Autonomous Delivery Bot: ROS Navigation + LiDAR Mapping DIY

Building an autonomous delivery bot with ROS and LiDAR is one of the most rewarding robotics projects you can tackle in 2024. Whether you’re a student, hobbyist, or engineer exploring last-mile automation, this guide walks you through every layer — from chassis selection to SLAM-based navigation — using components you can source right here in India. By the end, your bot will map its environment, plan collision-free paths, and deliver payloads without any human input.

What Is an Autonomous Delivery Bot?

An autonomous delivery bot is a ground vehicle that uses sensors, onboard compute, and software algorithms to navigate from point A to point B without human guidance. Commercial versions like Starship Technologies’ sidewalk robots and Amazon Scout use multi-layer sensor fusion. A DIY version built around the Robot Operating System (ROS) can achieve the same fundamental capability at a fraction of the cost.

The core pipeline is:

Perception — LiDAR scans the environment and builds a 2D or 3D occupancy map.
Localisation — The bot estimates its own position within that map using AMCL (Adaptive Monte Carlo Localisation).
Planning — The move_base stack generates a global path and local obstacle-avoidance trajectory.
Execution — Velocity commands flow to motor drivers, turning planned motion into physical movement.

For a typical indoor delivery scenario — think office corridors or college campus paths — a 2D LiDAR + differential-drive chassis is sufficient and cost-effective.

Hardware You Need

Here is a practical Bill of Materials (BOM) for a mid-range autonomous delivery bot build:

Compute: Raspberry Pi 4B (4GB) or NVIDIA Jetson Nano
LiDAR: RPLIDAR A1 or A2 (360° 2D scan, 6–10 m range)
Chassis: 4-wheel or mecanum-wheel platform with enough payload area
Motor Driver: L298N or TB6612FNG for differential drive
Power: 3S LiPo (11.1 V) or 18650 Li-ion pack with 5V BEC
IMU: MPU6050 for odometry fusion
Camera (optional): Raspberry Pi Camera Module for object detection
Encoder Motors: 12V DC motors with quadrature encoders for precise odometry

4 Wheels Car Chassis Acrylic Frame

Sturdy acrylic 4-wheel chassis — ideal base for your delivery bot with ample mounting space for LiDAR, Raspberry Pi, and payload compartment.

View on Zbotic

80mm-A Mecanum Wheel (Pack of 4) – Black

Mecanum wheels allow omnidirectional movement — perfect for tight indoor corridors where a delivery bot needs to navigate without turning in place.

View on Zbotic

Setting Up ROS on Raspberry Pi / Jetson Nano

For a Raspberry Pi 4, use ROS Noetic on Ubuntu 20.04 (64-bit). For Jetson Nano, JetPack 4.6 ships with Ubuntu 18.04, making ROS Melodic the natural choice. Here’s the condensed setup flow:

# On Ubuntu 20.04 (RPi 4)
sudo sh -c 'echo "deb http://packages.ros.org/ros/ubuntu focal main" > /etc/apt/sources.list.d/ros-latest.list'
sudo apt-key adv --keyserver hkp://ha.pool.sks-keyservers.net:80 --recv-key C1CF6E31E6BADE8868B172B4F42ED6FBAB17C654
sudo apt update && sudo apt install -y ros-noetic-desktop
echo "source /opt/ros/noetic/setup.bash" >> ~/.bashrc
sudo apt install -y python3-rosdep python3-rosinstall
sudo rosdep init && rosdep update

Create a catkin workspace and install the navigation stack:

mkdir -p ~/catkin_ws/src && cd ~/catkin_ws
catkin_make
sudo apt install -y ros-noetic-navigation ros-noetic-slam-gmapping ros-noetic-robot-localization

Your workspace is the home for all custom packages — robot description (URDF), launch files, and configuration YAML files for move_base.

Integrating LiDAR for Environment Mapping

The RPLIDAR A1 connects via USB and publishes /scan LaserScan messages at 5–10 Hz. Install the ROS driver:

sudo apt install -y ros-noetic-rplidar-ros
# Grant USB permission
sudo chmod 777 /dev/ttyUSB0
# Launch the driver
roslaunch rplidar_ros rplidar.launch

Verify the scan in RViz by adding a LaserScan display on topic /scan. You should see a 360° ring of distance readings. Mount the LiDAR at the highest point of your chassis, away from motor noise and metal interference, for the cleanest readings.

