实操 | Making my own Autonomous Robot in ROS / Gazebo, Day 2: Enable the robot

Day 2: Enable the robot

The goal of this post is to make the robot drivable.

Platform

Ubuntu 14.04
ROS indigo

Source

Sources for this tutorial can be found on GitHub

Prepare the repo

git checkout master
git branch day2_enable_robot
git push --set-upstream origin  day2_enable_robot

Now in a new branch we start working on enabling the robot.

Connect your robot to ROS

Alright, our robot is all nice and has this new car smell, but we can’t do anything with it yet as it has no connection with ROS. In order to add this connection we need to add gazebeo plugins to our model. There are different kinds of plugins:

World: Dynamic changes to the world, e.g. Physics, like illumination or gravity, inserting models
Model: Manipulation of models (robots), e.g. move the robots
Sensor: Feedback from virtual sensor, like camera, laser scanner
System: Plugins that are loaded by the GUI, like saving images

First of all we’ll use a plugin to provide access to the joints of the wheels. The transmission tags in our URDF will be used by this plugin the define how to link the joints to controllers. To activate the plugin, add the following to mybot.gazebo:

<gazebo>
  <plugin name="gazebo_ros_control" filename="libgazebo_ros_control.so">
    <robotNamespace>/mybot</robotNamespace>
  </plugin>
</gazebo>

Look at this tutorial for more information on how this plugin works.

With this plugin, we will be able to control the joints, however we need to provide some extra configuration and explicitely start controllers for the joints. In order to do so, we’ll use the package mybot_control that we have defined before. Let’s first create the configuration file:

roscd mybot_control
mkdir config
cd config
gedit mybot_control.yaml

This file will define three controllers: one for each wheel, connections to the joint by the transmission tag, one for publishing the joint states. It also defined the PID gains to use for this controller:

mybot:
  # Publish all joint states -----------------------------------
  joint_state_controller:
    type: joint_state_controller/JointStateController
    publish_rate: 50  
  

  # Effort Controllers ---------------------------------------
  leftWheel_effort_controller:
    type: effort_controllers/JointEffortController
    joint: left_wheel_hinge
    pid: {p: 100.0, i: 0.1, d: 10.0}
  rightWheel_effort_controller:
    type: effort_controllers/JointEffortController
    joint: right_wheel_hinge
    pid: {p: 100.0, i: 0.1, d: 10.0}

Now we need to create a launch file to start the controllers. For this let’s do:

roscd mybot_control
mkdir launch
cd launch
gedit mybot_control.launch

In this file we’ll put two things. First we’ll load the configuration and the controllers, and we’ll also start a node that will provide 3D transforms (tf) of our robot. This is not mandatory but that makes the simulation more complete:

<launch>

  <!-- Load joint controller configurations from YAML file to parameter server -->
  <rosparam file="$(find mybot_control)/config/mybot_control.yaml" command="load"/>

  <!-- load the controllers -->
  <node name="controller_spawner"
    pkg="controller_manager"
    type="spawner" respawn="false"
    output="screen" ns="/mybot"
    args="joint_state_controller
      rightWheel_effort_controller
      leftWheel_effort_controller"
  />


  <!-- convert joint states to TF transforms for rviz, etc -->
  <node name="robot_state_publisher" pkg="robot_state_publisher" type="robot_state_publisher" respawn="false" output="screen">
    <param name="robot_description" command="$(find xacro)/xacro.py '$(find mybot_description)/urdf/mybot.xacro'" />
    <remap from="/joint_states" to="/mybot/joint_states" />
  </node>

</launch>

We could launch our model on gazebo and then launch the controller, but to save some time (and terminals), we’ll start the controllers automatically by adding a line to the “mybot_world.launch” in the mybot_gazebo package :

<!-- ros_control mybot launch file -->
<include file="$(find mybot_control)/launch/mybot_control.launch" />

Now launch your simulations. In a separate terminal, if you do a “rostopic list” you should see the topics corresponding to your controllers. You can send commands manually to your robot:

  rostopic pub -1 /mybot/leftWheel_effort_controller/command std_msgs/Float64 "data: 1.5"
  rostopic pub -1 /mybot/rightWheel_effort_controller/command std_msgs/Float64 "data: 1.0"

