UW-Madison Tests a Unitree Quadruped Named Spuddy to Spot Potato and Onion Diseases Early

A Unitree B2 robot walking between crop rows is letting UW-Madison researchers catch vine disease and pest damage weeks before drones can see it from above.

Key Takeaways

  • Spuddy is a Unitree B2 quadruped carrying RGB, stereo-depth, infrared and lidar sensors through potato and onion fields.
  • Ground-level imaging catches vine disease and Colorado potato beetle damage before it's visible in overhead drone footage.
  • The robot carries up to 120 pounds of payload and streams real-time 3D data over a wireless connection.
  • In onion fields, the same imaging feeds machine learning models that count bulbs and estimate yield before harvest.

Researchers at the University of Wisconsin-Madison are testing a quadruped robot named Spuddy, built on the Unitree B2 platform, in potato and onion fields at the Hancock Agricultural Research Station. Associate professor Yi Wang's team reported on November 17, 2026, that walking the robot close to the ground between crop rows reveals vine disease and pest activity, including Colorado potato beetle damage, well before it becomes visible to aerial drones scanning the canopy from above.

Spuddy weighs roughly 60 to 70 pounds, can carry up to 120 pounds of additional equipment, and ships with front and rear RGB cameras alongside stereo depth sensors, infrared imaging and lidar. The team, which includes PhD researcher Taqdeer Gill and engineering doctoral students Liqun Xu and Zekai Xu, mounts extra cameras on the robot's top, sides and belly to capture a 360-degree view as it moves through rows.

Key takeaways

  • Spuddy is a Unitree B2 quadruped carrying RGB, stereo-depth, infrared and lidar sensors through potato and onion fields.
  • Ground-level imaging catches vine disease and Colorado potato beetle damage before it's visible in overhead drone footage.
  • The robot carries up to 120 pounds of payload and streams real-time 3D data over a wireless connection.
  • In onion fields, the same imaging feeds machine learning models that count bulbs and estimate yield before harvest.

Wang's team frames the robot as a scouting tool, not a replacement for grower judgment. The goal is to offload the most physically demanding and repetitive fieldwork — walking rows for hours to check plant health — so farmers get denser, earlier data without the labor cost.

Why ground-level robotics complements drones

Agricultural drones have become standard for scouting large acreages quickly, but their overhead vantage point has a structural blind spot: canopy cover hides early-stage disease and pest activity on lower leaves and stems until damage has already spread. A quadruped moving at crop height closes that gap, trading coverage speed for resolution and timing. That trade-off is why ground robots are emerging as a complement to, rather than a replacement for, aerial systems in precision agriculture.

The approach also sidesteps a cost problem that has slowed robotic adoption in row-crop agriculture: rather than building bespoke hardware, Wang's team adapted an off-the-shelf quadruped platform already proven in industrial inspection, which lowers the barrier for other university programs or extension services to replicate the setup.

What comes next

The research is still at the field-trial stage, run through a university extension program rather than a commercial product. Wang's team plans to extend the approach to corn, and the yield-estimation work in onions — counting and sizing bulbs from thousands of pre-harvest images — suggests the next step is pairing ground-level scouting with predictive yield modeling that growers could use for marketing and logistics decisions ahead of harvest.

Category: Robotics

Tags: Agricultural Robotics quadruped robot precision agriculture Unitree B2 UW-Madison crop disease detection

Related Articles

Frequently Asked Questions

What robot platform is Spuddy built on?

Spuddy is built on the Unitree B2 quadruped platform, equipped with RGB cameras, stereo depth sensors, infrared imaging and lidar.

Why use a ground robot instead of just drones for crop scouting?

Drones image the canopy from above and miss disease or pest damage happening beneath the leaves; a ground-level robot can detect these problems weeks earlier.

What crops has the robot been tested on?

The UW-Madison team has tested Spuddy primarily in potato and onion fields, with plans to extend the approach to corn.

Who is running this research?

Associate professor Yi Wang's team at the University of Wisconsin-Madison's Department of Plant and Agroecosystem Sciences, working at the Hancock Agricultural Research Station.