Autonomous Navigation in ROS2

In this work we built a differential drive robot with online asynchronous SLAM and autonomous navigation with ROS2 and Gazebo.

Project Page

System Architecture

The system is split into three independent launch domains that communicate via ROS2 topics, services, and actions:

1. Simulation & Bringup — spawns the robot in Gazebo, runs the differential drive controller, bridges Gazebo sensors to ROS topics, and arbitrates velocity commands.
2. SLAM — builds an occupancy grid from LiDAR scans and publishes the map.
3. Navigation (Nav2) — plans and executes paths based on the map and goal poses.

Prerequisites

• ROS2 Jazzy (other versions may require adaptations)
• slam_toolbox
• nav2
• colcon

Installation

1. Build the workspace:

colcon build --symlink-install

2. Source the setup files:

source /opt/ros/jazzy/setup.bash
source install/setup.bash

Quick Start

Launch the simulation with a custom world:

ros2 launch my_bot launch_sim.launch.py world:=src/my_bot/worlds/obstacles.world

Start the online asynchronous SLAM node:

ros2 launch slam_toolbox online_async_launch.py \
  use_sim_time:=true \
  slam_params_file:=./src/my_bot/config/mapper_params_online_async.yaml

Open RViz:

rviz2 -d src/my_bot/config/drive.rviz

In RViz, click "2D Pose Estimate" to provide an initial pose estimate for localization.

Navigation

Start the Nav2 navigation stack:

ros2 launch nav2_bringup navigation_launch.py use_sim_time:=true

You can now set navigation goals in RViz.

Key Topics & Actions

Name	Type	Publisher → Subscriber
/scan	sensor_msgs/LaserScan	Gazebo GPU LiDAR (via ros_gz_bridge) → SLAM, Nav2 costmaps, RViz
/camera/image_raw, /camera/camera_info	sensor_msgs/Image, CameraInfo	Gazebo camera (via ros_gz_bridge) → RViz display
/joint_states	sensor_msgs/JointState	Joint state broadcaster → Robot State Publisher
/cmd_vel	geometry_msgs/Twist	RViz teleop panel / keyboard / Nav2 → twist_mux → twist_stamper
/diff_cont/cmd_vel	geometry_msgs/TwistStamped	twist_stamper → differential drive controller
/odom	nav_msgs/Odometry	Differential drive controller → SLAM, Nav2, RViz
/tf, /tf_static	tf2_msgs/TFMessage	Diff controller (odom→base_link), RSP (fixed frames), SLAM (map→odom) → all nodes
/map	nav_msgs/OccupancyGrid	SLAM → Nav2, RViz
/map_updates	map_msgs/OccupancyGridUpdate	SLAM → RViz display
/initialpose	geometry_msgs/PoseWithCovarianceStamped	RViz "2D Pose Estimate" → SLAM
/goal_pose	geometry_msgs/PoseStamped	RViz "2D Pose Goal" → Nav2 action server
navigate_to_pose	nav2_msgs/NavigateToPose (action)	RViz goal tool / client → Nav2 bt_navigator

Acknowledgments

This project is inspired by the excellent ROS2 robot tutorial series by Josh Newans. See the full playlist for more details.