AI-powered robot learns how to harvest tomatoes more efficiently

2026-03-18 · Source: Artificial Intelligence News -- ScienceDaily · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Intermediate, quick

Summary

Researchers at Osaka Metropolitan University, led by Assistant Professor Takuya Fujinaga, have developed a new AI-powered tomato-picking robot that significantly improves harvesting efficiency. Unlike traditional systems that merely identify ripe fruit, this robot employs "harvest-ease estimation" to predict the difficulty of picking each tomato. It uses image recognition and statistical analysis to assess visual details like stems and leaf obstructions, then adjusts its approach, including changing angles, to optimize success. This intelligent strategy achieved an 81% success rate in testing, with a quarter of successful picks resulting from adaptive angle changes after initial attempts. The system represents a step towards collaborative farming where robots handle easier tasks and humans focus on more challenging ones.

Key takeaway

For AI Scientists developing agricultural robotics, this research demonstrates the value of moving beyond simple object detection to incorporate predictive "harvest-ease" metrics. Your next-generation systems should integrate statistical analysis with visual data to enable robots to dynamically adjust their approach, significantly boosting success rates in complex environments like clustered crops. Consider designing for adaptive angle changes to improve real-world performance and facilitate human-robot collaboration.

Key insights

Robots can improve harvesting efficiency by assessing "harvest-ease" and adapting their picking strategy.

Principles

• Robots can learn to predict task difficulty.
• Adaptive strategies improve robotic task success.

Method

The robot combines image recognition with statistical analysis to determine the optimal picking angle by assessing visual details like tomato position, stems, and leaf obstructions, enabling "harvest-ease estimation."

In practice

• Implement "harvest-ease" metrics for robotic tasks.
• Design robots for adaptive, multi-angle approaches.

Topics

• Agricultural Robotics
• Robotic Harvesting
• Image Recognition
• Decision-Making AI
• Human-Robot Collaboration

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, AI Engineer, Robotics Engineer, AI Researcher

Related on AIssential

• Eternal.ag raises €8M to automate greenhouse harvesting with AI-powered robots — AI-powered autonomous robots are critical for addressing greenhouse labor shortages and ensuring year-round food supply.
• A multi-armed robot for assisting with agricultural tasks — Robots can safely manipulate delicate plant branches using force-aware motion planning and real-time feedback.
• Improving Human-Robot Teamwork in Urban Search and Rescue Through Episodic Memory of Prior Collaboration — Robots can significantly improve human-robot teamwork by leveraging human-externalized collaboration patterns as episodic memory.
• An Intelligent Robotic and Bio-Digestor Framework for Smart Waste Management — An integrated robotic and bio-digestor framework automates waste segregation and optimizes biological conversion for smart waste management.
• Robots run this laboratory in Japan — and are changing how scientists work — AI-powered laboratory robots can autonomously manage complex biological experiments, accelerating discovery and freeing human researchers.
• RS-Claw: Progressive Active Tool Exploration via Hierarchical Skill Trees for Remote Sensing Agents — Active tool exploration via hierarchical skill trees improves remote sensing agent performance and context efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence News -- ScienceDaily.