8. karto mapping algorithm

karto: http://wiki.ros.org/slam_karto

map_server: https://wiki.ros.org/map_server

8.1. Introduction

Karto is a 2D laser SLAM solution based on a sparse graph optimization method with loop closure detection. The graph optimization method uses the mean value of the graph to represent the map. Each node represents a position point of the robot trajectory and a sensor measurement data set. Each new node is added and the calculation is updated. Karto uses spa (karto_slam) or g2o (nav2d) optimization library, and the front-end and back-end use a single thread.

The ROS version of Karto_SLAM, in which the Spare Pose Adjustment (SPA) used is related to scan matching and loop closure detection. The more landmarks there are, the greater the memory requirements. However, compared with other methods, the graph optimization method has greater advantages in mapping in large environments because it only contains point graphs (robot poses), and the map is obtained after the pose is obtained.

overall program framework

8.2. Use

Note: When building a map, the slower the speed, the better the effect (note that the rotation speed should be slower). If the speed is too fast, the effect will be poor.

According to different models, you only need to set the purchased model in [.bashrc], X1 (normal four-wheel drive) X3 (Mailun) Take X3 as an example

xxxxxxxxxx
#Raspberry Pi 5 master needs to enter docker first, please perform this step
#If running the script into docker fails, please refer to ROS/07, Docker tutorial
~/run_docker.sh

Open the [.bashrc] file

xxxxxxxxxx
sudo vim .bashrc

Find the [ROBOT_TYPE] parameters and modify the corresponding car model

xxxxxxxxxx
export ROBOT_TYPE=X3 # ROBOT_TYPE: X1 X3 X3plus R2 X7

Note: Due to the difference in radar laser data between 4ROS radar and A1 radar, you need to modify the karto source code part before it can run normally and modify the file.

xxxxxxxxxx
sudo gedit ~/software/library_ws/src/open_karto-melodic/include/open_karto/Karto.h

Locate line 4165, the source code is,

xxxxxxxxxx
void Update()
     {
       m_NumberOfRangeReadings = static_cast<kt_int32u>(math::Round((GetMaximumAngle() -
                                                                     GetMinimumAngle())
                                                                     / GetAngularResolution())+1);
     }

Need to be changed to,

xxxxxxxxxx
  void Update()
     {
       m_NumberOfRangeReadings = static_cast<kt_int32u>(math::Round((GetMaximumAngle() -
                                                                     GetMinimumAngle())
                                                                     /GetAngularResolution()));
     }

After saving, switch to the ~/software/library_ws directory and use catkin_make to compile.

xxxxxxxxxx
cd ~/software/library_ws
catkin_make

8.2.1. Start

Start the command (robot side). For the convenience of operation, this section takes [mono + laser + yahboomcar] as an example.

xxxxxxxxxx
roslaunch yahboomcar_nav laser_bringup.launch # laser + yahboomcar
roslaunch yahboomcar_nav laser_usb_bringup.launch # mono + laser + yahboomcar
roslaunch yahboomcar_nav laser_astrapro_bringup.launch # Astra + laser + yahboomcar

Mapping command (robot side)

<PI5 needs to open another terminal to enter the same docker container

xxxxxxxxxx
roslaunch yahboomcar_nav yahboomcar_map.launch use_rviz:=false map_type:=karto

• [use_rviz] parameter: whether to enable rviz visualization.

• [map_type] parameter: Set the mapping algorithm [karto].

Turn on the visual interface (virtual machine side)

xxxxxxxxxx
roslaunch yahboomcar_nav view_map.launch

The gap at the back of the robot is due to the obstruction caused by the installation position of the display screen, so a certain range of radar data is blocked. The shielding range can be adjusted, or it can not be blocked according to the actual situation. For specific operations, see [01. Radar Basic Course].

8.2.1. Controlling the robot

• Keyboard controls robot movement

xxxxxxxxxx
rosrun teleop_twist_keyboard teleop_twist_keyboard.py # System integration
roslaunch yahboomcar_ctrl yahboom_keyboard.launch # Custom

• Control the robot movement with the handle

Make the robot cover the area to be mapped and the map should be as closed as possible.

There may be some scattered points during the mapping process. If the mapping environment is well closed, relatively regular, and the movement is slow, the scattering phenomenon will be much smaller.

8.2.1. Map saving

xxxxxxxxxx
rosrun map_server map_saver -f ~/yahboomcar_ws/src/yahboomcar_nav/maps/my_map # The first way
bash ~/yahboomcar_ws/src/yahboomcar_nav/maps/map.sh # The second way

The map will be saved to the ~/yahboomcar_ws/src/yahboomcar_nav/maps/ folder, a pgm image and a yaml file.

map.yaml

xxxxxxxxxx
image: map.pgm
resolution: 0.05
origin: [-15.4,-12.2,0.0]
Negate: 0
occupied_thresh: 0.65
free_thresh: 0.196

Parameter analysis:

• image: The path of the map file, which can be an absolute path or a relative path.

• resolution: resolution of the map, meters/pixel

• Origin: 2D pose (x, y, yaw) in the lower left corner of the map. The yaw here is rotated counterclockwise (yaw=0 means no rotation). Many parts of the current system ignore the yaw value.

• negate: whether to reverse the meaning of white/black and free/occupied (the interpretation of the threshold is not affected)

• occupied_thresh: Pixels with an occupation probability greater than this threshold will be considered fully occupied.

• free_thresh: Pixels with occupancy probability less than this threshold will be considered completely free.

8.3. Topics and services

Node view

xxxxxxxxxx
rqt_graph

8.4. Configuration parameters

• Common parameters

• Correction parameters

• Loopback parameters

• Scan Matcher parameters

8.5, TF transformation

View tf tree

xxxxxxxxxx
rosrun rqt_tf_tree rqt_tf_tree