Oregon State University · MIME

Robots for productive, resilient, and lower-energy field operations.

The Sustainable Field Robotics Lab develops intelligent, energy-aware robotic systems for agriculture and other demanding outdoor environments. We combine machine vision, robotic manipulation, artificial intelligence, and renewable-energy-aware design to reduce the energy, labor, and environmental cost of field operations.

SFRL robot with articulated arm reaching into apple tree canopy in a commercial orchard
20+Peer-reviewed journal papers
5,800+Citations
3Invention disclosures
2027Lab opens at OSU
Aerial view · autonomous robot navigating commercial apple orchard rows

What we work on

Agriculture first. Field robotics broader. Energy-aware everywhere.

Sustainable autonomy

Designing robots that run on renewable power and use energy efficiently, so they can work in the field for an entire growing season without interruption.

Perception and intelligence

Teaching machines to see and reason in outdoor environments where lighting, occlusion, and weather change constantly, using camera, depth, and LiDAR data alongside modern vision-language models.

Precision field manipulation

Building end effectors and motion plans that can handle fruit, flowers, shoots, and branches without bruising or breaking them.

Resilient deployment

Testing and hardening systems so they work reliably across orchards, vineyards, greenhouses, forests, and other demanding outdoor conditions.

Field operations

Our primary focus is specialty crops, including fruit, vegetables, and nursery stock, where much of the work is still done by hand. We develop robotic systems for picking and harvesting, weeding, pruning, green-fruit thinning, pollination, and crop monitoring and yield estimation.

Where we deploy

Our systems are tested and deployed in orchards, vineyards, greenhouses, forests, wildfire-affected terrain, and row-crop fields. These environments demand robust, field-ready engineering that can handle uneven ground, variable lighting, and unpredictable weather.

Method stack

Machine vision
Detection and segmentation of fruit, canopy, and obstacles using RGB-D cameras, LiDAR, and 3D shape fitting
Vision-language models
Open-vocabulary recognition and vision-language-action policies for tasks that resist fixed label sets
Agentic AI
Autonomous tool-using agents with built-in safety, trust evaluation, and risk management
Robotic manipulation
Custom and soft end effectors, air-assisted occlusion clearing, and contact-safe motion planning
Synthetic data
Generated training images and annotations that reduce the cost of manual field labeling
Energy-aware design
Solar-aware task scheduling, efficient onboard compute, and season-long deployment planning

Featured projects

All projects →
Robotic king-flower pollination

Active

Robotic king-flower pollination

A vision-guided robot that finds king flowers in apple canopies and delivers pollen precisely, reducing dependence on managed bees during unreliable bloom weather.

Partners
Washington State University, Cornell AgRobotics Lab
Funding
WSU Commercialization Gap Fund ($50,000, 2025)
Field site
Commercial apple orchards, Washington State

pollinationmanipulationmachine vision

Robotic green fruit thinning

Active

Robotic green fruit thinning

Detecting and sizing immature fruitlets, then removing excess fruit with a soft robotic manipulator so trees carry an optimal crop load.

Partners
Cornell University
Funding
Cornell CALS Moonshot Seed Grant, Proposal Development Award ($50,000, 2025-2026)
Field site
Commercial apple orchards

thinningsoft roboticsYOLO

Latest news

All news →

Selected publications

All →
  1. Robotics for crop pollination: Recent advances and future direction

    Sapkota, R., Whiting, M. D., Ahmed, D., & Karkee, M.

    Journal of Field Robotics, 2026

  2. AI agents vs. agentic AI: A conceptual taxonomy, applications and challenges

    Sapkota, R., Roumeliotis, K. I., & Karkee, M.

    Information Fusion, 126, 103599, 2026

  3. Multi-modal LLMs in agriculture: A comprehensive review

    Sapkota, R., Qureshi, R., Hadi, M. U., Hassan, S. Z., Sadak, F., Shoman, M., et al., & Karkee, M.

    IEEE Transactions on Automation Science and Engineering, 2025

  4. Comparing YOLOv8 and Mask R-CNN for instance segmentation in complex orchard environments

    Sapkota, R., Ahmed, D., & Karkee, M.

    Artificial Intelligence in Agriculture, 13, 84-99, 2024