Lexseal / irobot_create_nav2

create2 nav2 bringup

Navigation Stack

Overview

This package enables the irobot create or the empty belly robot to navigate around using ros2 nav2. In addition, it includes a dockerfile to build and run the entire stack on a jetson (can be easily ported to run on non-arm architectures).

Installation

Install the dependencies.

apt update && \
apt install python3-pip vim zsh -y && \
apt install ros-$ROS_DISTRO-navigation2 ros-$ROS_DISTRO-nav2-bringup ros-$ROS_DISTRO-turtlebot3* -y

Then,

For Empty Belly

1. You will need to get the driver code that enables moving the wheels and sending odometry data.

   mkdir -p ~/ws_nav2/src && \
   git clone --recursive https://github.com/Lexseal/zlac8015d_ros.git && \
   cd ~/ws_nav2/src/zlac8015d_ros/ZLAC8015D_python && \
   pip3 install -e .
   

   The above will install the underlying python communication library for ZLAC8015D, which is the motor driver for the empty belly robot.

2. You will need to clone the lidar package.

   cd ~/ws_nav2/src && \
   git clone -b ros2 https://github.com/Slamtec/rplidar_ros.git
   

3. You will obviously also need this very repository.

   cd ~/ws_nav2/src && \
   git clone -b empty_belly https://github.com/Lexseal/irobot_create_nav2.git
   

4. Build the workspace.

   cd ~/ws_nav2 && \
   source /opt/ros/$ROS_DISTRO/setup.bash && \
   colcon build --symlink-install
   

For iRobot Create

1. Follow https://iroboteducation.github.io/create3_docs/ to get the robot running.

2. You will need to do step 2-4 from the above. Note that when cloning irobot_create_nav2 you should clone the master branch.

Running

There is a tmux script called launch.sh that should bring up everything. Please change the port number in the script as needed.

If you are using docker on jetson, you can pull xinsonglin/jetson_nav:latest and start the container with

docker run -it --network host --device /dev/ttyUSB0:/dev/ttyUSB0 --device /dev/ttyUSB1:/dev/ttyUSB1 -v /dev:/dev --device-cgroup-rule "c 81:* rmw" --device-cgroup-rule "c 189:* rmw" xinsonglin/jetson_nav:latest`

The cgroup rules are needed for realsense camera to work.

Additional Notes

If you are using L515, please see the dockerfile on how to install proper version of the realsense library as it is discontinued.

There is also a script to convert depth image into 2d lidar scan for either obstacle avoidance or SLAM. However, the realsense cameras have a high minimum range, so it is not very useful for obstacle avoidance.

For localization, there are a couple of choices:

1. The default nav2 uses AMCL. This is not accurate enough in practice and also requires an additional map. To build the map, you may run the cartographer package ros2 launch turtlebot3_cartographer cartographer.launch.py resolution:=0.01, drive it around with ros2 run turtlebot3_teleop teleop_keyboard. When you are satisfied with the map, save it with ros2 run nav2_map_server map_saver_cli -f ~/ws_nav2/src/irobot_create_nav2/maps/map.

2. The second option is to use slam_toolbox. This is the current approach.

3. One can also swap out slam_toolbox with cartographer if the odometry is not accurate enough. cartogarpher has their own odometry estimation.