Robotics

Autonomous Navigation Robot

ROS 2 mobile robot with mapping, localization and autonomous navigation

ROS 2
Nav2
SLAM Toolbox
Gazebo
Raspberry Pi
Arduino

Problem Statement

Indoor mobile platforms in institutions and small factories still rely on manual driving or fixed line-following, which cannot adapt to changing layouts, dynamic obstacles or rescue scenarios.

Objectives

Build a ROS 2 differential-drive robot capable of autonomous point-to-point navigation.
Generate reliable 2D maps of indoor environments.
Achieve stable localization and dynamic obstacle avoidance.
Create a modular base for future Surveillance and Rescue payloads.

Hardware Used

Raspberry Pi 4 (ROS 2 compute)
Arduino motor controller
2D LiDAR
DC geared motors with quadrature encoders
Motor driver and Li-ion power distribution

Software Used

ROS 2
Nav2
SLAM Toolbox
Gazebo
RViz
Python
C++
Linux

System Architecture

Sensor layer: LiDAR scans and wheel encoders publish /scan and /odom.
State layer: robot_state_publisher and TF tree maintain frames.
Mapping layer: SLAM Toolbox builds and serves the occupancy grid.
Navigation layer: Nav2 planner, controller and behaviour trees output /cmd_vel.
Actuation layer: microcontroller firmware converts velocity to PWM.

Challenges

TF timing mismatches causing localization drift.
Noisy encoder odometry on uneven floors.
Tuning Nav2 costmaps for narrow corridors.

Solutions

Synchronised timestamps and corrected the TF tree hierarchy.
Applied odometry filtering and recalibrated wheel parameters.
Tuned inflation radius, footprint and controller gains iteratively in Gazebo before hardware runs.

Key Features

2D occupancy mapping with SLAM Toolbox
AMCL localization and Nav2 path planning
Gazebo simulation before hardware deployment
Differential-drive base with encoder odometry
Obstacle avoidance using LiDAR scans

Future Improvements

Integrate a camera for Surveillance and Rescue (SAR) mode.
Add YOLO-based human detection for rescue alerts.
Multi-floor navigation and fleet coordination.