How a greenhouse robot picks tomatoes

In a high wire tomato greenhouse the heating pipes between rows double as rails, so a harvesting trolley can roll from plant to plant. At each stop a camera finds ripe fruit, the arm moves a soft gripper around one fruit at a time and detaches it, and the fruit goes into a crate on the trolley. Output per hour comes from how long each pick takes, and the share of fruit left behind comes from how often the camera and gripper succeed.

Controls

Labels

Animation speed

Pick cycle time

Detection success

Plants per row

Step through it

Pipes that are also railsHot water pipes run along the floor between rows to heat the greenhouse. Trolleys with grooved wheels roll on them, so the robot needs no steering to follow a row.
Find the ripe fruitA camera on the mast scans the plant at each stop and a vision model picks out fruit that is ripe and reachable. Leaves and stems often hide fruit, which is why detection is never perfect.
Reach and gripThe arm plans a path around stems and moves a soft gripper over one fruit. Padded fingers hold it without bruising while the wrist twists or a blade cuts the stalk.
Into the crateEach fruit is set down in a crate on the trolley, then the trolley rolls to the next plant. Fruit the camera missed or the gripper dropped stays on the vine.
Fast but blindSet a 6 second cycle but only 60 percent detection. The robot finishes rows quickly, yet nearly half the ripe fruit is left for a human crew.

The harvest arithmetic

Time at each plant is a fixed 4 seconds to roll and scan, plus one pick cycle for every fruit the camera finds. This model assumes 3 ripe fruit per plant per pass and that 90 percent of detected fruit are picked cleanly. Fruit per hour is the fruit picked at one plant divided by the time spent there, times 3,600 seconds; time per row is plants times seconds per plant.

Fruit left behind is 1 minus detection success times grip success. These values are simplified for teaching. Published greenhouse trials have reported cycle times of roughly 20 to 50 seconds per fruit and harvest success from below 20 percent in a commercial crop to above 80 percent in simplified test setups.

Try it: start at a 12 second cycle and 85 percent detection, then cut the cycle to 6 seconds. Fruit per hour nearly doubles but the share left behind does not move at all; only better detection lowers it.

Why greenhouses go first

Rails, rows of identical plants, controlled light and a year round harvest make greenhouses an easier place for picking robots than open fields. Growers also train plants so fruit trusses hang clear of leaves, which raises detection and grip success far more than a faster arm would.

Where Avid uses this

Avid works on agricultural robotics, where pick rate and detection rate are the numbers that decide whether a harvesting robot earns its place.

Sources

Selective Harvesting Robotics: Current Research, Trends, and Future DirectionsWageningen University and Research (eDepot)
Edge-deployable greenhouse tomato cluster harvesting robot integrating YOLOv8n-BiFPN-WIoU and ROS-based autonomous controlFrontiers in Plant Science, via PubMed Central (NIH National Library of Medicine)

How to cite this model

Avid Solutions International. (2026, November 11). How a greenhouse robot picks tomatoes [Interactive 3D model]. Avid Field Lab. https://avidsolutionsintl.com/explore/robotic-fruit-picker

Last updated November 11, 2026

Keep exploring

Controlled Environment AgHow aeroponics feeds roots with mist Controlled Environment AgHow aquaponics turns fish food into lettuce Controlled Environment AgWhy greenhouses add carbon dioxide Controlled Environment AgWhy deep water culture needs bubbles