Gazebo built a virtual environment to complete ROS robot positioning and navigation simulation and YOLO detection and identification of marked objects

---

Preface

Environment description:
This process uses Ubuntu18.04+ROS Melodic+Gazebo+yolo3

---

1. Preliminary preparation

(1) Create a workspace

1. Create ros work area

   mkdir -p ~/robot_positioning_ws/src
   

2. Switch to work area directory

   cd ~/robot_positioning_ws/src
   

3. Initialize the ROS workspace

   catkin_init_workspace
   

(Two) function package racecar

1. Download Gazebo to build a track

   git clone https://github.com/xmy0916/racecar.git
   

   

2. Install controllers related

   sudo apt-get install ros-melodic-controller-manager
   sudo apt-get install ros-melodic-gazebo-ros-control
   sudo apt-get install ros-melodic-effort-controllers
   sudo apt-get install ros-melodic-joint-state-controller
   

3. Racecar feature pack compilation

   cd ..
   catkin_make
   

   Problems encountered in compiling
   ① Could not find a package configuration file provided by "driver_base" with any of the following names
   solution

   sudo apt-get install ros-melodic-driver-base
   

   ② Could not find a package configuration file provided by "OpenCV" with any of the following names
   
   Solution

   locate OpenCVConfig.cmake
   sudo gedit ~/robot_positioning_ws/src/racecar_gazebo/CMakeLists.txt
   modify the path on line 7 to your path: set(OpenCV_DIR /opt/ros/kinetic/share/OpenCV-3.3.1-dev/)

   
   ③ ackermann_msgs/AckermannDrive.h: No such file or directory
   
   Solution

   sudo apt-get install ros-melodic-ackermann-msgs
   

   Compiled successfully
   

2. Gazebo builds its own virtual environment

(1) Related settings

1. Set environment variables

   echo "source ~/robot_positioning_ws/devel/setup.bash" >> ~/.bashrc
   source ~/.bashrc
   

(2) Running the car model

1. Run the statement

   roslaunch racecar_gazebo racecar.launch 
   

2. Operation effect
   Description: The small window of the control interface of tk controls the movement of the car, W means forward, D means turn left, S means backward, A means turn right
   

(3) Manually build the environment

1. Open gazebo

   roslaunch gazebo_ros empty_world.launch
   

2. Click Edit->Build Editor, create a model, and save the model
   

3. Import the track model framework you just created to
   run the car model

   roslaunch racecar_gazebo racecar.launch 
   

   Import and create the environment model
   insert->select the corresponding saved model,
   
   add obstacles and
   
   save as a world file
   

(4) Operation and creation environment

1. Create a launch file and configure track parameters

   cd ~/robot_positioning_ws/src/racecar/racecar_gazebo/launch
   sudo gedit lyy.launch
   

   lyy.launch content

   <?xml version="1.0"?>
   <launch>
     <!-- Launch the racecar -->
     <include file="$(find racecar_gazebo)/launch/racecar.launch">
       <arg name="world_name" value="lyy"/>
     </include>
   </launch>
   

2. Run gazebo

   roslaunch racecar_gazebo lyy.launch 
   

   Shown as follows
   

Three, perform gmapping mapping

1. gmapping mapping

   roslaunch racecar_gazebo slam_gmapping.launch
   

   Problem description: ERROR: cannot launch node of type [gmapping/slam_gmapping]: gmapping
   
   solution

   sudo apt-get install ros-melodic-gmapping
   

2. Run a complete circle (through the kt control window)
   the starting point of the
   
   trolley, the end point of the trolley is
   
   displayed in Rviz
   

3. Save the map created by gmapping

   sudo apt-get install ros-melodic-map-server
   cd ~/robot_positioning_ws/src/racecar/racecar_gazebo/map
   rosrun map_server map_saver -f lyy_map
   

   

Fourth, the car's autonomous positioning and navigation

(1) The environment created by myself

1. Editing the launch file

   cd ~/robot_positioning_ws/src/racecar/racecar_gazebo/launch
   sudo gedit lyy_auto.launch
   

   lyy_auto.launch file content

   <?xml version="1.0"?>
   <launch>
     <!-- Launch the racecar -->
     <include file="$(find racecar_gazebo)/launch/racecar.launch">
       <arg name="world_name" value="lyy"/>
     </include>
     
     <!-- Launch the built-map -->
     <node name="map_server" pkg="map_server" type="map_server" args="$(find racecar_gazebo)/map/lyy_map.yaml" />
   
