SalviSantilio - Mobile Robotics Project

This repository contains a minimal setup for an Ackermann drive robot using the nav2 stack on ROS 2 Foxy. It provides launch files and parameters to simulate the vehicle in Gazebo and autonomously navigate around a map.

Rationale

The project aims to demonstrate the capabilities of the ROS 2 navigation stack with an Ackermann steering model, which is commonly used in autonomous vehicles.
At first, we considered using the AWS DeepRacer platform, but it was not compatible with ROS 2 Foxy. Therefore, we created a custom Ackermann robot model and navigation stack that works seamlessly with the nav2 stack.
Then, we tried to use the ackermann steering model from the official ROS2 Gazebo / Nav2 stack, but there were no out-of-the-box solutions for Ackermann steering in ROS 2 Foxy.
Thus, we implemented our own custom odometry publisher and control bridge to handle Ackermann steering kinematics, allowing the robot to navigate effectively in simulated environments.

Repository Structure

• src/ackermann_robot_description – URDF model and Gazebo launch files for the robot
• src/ackermann_control_bridge – node that converts /cmd_vel into steering and wheel commands
• src/ackermann_robot_navigation – navigation configuration and launch files
• The robot description now includes a 2D LIDAR publishing /scan and a depth camera publishing /depth/image_raw in Gazebo.

Documentation

You can find detailed documentation for each package in the docs directory:

• ackermann_robot_description – Overview of the Ackermann robot structure and URDF
• ackermann_bridge – Overview of the control bridge node and its functionality
• ackermann_control_bridge – Custom odometry publisher for Ackermann steering
• ackermann_robot_navigation – Instructions for mapping and navigation setup

How the nav2 Stack Works

The nav2 stack is a collection of ROS 2 nodes that provide navigation capabilities:

1. map_server – loads an occupancy grid map for global planning.
2. amcl – localizes the robot using Adaptive Monte Carlo Localization.
3. planner_server – computes paths from the current pose to a goal pose.
4. controller_server – follows the global plan by generating velocity commands.
5. bt_navigator – runs a Behavior Tree (BT) that sequences high‑level navigation behaviors.
6. recoveries_server – executes recovery behaviors when navigation fails.
7. nav2_lifecycle_manager – manages node startup and shutdown.

The behavior tree used here is navigate_w_replanning_and_recovery.xml, which repeatedly replans the path at 1 Hz and performs clearing, spinning, waiting, and backing up if issues occur.

Bringing Up the Navigation Stack

The entire navigation stack, along with the robot simulation and visualization, can be launched using the bringup_all.launch.py file:

ros2 launch ackermann_robot_navigation bringup_all.launch.py

This will:

• Launch the Gazebo simulation with the Ackermann robot.
• Start the control bridge and odometry nodes.
• Bring up the nav2 stack (map server, AMCL, planner, controller, etc.).
• Open RViz2 for visualization.

Quickstart: Install System Dependencies

Before building the workspace, install all required apt packages:

bash install_ros2_foxy_dependencies.sh

Building the Workspace

From the repository root:

source /opt/ros/foxy/setup.bash   # make sure ROS 2 Foxy is installed
colcon build --symlink-install

Running the Simulation and nav2

1. Start Gazebo and the robot

   ros2 launch ackermann_robot_navigation bringup_all.launch.py

2. Start the nav2 stack (in another terminal):

   ros2 launch ackermann_robot_navigation nav2_bringup.launch.py

The robot will read the map in src/ackermann_robot_navigation/maps/ackermann_steering_map.yaml and you can send goals using RViz2 or the /navigate_to_pose action.

The LiDAR and depth camera data is available on /scan and /depth/image_raw.

Sending Navigation Goals

You can send navigation goals using RViz2 or directly via command line:

ros2 action send_goal /navigate_to_pose nav2_msgs/action/NavigateToPose "{pose: {position: {x: 1.0, y: 1.0, z: 0.0}, orientation: {w: 1.0}}}"

3. Visualize in RViz2 (in another terminal):

   ros2 run rviz2 rviz2 -d src/ackermann_robot_navigation/rviz/nav2_default_view.rviz

4. Send goals using the RViz2 interface or command line as shown above.
5. Monitor the robot's state in RViz2, where you can see the robot's position, planned path, and sensor data.

---

Bibliography and References

1. Foxy support on Ackermann
2. AWS Deepracer Navigation Stack
3. ROS 2 Navigation Documentation
4. ROS2 Nav2 Foxy