ROS 2D Navigation Example

This respository contains example code, configuration, and launch files to demonstrate how to use the ROS navigation stack using a Gazebo simulation.

The contents of this repository are meant to accompany lectures on the topic. The official documentation for the ROS navigation stack is available here: http://wiki.ros.org/navigation. This is a good resource to use to expand upon the example code in this repository to implement the navigation on a real robot.

diff_robot_gazebo Package

This package implements a differential drive robot model for Gazebo and implements a simulated drive control system. The drive control plugin emulates a similar ROS topic interface to drive control systems on real robots.

In addition to the Gazebo robot model, this package contains a launch file called run_simulation.launch that starts Gazebo and spawns the robot model. By default, it also loads a world containing a maze for the vehicle to navigate:

There are a few arguments to run_simulation.launch, but the one that is meant to be changed from default is the ground_truth_mode argument.

roslaunch diff_robot_gazebo run_simulation.launch ground_truth_mode:=true

or

roslaunch diff_robot_gazebo run_simulation.launch ground_truth_mode:=false

When ground_truth_mode is set to true, a TF frame and corresponding nav_msgs/Odometry message is published for the vehicle's current position and orientation relative to its position in the Gazebo world. This represents ideal ground truth localization information that can be used to focus on path planning algorithms without having to worry about localization.

When ground_truth_mode is set to false, the TF frame and odometry message are not published at all. To successfully navigate the maze in this mode, a localization system using LIDAR SLAM is necessary.

diff_drive_control Package

This package contains an example of how to implement the drive control of a differential drive mobile robot. The example node subscribes to a geometry_msgs/Twist topic and uses the speed and yaw rate values to generate left and right wheel speeds for the robot.

dead_reckoning_odom Package

This package contains an example of how to implement a discrete state space model, and how to apply it to the task of dead reckoning navigation. The example node subscribes to a geometry_msgs/TwistStamped topic containing the measured speed and yaw rate of a robot, and produces an estimate of the vehicle's position and orientation in a global frame. This is done by recursively updating a discrete state space model using the speed and yaw rate data.