Krishi Cobot (KC) theme for eYRC 2025-26

This repository contains the simulation setup and hardware implementation codes for the Krishi Cobot (eYantra Robotics Competition 2025-26).

Project Overview

Krishi Cobot (KC) is a ROS2-based autonomous agricultural automation system developed as part of the e-Yantra Robotics Competition (eYRC 2025–26). The project focuses on building a coordinated mobile manipulation pipeline for a smart vertical farm environment.

The system integrates an autonomous mobile robot (eBot) and a UR5 robotic arm to perform crop monitoring, plant health inspection, fertilizer coordination, and defective fruit removal. The mobile robot navigates greenhouse lanes using LiDAR-based localization and shape detection, while the UR5 arm performs vision-based object pose estimation and precise pick-and-place manipulation.

This project combines perception, navigation, manipulation, and inter-robot coordination into a unified ROS2 architecture. The perception pipeline uses ArUco marker detection and depth-based 3D localization, while the control layer integrates waypoint navigation, servo-based arm motion, and state-driven task execution.

System Architecture

The Krishi Cobot system is built on a modular ROS2 architecture integrating perception, navigation, manipulation, and coordination layers. The system consists of two primary robotic subsystems:

Mobile Navigation Subsystem (eBot)

• LiDAR-based autonomous navigation
• Waypoint tracking inside greenhouse lanes
• Real-time shape detection during motion
• Publishes plant condition and task triggers

Vision & Manipulation Subsystem (UR5)

• ArUco marker detection using OpenCV
• Depth-assisted 3D pose estimation for detected objects
• TF2-based frame transformations (camera → base → world)
• Servo-based end-effector control
• Pick-and-place execution for defective fruit removal

Control & Coordination Layer

• State-driven task sequencing
• ROS2 topic-based communication between eBot and UR5
• Trigger-based collaboration logic
• Execution Flow: Navigate to Waypoint → Detect Target (LiDAR / Vision) → Compute 3D Pose (TF Transform) → Execute Manipulation → Update System State

The architecture enables an end-to-end autonomous workflow where the mobile robot navigates and detects plant conditions, then coordinates with the UR5 arm for corrective actions, completing a full agricultural automation cycle.

Implementation Phases

Phase 1– Core Foundations

• 1A - LiDAR-Based Waypoint Navigation
• 1B - Vision-Based Pose Estimation (ArUco + Fruit Detection)
• 1C - Waypoint-Based Servo Control for Robotic Arm Manipulation

Phase 2– Autonomous Integration

• 2A - Autonomous Navigation with Shape Detection
• 2B - Pick and Place Operations
• 3B - eBot-UR5 Arm Collaboration
• 4C – Full Simulation Run

Phase 3– Hardware Deployment

• 3A - Remote Hardware Task - Object Pose Estimation
• 4A - Remote Hardware: UR5 Arm Pick and Place
• 4B - Remote Hardware: Autonomous Navigation with Shape Detection
• 5 - Hardware Mini-Theme Run
• 6 - Full Hardware System Run

Workspace Structure

• HW_5076_eBot_nav/ - Hardware AMR files
• HW_5076_ur5_control/ - Hardware ARM files
• SW_5076_eBot_nav/ - Simulation AMR files
• SW_5076_ur5_control/ - Simulation ARM files
• eBot_description/ - Contains eBot robot description, launch files, and configurations
• eyantra_warehouse/ - Warehouse simulation environment and models
• ur_description/ - Universal Robots arm description
• ur_simulation_gz/ - Gazebo simulation for Universal Robots arm and its controllers
• ur5_control/ - Control packages for UR5 arm
• linkattacher_msgs/ - Custom messages for link attacher

Installation & Setup

Prerequisites

• Ubuntu 22.04 (Jammy Jellyfish)
• ROS 2 Humble
• Gazebo Fortress

Clone Repository

cd ~/eyrc/src
git clone https://github.com/NSGourav/eYRC-2025-26-krishi-cobot.git

Build Repository

cd ~/eyrc
colcon build
source install/setup.bash

• ROS Bag for Testing Hardware: Download ROS Bag

Launch Commands

To launch Gazebo World for Task 4C:

ros2 launch eyantra_warehouse task4c.launch.py

Launch Navigation Module

ros2 launch ebot_nav eBot.launch.py

Launch Manipulation Module

ros2 launch ur5_control arm.launch.py

Run Control file

ros2 run ebot_nav main.py

Results & Performance

• Successfully implemented a complete autonomous mobile–manipulator pipeline in simulation and hardware environments
• Achieved stable LiDAR-based navigation with real-time shape detection
• Implemented reliable ArUco-based 3D pose estimation with TF frame transformations
• Executed coordinated pick-and-place operations using UR5 with servo-based control
• Validated end-to-end greenhouse workflow from navigation to manipulation

Competition Outcome: Ranked 16th out of 400+ teams in the e-Yantra Robotics Competition 2025–26

My Contributions

• Implemented vision-based bad fruit detection and ArUco marker detection with depth-assisted 3D pose estimation
• Built TF2 frame transformation pipeline for accurate coordinate mapping
• Implemented servo-based manipulation logic for UR5 pick-and-place
• Integrated ROS2 topic-based communication between navigation and manipulation modules
• Executed simulation-to-hardware transition and system validation