-freiburg.de/teaching/ws15/mapping/pdf/slam10-gridmaps.pdf
• Bilibili https://www.bilibili.com/video/BV1nV411Z7W5?from=search&seid=8833147598009061857
• code view https://zhuanlan.zhihu.com/p/48010119

blogs

• 泡泡机器人SLAM https://mp.weixin.qq.com/s?__biz=MzI5MTM1MTQwMw==&mid=2247484537&idx=1&sn=86200d961cf933896a9781bbe58442cc&chksm=ec10ba7ddb67336ba920a3c6b7e6414a0131bb775d6695e526d25dc6d377e31578684e83f802&scene=1&srcid=0925Vlxe9psorns6CY8O4sC7#rd

Ch0 Basic Concepts

history

1. 2D scan matching
   • scan to scan matching – ICP,etc
   • scan to map matching – Hector SLAM,etc
2. GMapping
   • 使用了 IMU
     • 作用1：将倾斜的雷达数据投影在水平面上面，如果机器人是水平移动，完全可以不适用
     • 作用2：提供一个 initial guess 用于 scan matching
3. cartographer
   • 2016

lib

• Ceres / g2o
  • Ceres user handbook http://ceres-solver.org/users.html
• Eigen

advantage

• 2016

• 论文精度：室内 5cm resolution

• 用途：室内

• Sensor：激光雷达，IMU，里程计

• Input:

  • IMU linear_acceleration angular_velocity；
  • Odometry current Pose
  • Rangefinder（激光雷达） 的 origin和ranges
    • origin是rangefinder在robot坐标系下的位置
    • ranges是robot坐标系下的点云（3维）

• 重点

  • submap 的选择
  • 闭环检测的加速策略

• 主要代码：（2d 相关）

  1. map_builder是最顶层，cartographer_ros用cartographer就是通过这个。

  2. global_trajectory_builder： 使用了全库核心，数据给到local_trajectory_builder， 结果再给sparse_pose_graph（local是相对于global的，是指没有全局优化，并不是说robot坐标系； ）

  3. submaps 可以看到怎么管理子图，子图的position和robot pose的关系 （submap的local pose指的submap在世界下的位置）

  4. ceres_scan_matcher和几个使用的functor可以看到怎么在做scan to map match。 它的优化除了论文上的损失函数，还加了对初值的约束；计算residual查找grid中的值时用了插值ha。

  5. probability_grid和range_data_inserter写了grid map是在怎么迭代更新的，写成查表很快

  6. 3d和2d在同一个框架下，差别不大

     主要差别：

     • 3d前端用了ukf，代码中写了参考文献的
     • 另外在优化时，3d多了些优化，包括优化imu的重力方向
     • 3d没有那么多拧巴的维度转换，不需要水平化，应该会看得通畅些

Ch1 Grid Maps

refitem: http://ais.informatik.uni-freiburg.de/teaching/ws15/mapping/pdf/slam10-gridmaps.pdf

• Discretize the world into cells 将世界分割成一个个cell
• Grid structure is rigid Grid结构是坚固的
• Each cell is assumed to be occupied or free space 每个cell有两种状态
• Non-parametric model 无参数模型
• Large maps require substantial memory resources 大地图需要较高的内存
• Do not rely on a feature detector 不依赖于特征检测

Assumption 1

• The area that corresponds to a cell is either completely free or occupied. 一个cell只能有三种种状态：空闲的或者占据的或者Unknow

Occupancy Probability

• Each cell is a binary random variable that models the occupancy
• Cell is occupied: $p(m_i) = 1$
• Cell is not occupied: $p(m_i) = 0$
• No knowledge: $p(m_i) = 0.5$

Assumption 2

• The world is static. 世界是静态的

Assumption 3

• The cells (the random variables) are independent of each other. 每个cell在概率上彼此独立。

Representation

• The probability distribution of the map is given by the product over the cells. \(p(map) = \prod_{i} p(m_i)\)

Estimating a Map From Data

Static State Binary Bayes Filter

From Ratio to Probability

Log Odds Notation

Occupancy Mapping Algorithm

Ch2 Local SLAM

• post https://blog.csdn.net/weixin_36976685/article/details/84994701

Ch3 Closing Loops

• post https://gaoyichao.com/Xiaotu/?book=Cartographer%E6%BA%90%E7%A0%81%E8%A7%A3%E8%AF%BB&title=%E5%88%86%E6%94%AF%E5%AE%9A%E7%95%8C%E9%97%AD%E7%8E%AF%E6%A3%80%E6%B5%8B%E7%9A%84%E5%8E%9F%E7%90%86%E5%92%8C%E5%AE%9E%E7%8E%B0

Contact

Feel free to contact me windmillyucong@163.com anytime for anything.

License

Creative Commons BY-SA 4.0

• Previous
  mutable
• Next
  paper:Learning High-Speed Flight in the Wild