Why strawberries are the hard case
Strawberries are one of the most demanding crops to automate. They ripen unevenly, hide under leaves, bruise on contact, and have to be picked selectively — only the ripe fruit, and only at the right moment. They are also among the most labour-intensive crops to harvest by hand, which is exactly why a shrinking, more expensive seasonal workforce pushes growers toward robotics.
That difficulty is also why strawberries became a benchmark for the field. Xiong et al. (2020), in the Journal of Field Robotics, designed, built, and field-evaluated an autonomous strawberry-harvesting robot in real growing conditions — establishing that robotic picking of soft fruit is viable outside the lab. HarvBot, CoFarmer's proprietary harvesting robot, applies the same approach in production.
How the robot sees: detection plus 3D location
A harvesting robot needs two perception skills. First, it has to detect ripe fruit in a cluttered, unstructured canopy. Yu et al. (2019), in Computers and Electronics in Agriculture, showed that AI computer-vision models (Mask R-CNN) can reliably detect strawberries even when leaves and other berries get in the way.
Second, it has to know exactly where each berry is in space so the arm can reach it without crushing it. Ge et al. (2022), in Precision Agriculture, compared 3D-location methods that pinpoint fruit position and size. HarvBot uses 3D vision to do both jobs at once — locating each berry and grading it by size, colour, and ripeness as it picks, around the clock.
From a single robot to a coordinated fleet
A single robot is a proof of concept; a farm needs throughput. The next step is coordination — several machines working the same field without colliding, each scheduled against crop readiness, weather, and labour availability.
CoFarmer Fleet manages multiple robots from one dashboard with real-time telemetry, route planning, and AI-driven task allocation. Because the robots run on the same platform as the farm's monitoring and operations, harvesting decisions use the same live crop data as the rest of the farm — not a separate, disconnected system.
The takeaway
Autonomous strawberry harvesting is no longer theoretical: the perception stack — fruit detection plus 3D location — is validated in peer-reviewed research, and robots like HarvBot bring it to the field 24/7, grading by ripeness as they pick.