The robot should start moving. Congratulations, you can now control your joints through ROS ! You can also monitor the joint states by doing :

  rostopic echo /mybot/joint_states

Issue 1

Failed to load plugin libgazebo_ros_control.so

the installation of this packages cleared out all errors for me (tested under indigo) :

ros-indigo-youbot-gazebo-robot
ros-indigo-youbot-gazebo-control
ros-indigo-youbot-description
ros-indigo-youbot-driver
ros-indigo-youbot-driver-ros-interface
ros-indigo-youbot-gazebo-worlds
ros-indigo-youbot-simulation
ros-indigo-gazebo-ros-control
ros-indigo-effort-controllers
ros-indigo-joint-state-controller
ros-indigo-joint-trajectory-controller

copy/paste command:

sudo apt-get install ros-indigo-youbot-gazebo-robot ros-indigo-youbot-gazebo-control ros-indigo-youbot-description ros-indigo-youbot-driver ros-indigo-youbot-driver-ros-interface ros-indigo-youbot-gazebo-worlds ros-indigo-youbot-simulation ros-indigo-gazebo-ros-control ros-indigo-effort-controllers ros-indigo-joint-state-controller ros-indigo-joint-trajectory-controller

Issue 2

No valid hardware interface element found in joint

In macro.xacro:

modify

    <transmission name="${lr}_trans">
      <type>transmission_interface/SimpleTransmission</type>
      <joint name="${lr}_wheel_hinge"/>
      <actuator name="${lr}Motor">
        <hardwareInterface>EffortJointInterface</hardwareInterface>
        <mechanicalReduction>10</mechanicalReduction>
      </actuator>
    </transmission>

    <transmission name="${lr}_trans">
      <type>transmission_interface/SimpleTransmission</type>
      <joint name="${lr}_wheel_hinge">
          <hardwareInterface>EffortJointInterface</hardwareInterface>
      </joint>
      <actuator name="${lr}Motor">
        <hardwareInterface>EffortJointInterface</hardwareInterface>
        <mechanicalReduction>10</mechanicalReduction>
      </actuator>
    </transmission>

Teleoperation of your robot

Ok you can control joints individually, but that’s not so convenient when you want to make your mobile robot move around. Let’s use another plugin called differential drive to make it easier. Add this in the gazebo file of your model :

<gazebo>
  <plugin name="differential_drive_controller" filename="libgazebo_ros_diff_drive.so">
    <alwaysOn>true</alwaysOn>
    <updateRate>100</updateRate>
    <leftJoint>left_wheel_hinge</leftJoint>
    <rightJoint>right_wheel_hinge</rightJoint>
    <wheelSeparation>${chassisWidth+wheelWidth}</wheelSeparation>
    <wheelDiameter>${2*wheelRadius}</wheelDiameter>
    <torque>20</torque>
    <commandTopic>mybot/cmd_vel</commandTopic>
    <odometryTopic>mybot/odom_diffdrive</odometryTopic>
    <odometryFrame>odom</odometryFrame>
    <robotBaseFrame>footprint</robotBaseFrame>
  </plugin>
</gazebo>

This plugin will subscribe to the cmd_vel topic specified with the « commandTopic » tag and convert the messages to the proper commands on the wheels. It also provides some odometry data.

Now, you can start gazebo with the usual launch file.

To teleoperate your robot with the keybord you can use a teleoperation node as provided in turtlesim or turtlebot packages. We just need to remap the topic name to connect it to our robot :

    rosrun turtlesim turtle_teleop_key /turtle1/cmd_vel:=/mybot/cmd_vel
    rosrun turtlebot_teleop turtlebot_teleop_key /turtlebot_teleop/cmd_vel:=/mybot/cmd_vel

Enjoy the ride !

Issue

package "turtlebot_teleop" is not found

Fix by typing

sudo apt-get install ros-indigo-turtlebot-teleop

References

[Tutorial] Simulating Sensors in Gazebo (part 2)
Gazebo Plugin Library
Gazebo Camera Tutorial