Coordinate Frames: Define a URDF with a base_link and laser frame, linked by a fixed transform. Publish this transform with a static transform publisher:

rosrun tf static_transform_publisher 0 0 0.1 0 0 0 base_link laser 100

SLAM (Simultaneous Localisation and Mapping) lets the bot build a map while figuring out where it is within that map — chicken-and-egg solved simultaneously. GMapping is the easiest SLAM algorithm to start with on a 2D LiDAR bot:

roslaunch slam_gmapping slam_gmapping.launch
# Save map after exploring
rosrun map_server map_saver -f ~/maps/my_map

Once the map is saved, switch to navigation mode with AMCL for localisation:

roslaunch my_bot navigation.launch map_file:=/home/pi/maps/my_map.yaml

The move_base node ties together the global planner (Dijkstra), local planner (DWA — Dynamic Window Approach), costmaps, and recovery behaviours. Key YAML parameters to tune:

inflation_radius: Set slightly larger than bot width for safe margins.
max_vel_x: Keep at 0.3–0.5 m/s indoors to avoid overshooting.
xy_goal_tolerance: 0.05 m for precise docking.
sim_time (DWA): 2–3 s lookahead for smooth trajectories.

Send navigation goals via rviz (2D Nav Goal button) or programmatically through an actionlib client in Python.

Motor Control & Drive System

The ROS cmd_vel topic (geometry_msgs/Twist) carries linear and angular velocity commands. A microcontroller node (Arduino or ESP32 running rosserial) subscribes to cmd_vel and translates it to PWM signals for the motor driver.

// Arduino rosserial sketch snippet
#include <ros.h>
#include <geometry_msgs/Twist.h>
void cmd_velCB(const geometry_msgs::Twist& msg){
  float linear = msg.linear.x;
  float angular = msg.angular.z;
  int leftPWM  = constrain((linear - angular * WHEEL_BASE/2) * PWM_SCALE, -255, 255);
  int rightPWM = constrain((linear + angular * WHEEL_BASE/2) * PWM_SCALE, -255, 255);
  setMotors(leftPWM, rightPWM);
}

Encoder feedback is critical — without it, odometry drifts and localisation breaks down within metres. Connect encoder channels to interrupt pins on your microcontroller and publish /odom using nav_msgs/Odometry.

80mm-A Mecanum Wheel (Pack of 4) – Blue

Colour-coded mecanum wheels make assembly easier — blue for left, black for right — while providing omnidirectional motion for your autonomous bot.

View on Zbotic

4mm Hex Coupling for Robot Smart Car Wheel (30mm)

Secure your wheels to motor shafts with precision hex couplings — essential for consistent odometry and reliable drive performance.

View on Zbotic

Payload Delivery Mechanism

The delivery mechanism is what differentiates a delivery bot from a regular autonomous rover. Common approaches include:

Servo-actuated tray: An SG90 or MG996R servo tilts the payload tray to drop items at the destination.
Electromagnet latch: Energise a 12V electromagnet to hold the payload; cut power at destination to release.
Linear actuator: Push a drawer-style compartment open using a small linear actuator.

A ROS action server running on the Pi can trigger delivery upon reaching the goal:

if result.status == 3:  # SUCCEEDED
    servo_pub.publish(90)  # open tray
    rospy.sleep(3)
    servo_pub.publish(0)   # close tray

TowerPro SG90 180 Degree Rotation Servo Motor

Reliable, lightweight servo for actuating your delivery tray or latch mechanism — easy to control via Arduino or Raspberry Pi GPIO.

View on Zbotic

Frequently Asked Questions

Can I run ROS on an Arduino alone?

No. Arduino can run rosserial as a ROS node for motor control and sensor reading, but the full ROS navigation stack (GMapping, AMCL, move_base) requires a Linux SBC like Raspberry Pi or Jetson Nano.

What is the range of RPLIDAR A1?

The RPLIDAR A1 has a typical range of 6 m (up to 12 m in ideal conditions) and a 360° field of view at 5–10 Hz. For indoor delivery in corridors and rooms, this is more than adequate.

How accurate is GMapping SLAM?

GMapping achieves centimetre-level map accuracy in small to medium indoor environments (up to ~50m×50m). For larger spaces, consider Cartographer or RTAB-Map.

Do I need encoder motors?

Encoder feedback is strongly recommended. Without it, odometry is purely time/velocity-based and drifts significantly, making AMCL localisation unreliable after a few metres of travel.

What battery life can I expect?

A 3S 5000mAh LiPo typically gives 1–2 hours of autonomous operation depending on motor load, compute, and LiDAR power draw (~2.5W for RPLIDAR A1).

Ready to Build Your Autonomous Delivery Bot?

All the components you need — chassis, mecanum wheels, hex couplings, and servo motors — are available at Zbotic.in with fast shipping across India. Start with the chassis and wheels, get your ROS stack running in simulation first, then integrate hardware iteratively.

Shop Robotics Components at Zbotic