RR100 Navigation package

Dependencies

This project should only depend on docker to be built and function. You can see install instructions by following this link.

If you want to have access to hardware acceleration inside the container using an nvidia GPU, follow the install instructions for the nvidia-container-toolkit located here.

Building and running the container

First, clone this repository and its submodules :

user@machine:~$ git clone git@github.com:Soralsei/rr100-rhoban.git
user@machine:~$ cd rr100-rhoban
user@machine:~/rr100-rhoban$ git submodule update --init --recursive

Next, if you intend to rebuild the image often, in contexts where a good network connection isn't available (or if you don't have access to the internet at all), pull the required parent docker image from dockerhub:

docker pull ros:noetic

Then, for convenience, a bash script (rr100-rhoban/build_and_run.sh) to build and run the project is provided at the root of the repository.

Usage example :

user@machine:~/rr100-rhoban$ ./build_and_run.sh --image-tag rr100-sim --target simulation -g

Script parameters

-i, --image-tag <tag-name>              Used to specify the docker image to run/build REQUIRED
-c, --container-name <container-name>   Used to specify the docker container name when running the image, OPTIONAL, default: 'ros'
-r, --rebuild                           Rebuild the image, OPTIONAL, default: false
-t, --target [real, simulation]         The build target, OPTIONAL, default: 'real'
-g, --gpu                               Add GPU support, OPTIONAL, default: false
-h, --help                              Show this message

Running the project

Before we start, if you want to have access to the GUI, it is necessary to give access to your X server to the docker container. To do this, a quick and dirty way would be to add the docker user to the authorized users for your X server by executing xhost +local:docker. However, this isn't a safe way of doing this but since we aren't exposing anything to the internet, we can get away with going about it like this. You can also add it to your .bashrc to avoid having to retype it on every reboot.

In simulation

First, in a first terminal in the container, run the following command :

roslaunch rr100_gazebo rr100_<empty|playpen>.launch

Here, you will get a few errors concerning a missing "p gain" parameter :

However these errors are false positives and won't affect the simulated robot. Nonetheless, if you still wish to remove these error messages, you can add following parameters in the configuration file at /opt/ros/$ROS_DISTRO/share/rr100_configuration/rr100_control/control.yaml

gazebo_ros_control: 
  pid_gains:
    front_left_steering_joint:
      p: <p value>
      i: <i value>
      d: <d value>
    # front_left_wheel ...

Warning

Be advised that if you modify these files inside of a container, the changes made will only exist as long as the container exists, meaning that if run a new container, you will have to modify the same files all over again. To address this, one could copy the rr100_configuration package to the project's workspace and modify the configuration files in that copied package. This should give priority to the locally copied package over the preinstalled one, resulting in the changes being permanent

Next, open a new terminal inside the container :

# Outside of the container, to attach a new terminal to the container
docker exec -ti ros bash

Then, run the following command to run the navigation stack without a static map (using the one generated by SLAM) :

# Inside the container
roslaunch rr100_navigation rr100_navigation.launch \
    imu_topic:=imu7/data \
    simulated:=true \
    use_static_map:=false \
    use_pointcloud_nodelets:=false \

If you have a slam_toolbox serialized pose graph (.posegraph/.data format) you can avoid generating a new map by setting launch arguments use_static_map to true (and optionally generate_map to false) and providing a path to the map through the argument map_path

You should now have two windows open :

• A Gazebo window with the simulated environment
• The Rviz visualisation window

In Rviz, you can then click on the 2D Nav Goal tool in the top toolbar then click anywhere and drag the arrow in the desired orientation and the navigation stack will do its best to reach this goal.

Example

With the real robot

Before starting, you will need to be able to communicate with the RR100 by using its name (rr-100-07 in our case). For this you can either configure your DNS to provide that name/address translation, or if you're in a rush, you can add these translation to your /etc/hosts file like so:

# RR100 addresses
192.168.123.123   rr-100-07   # Replace with your local network configuration
192.168.1.102     rr-100-07   # RR100 internal network

Alternatively, you can alter the robot's configuration and set its ROS_IP environment variable to its IP address (LAN or WLAN depending on which network you're connected to).

Next, because your computer's clock and the robot's computer's clock will most likely not be synchronized, you will need to configure an NTP client to synchronize to the robot's chrony NTP server.

First, install chrony:

sudo apt install chrony

Then, you will have to alter the default /etc/chrony/chrony.conf and add either the robot's IP addresses or the robot's name (since we added it to our hosts file, this will also work). After that, you will likely have to comment-out the other ntp server pools present in the configuration file :

# Use servers from the NTP Pool Project. Approved by Ubuntu Technical Board
# on 2011-02-08 (LP: #104525). See http://www.pool.ntp.org/join.html for
# more information.
## Comment these out #############
#pool 0.ubuntu.pool.ntp.org iburst
#pool 1.ubuntu.pool.ntp.org iburst
#pool 2.ubuntu.pool.ntp.org iburst
#pool 3.ubuntu.pool.ntp.org iburst
##################################

Finally, start/restart the chrony service to take these changes into effect:

sudo systemctl restart chrony.service

To execute the navigation package on the real robot, you must first build the docker image with the target real by using the provided script in a similar fashion :

./build_and_run.sh --image-tag rr100-real --target real -g -a <YOUR IP ADDRESS>

With this command, the container should be automatically built and then run. Make sure you're connected to the same network as the robot or directly connected to the robot's hotspot and then, in the newly running container, launch the navigation package like this :

roslaunch rr100_navigation rr100_navigation.launch \
    use_pointcloud_to_laserscan_nodelet:=true \
    nodelet_manager:=rslidar_nodelet_manager \
    use_static_map:=false

After running the command, an Rviz window will open and you will be able to control the robot remotely like described previously in simulation.

Example