     <!--Launch the move base with time elastic band-->
     <param name="/use_sim_time" value="true"/>
     <node pkg="move_base" type="move_base" respawn="false" name="move_base" output="screen">
       <rosparam file="$(find racecar_gazebo)/config/costmap_common_params.yaml" command="load" ns="global_costmap" />
       <rosparam file="$(find racecar_gazebo)/config/costmap_common_params.yaml" command="load" ns="local_costmap" />
       <rosparam file="$(find racecar_gazebo)/config/local_costmap_params.yaml" command="load" />
       <rosparam file="$(find racecar_gazebo)/config/global_costmap_params.yaml" command="load" />
       <rosparam file="$(find racecar_gazebo)/config/teb_local_planner_params.yaml" command="load" />
   
       <param name="base_global_planner" value="global_planner/GlobalPlanner" />
       <param name="planner_frequency" value="0.01" />
       <param name="planner_patience" value="5.0" />
       <!--param name="use_dijkstra" value="false" /-->
       
       <param name="base_local_planner" value="teb_local_planner/TebLocalPlannerROS" />
       <param name="controller_frequency" value="5.0" />
       <param name="controller_patience" value="15.0" />
   
       <param name="clearing_rotation_allowed" value="false" />
     </node>
     
   </launch>
   

   

2. Run the created environment

   roslaunch racecar_gazebo lyy_auto.launch
   

3. Start rviz

   roslaunch racecar_gazebo racecar_rviz.launch
   

   

4. Design the trajectory of the car.
   Use 2D Nav Goal to publish the target.
   
   Problem description: 使用2D Nav Goal发布目标时，始终没有反应
   Solution

   sudo apt-get install ros-melodic-teb-local-planner
   

   Since there is no wall separating the start point and the end point, now move the end point to the opposite side

5. Start the path_pursuit.py script file

   rosrun racecar_gazebo path_pursuit.py
   

   The actual situation is
   
   temporarily unclear what caused it.

(2) The environment is included in the downloaded function package

1. Operating environment

   roslaunch racecar_gazebo racecar_runway_navigation.launch 	
   

2. Start rviz

   roslaunch racecar_gazebo racecar_rviz.launch
   

3. Design the car trajectory
   using 2D Nav Goal to publish the target position

4. Start the path_pursuit.py script file

   rosrun racecar_gazebo path_pursuit.py
   

   Running results (part of the movement process)
   

Five, use YOLO to detect and identify marked objects

(1) Load YOLO

1. ssh settings

   ssh-keygen ##输入之后根据提示输入两次回车，两次密码
   ls ~/.ssh
   eval `ssh-agent`
   ssh-add ~/.ssh/id_rsa
   cat ~/.ssh/id_rsa.pub #得到一串密匙
   

   Log in to GitHub, select settings, ssh and GPG keys in the account drop-down option, select new ssh key, enter the name Default public key, and then copy the obtained key to the text box, and then click add ssh key, the key setting is complete.

2. Download darknet-ros

   cd ~/robot_positioning_ws/src
   git clone --recursive [email protected]:leggedrobotics/darknet_ros.git
   

3. Compile

   cd ~/robot_positioning_ws
   catkin_make -DCMAKE_BUILD_TYPE=Release
   

   
   说明：
   Start to compile the entire project. After the compilation is completed, it will be checked whether there are two model files, yolov2-tiny.weights and yolov3.weights, in the ./darknet_ros/darknet_ros/yolo_network_config/weights file. By default, the downloaded code does not include this in order to save volume. Two model files. Therefore, the model file will be downloaded automatically after compilation, and there will be a long waiting time at this time. Before starting to compile, you can download it in advance, copy the model file to the above folder, and it will not download again.

(2) Realize the detection of marked objects

1. Open gazebo

   roslaunch racecar_gazebo racecar_runway_navigation.launch
   

2. Open the racecar.gazebo file to view topics
   

3. Modify the topic subscribed in the ros.yaml file
   Original file
   
   Modified file
   

4. Start YOLO V3

   roslaunch darknet_ros darknet_ros.launch
   

   This process is very easy to get stuck, so you need to try several times.
   
   In the end, the recognition result is not very accurate. The possible reason is that the camera did not capture all the objects, and some of the objects were recognized as other objects. This process did take a long time, and it was too easy to get stuck. It takes more patience to achieve the final effect. When building the environment by yourself, you need to pay attention to what separates the starting point and the end point. When you don’t plan the path later, it won’t be drawn.

---

Reference connection

1. Under Ubuntu 18.04, use Gazebo to build a track to complete ROS robot positioning and navigation simulation + load YOLO to detect and identify marked objects [smart car]
2. Realize ROS smart car positioning and navigation simulation under Ubuntu 18.04
3. Implementation of darknet_ros (YOLO V3) detection under ROS
4. Embedded system application development and big job-ROS robot positioning and navigation simulation