Mahmoud Abas, Full-Stack Software Engineer & AI

ROS 2 Practical - TurtleBot3 Lab 06 - SLAM (Simultaneous Localization and Mapping)

In the print dialog, choose "Save as PDF" as the destination.

ROS2 Practical - Lab 06

TurtleBot3 Lab 06: SLAM

Simultaneous Localization and Mapping with SLAM Toolbox.

Learning Objectives

Understand the fundamentals of SLAM algorithms
Build real-time maps from laser scanner data
Learn coordinate frame transformations in robotics
Practice teleop-driven exploration
Visualize mapping data live in RViz

Lab Setup

bash

sudo apt install ros-jazzy-slam-toolbox
export TURTLEBOT3_MODEL=waffle_pi

ros2 launch turtlebot3_lab06 lab06.launch.py

Terminal 1: install SLAM Toolbox and launch the lab

In a second terminal, drive the robot to explore:

bash

ros2 run turtlebot3_teleop teleop_keyboard

Terminal 2: teleop keyboard control

Required Step

Activating SLAM Toolbox

unconfigured

configure

inactive

activate

active [3]

bash

ros2 lifecycle get /slam_toolbox
ros2 lifecycle set /slam_toolbox configure
ros2 lifecycle set /slam_toolbox activate

Run the transition

bash

ros2 lifecycle get /slam_toolbox   # should show: active [3]
ros2 topic echo /map --once

Verify: state should read active [3]

What You Should See

Gazebo: robot responds to teleop, red laser lines from the scanner
RViz Map display, updating live:

Color	Meaning
Gray	Unknown
White	Free space
Black	Obstacle

Robot Model + Laser Scan + Path (green trajectory)
TF frames: map to odom to base_link

SLAM Concepts in Action

Real-Time Mapping: The map updates about every 0.1s; laser data becomes an occupancy grid.

Localization: Odometry combined with scan matching estimates the robot's position within the growing map.

Loop Closure: Revisiting a mapped area corrects accumulated drift.

Sensor Integration: 2D LiDAR + IMU + wheel encoders all feed the position estimate.

Coordinate Frames

map

global frame, corrected by loop closure

odom

drifts over time

base_link

robot's own local frame

Sync vs Async SLAM Toolbox

	Sync	Async
Determinism	Deterministic, repeatable	Non-deterministic
Debugging	Simple, sequential	Harder, multi-threaded
Performance	Single-core, higher latency	Multi-core, lower latency
Best for	Learning, limited resources	Production, high-frequency sensors

This lab uses Sync SLAM Toolbox for educational reliability
Both variants still require lifecycle configure + activate

Key SLAM Parameters

Parameter	Value
Resolution	`0.05m` per pixel (`5cm` grid)
Update rate	`50Hz` transforms, `2Hz` map updates
Scan range	up to `12m` (TurtleBot3 Waffle Pi LDS)
Loop closure	enabled for map consistency

Troubleshooting

"No map received" in RViz: SLAM Toolbox was not activated. Run configure then activate, and confirm the state reads active [3].

Map not updating: SLAM needs motion. Check that /scan is publishing and the laser is visible in Gazebo.

Robot not responding to teleop: Verify TURTLEBOT3_MODEL is set and /cmd_vel exists.

RViz frame errors: Fixed Frame should be odom. Use view_frames to debug the TF tree.

bash

ros2 lifecycle set /slam_toolbox configure
ros2 lifecycle set /slam_toolbox activate
ros2 run tf2_tools view_frames

Recover and re-check the TF tree

Key Takeaways

SLAM Toolbox nodes are lifecycle nodes: unconfigured to inactive to active, manual activation required every launch
SLAM builds a map and tracks the robot's position in it simultaneously, fusing LiDAR, IMU, and odometry
Loop closure is what corrects drift when the robot revisits known territory
Sync SLAM trades performance for determinism and easier debugging, the right choice while learning
map, odom, and base_link form the core TF chain that every SLAM/navigation feature depends on
Motion is required to build a map: a stationary robot produces no new map data

Open this lesson

Mahmoud AbasTurtleBot3 Lab 06 - SLAM (Simultaneous Localization and Mapping)

Color

Meaning

Gray

Unknown

White

Free space

Black

Obstacle

Sync

Async

Determinism

Deterministic, repeatable

Non-deterministic

Debugging

Simple, sequential

Harder, multi-threaded

Performance

Single-core, higher latency

Multi-core, lower latency

Best for

Learning, limited resources

Production, high-frequency sensors

Parameter

Value

Resolution

0.05m per pixel (5cm grid)

Update rate

50Hz transforms, 2Hz map updates

Scan range

up to 12m (TurtleBot3 Waffle Pi LDS)

Loop closure

enabled for map consistency