Showcase: Beam AgroBot V2 – Autonomous Agricultural Robot Built with ROS 2, Nav2, MoveIt 2 and RealSense

Projects
, , ,
1

image
image1920×814 148 KB

image
image786×576 57.8 KB

Nav2 Demo

image
image1920×1395 278 KB

image
image1920×878 524 KB

MoveIt2 Demo

image
image640×480 39.7 KB

Hello everyone,

I’d like to share Beam AgroBot V2, an autonomous agricultural robot that I’ve been developing over the past several months as a personal robotics project.

My name is Hatim Ahmed Hassan, a Mechanical Engineering graduate with a strong interest in robotics, autonomous systems, and embedded development. The primary objective of this project has been to gain a deep understanding of how the different components of a modern robotic system work together—from mechanical design and robot modeling to navigation, perception, manipulation, and system integration.

Beam AgroBot V2 is currently built entirely on ROS 2 Humble and serves as the simulation platform before transitioning to physical hardware.

Repository

GitHub:

Current Features

  • Autonomous navigation using Nav2
  • Mapping with SLAM Toolbox
  • Localization using AMCL
  • 5-DOF robotic manipulator controlled with MoveIt 2
  • TRAC-IK inverse kinematics solver
  • RGB-D crop localization using an Intel RealSense depth camera
  • URDF/Xacro robot description
  • ROS 2 Control differential drive controller
  • Gazebo simulation environment
  • End-to-end autonomous pipeline for navigation, crop detection, arm motion, and precision watering

Project Goal

Rather than building isolated demonstrations, my goal was to integrate the major components of a complete autonomous robotic system into a single platform.

The project combines mechanical design, simulation, robot description, navigation, perception, manipulation, and control into one ROS 2 workspace. Although the current version is simulation-based, the software architecture has been designed with future deployment to physical hardware in mind.

Looking for Feedback

I would greatly appreciate feedback from the ROS community, particularly regarding:

  • Overall ROS 2 system architecture
  • Package organization and maintainability
  • Navigation pipeline and Nav2 integration
  • MoveIt 2 integration and manipulation workflow
  • Computer vision and perception pipeline
  • General ROS best practices
  • Design decisions that could be improved before moving to hardware

I’m not currently looking for collaborators on this project because one of my primary goals is to learn each subsystem by implementing it myself. However, I would genuinely appreciate constructive criticism, suggestions, and recommendations from experienced ROS developers that could help improve both the project and my understanding of robotics.

Thank you for taking the time to read through this. Any feedback, whether positive or critical, is greatly appreciated.

1 Like
ROS News for the Week of July 6th, 2026
2

This looks interesting and smaller farmers are really eager after any technology that would ease them making a high-quality low-pesticide food. Precision watering is definitely one of the areas of active interest.

Regarding the design - the wheels don’t seem to be designed to move in mud. You need mich wider tyres. Especially the shopping-cart front wheels.

Also, the idea from the render that the robot can turn mid-lane is not realistic. The robot has to go absolutely straight inside the lane and can only rotate at the ends (imagine how heavy the robot will be with full tank). Otherwise, the lane would quickly become an offroad driving arena. Nav2 already has some kind of support for this I think.

Regarding the 2D lidar - unless it’s just for a field behind your home, it’s insufficient. Don’t hold yourself back and invest into some 3D lidars. You can get a Livox for about $600, which is basically nothing in the agri world. And you also need to think about safety and security, so at least one lidar front, one lidar back. And regarding safety, it should also have active bumpers which will stop the robot before it bumps into people.

3

appreciate the detailed feedback. you’re completely right about the caster wheels—I’m actually planning to swap to tank treads for V3 to handle the mud and weight better.

speaking of the terrain, what you said about mid-lane turns tearing up the ground makes total sense. I didn’t realize Nav2 had specific support for strict straight-line navigation, so I’ll definitely dig into that configuration.

on the hardware side, 3D lidar is already on my upgrade list, but $600 for a Livox is a really solid price point. adding a rear lidar and active bumpers for safety is a great call too, I’ll definitely add those to the requirements before I start buying parts.

4

being at early scale is thats too good progress, i am working on something similar a 5dof arm on a tracked based rover more likely same setup stereo cam and lidar, thats real we require a 3d lidar for more prescise persepction. If you are going into farm fields then the caster wheel won’t work their.

but great work their lets do somthing togther like i into mechatronics lets build it