From mapping to navigation in ROS 2 — a minimal Jazzy and Nav2 procedure

Contents — find the section you need

Change parameters and verify

Open the panel, then press Run to load Python. You can stop execution and reset parameters. Results are computed on this device. No Python installation is required.

Local execution steps below are optional for reproducing the source results; they are not required for the browser experiment.

Open experiment panel in a new tab

Download reproduction source

Background and reading guide

Background A browser for this Lab; Ubuntu 24.04, ROS 2 Jazzy and TF knowledge for the optional runtime procedure.

Hands-on · Procedure checked against official sources; Gazebo and physical driving were not executed for this article.

Explore a map, pose and goal in your browser

Change the controls above and compare A/B results. No ROS installation is needed. This exercise checks a fixed synthetic map and supplied points; it does not run SLAM, AMCL, Nav2 or Gazebo. The real-runtime procedure below is optional.

Try mapping, then saved-map mode. Mapping uses an already prepared grid: no observations update it. Saved-map mode requires the modeled image/YAML pair; choosing either file alone hides the map and makes occupancy checks unavailable. These are prepared file-availability cases, not a save operation or an image/YAML parser. Both modes use the same fixture. The expected map→odom owner changes from SLAM to AMCL. Duplicate or missing publishers are independent configuration warnings; no transform is executed. odom→base_link is assumed to have one odometry source.

The grid contains 12 × 10 cells, resolution 0.5 m, origin (−1, −1) m and origin yaw 0. Values are 0=free, 100=occupied and −1=unknown in this fixture. Cells are stored with x changing fastest, starting at the bottom-left; the display flips y for the screen. Coordinates are in the map frame. Column=floor((x+1)/0.5), row=floor((y+1)/0.5); the map covers −1≤x<5 and −1≤y<4. A point exactly on an internal boundary belongs to the cell above/right. The data download retains cell indices and values. Gray question marks are unknown cells, not free space.

The blue circle/heading is a supplied pose, not a localization estimate or ground truth. In saved-map mode it represents an initial-pose seed; a supplied seed does not establish convergence. Heading is counterclockwise from +x. The red cross is a goal point without a goal orientation. With no pose supplied, distance and pose occupancy are unavailable. Straight-line distance is Euclidean separation between supplied points, even when outside the map; it is not path length. The default points are free but the straight segment between them crosses a wall. No segment, connectivity, footprint, inflation, collision, path or arrival check runs. Passing the independent checks is not proof of navigation readiness.

Use the unknown, occupied and outside presets to see distinct goal states. A free cell alone does not establish footprint clearance or reachability. A/B starts fresh and preserves the fixed map and coordinate scale. No uploads, map saving, sensor simulation or real-device performance are included.

Model references: ROS grid ordering, map metadata, Nav2 map loading, SLAM mapping roles and REP-105 frame authorities. Reviewed 2026-09-20. The fixture uses already classified cells; it does not reproduce image thresholds or map-loader behavior. See also the separate replay-condition Lab.

For a first Nav2 experiment, separate mapping from navigation on a saved map using the official TurtleBot simulation. This procedure follows Jazzy documentation checked on September 7, 2026. Gazebo was not run in the writing environment.

Record the environment

Use Ubuntu 24.04 with a GUI, ROS 2 Jazzy and its compatible Gazebo environment. With ROS 2 already installed, prepare packages following the Nav2 Quickstart. Do not mix Humble-era Gazebo Classic instructions into this setup.

source /opt/ros/jazzy/setup.bash
sudo apt update
sudo apt install ros-jazzy-navigation2 ros-jazzy-nav2-bringup ros-jazzy-slam-toolbox ros-jazzy-nav2-minimal-tb3-sim
dpkg-query -W ros-jazzy-navigation2 ros-jazzy-nav2-bringup ros-jazzy-slam-toolbox ros-jazzy-nav2-minimal-tb3-sim
ros2 launch nav2_bringup tb3_simulation_launch.py --show-args

Save the installed versions and verify slam, map and use_sim_time arguments against the Jazzy launch source.

Start mapping mode

Connect Gazebo, SLAM Toolbox and Nav2, save the map, then restart with AMCL

Use a dedicated ROS environment without physical robots. The domain ID below is an example; confirm that no real robot shares it. Use the same setup in each new terminal.

source /opt/ros/jazzy/setup.bash
export ROS_DOMAIN_ID=73
ros2 launch nav2_bringup tb3_simulation_launch.py slam:=True headless:=False use_sim_time:=True

In the official SLAM procedure, SLAM supplies /map and map → odom. Do not add AMCL as a second publisher of that transform during mapping.

Try nearby goals

In another terminal, source the environment and set export ROS_DOMAIN_ID=73, then inspect the interfaces.

ros2 topic list
ros2 topic hz /scan
ros2 run tf2_ros tf2_echo map base_link

Check that the map, LaserScan and robot pose agree in RViz. Set a Nav2 Goal in nearby observed free space, then move toward the next area. A goal deep inside unknown space can fail planning even while SLAM works correctly.

Save the map and restart

From another terminal with the same environment, save into a new working directory. Choose a filename that will not overwrite a map you need.

mkdir -p nav2-map-lab
cd nav2-map-lab
ros2 run nav2_map_server map_saver_cli -f map --ros-args -p use_sim_time:=true

Keep both map.yaml and its referenced image. Stop the original launch with Ctrl+C and wait for shutdown. From the nav2-map-lab directory, restart with the saved map.

ros2 launch nav2_bringup tb3_simulation_launch.py slam:=False map:="$PWD/map.yaml" headless:=False use_sim_time:=True

Use RViz's 2D Pose Estimate to set the robot's current simulated location and heading. Request a nearby goal and check both path generation and arrival. Saved-map localization serves a different purpose from updating a map with SLAM.

Diagnose failures by interface

Symptom	First check
Missing robot or world	Simulator logs and installed models
Missing map	`/scan`, SLAM node and simulation clock
Missing TF	Shared domain, timestamps and frame names
Misaligned saved map	YAML image reference and initial pose
No path	Known free-space goal and robot footprint clearance

Keep versions, commands, maps and success or failure logs at each stage. The newbot implementation record includes 3D LiDAR and a different odometry source; identify each change from this minimal setup as you read it.