[Interview] It Picks Alone Through the Night… 70 Million Won Strawberry Harvesting Robot “Picky” to Be Commercialized at Year-End

A robot that moves around a greenhouse while people are asleep, picking only ripe strawberries, sorting them by size, and placing them on trays will hit the market at the end of 2026. It is “Picky,” developed by strawberry harvesting robot developer Beyond Robotics Co., Ltd. (CEO Byun Sung-ho), which has completed verification and is preparing for its official launch. The price is 70 million won per unit. This is not a figure arrived at by first building the performance and then setting the price, but one calculated by first deciding on a price that farms can afford and then fitting the performance within it.

70 million won calculated backward from labor costs

CEO Byun Sung-ho said, “Many farms say they will not use it if it costs more than 100 million won, and quite a few say they would not buy it even at 50 million to 60 million won without a support program.”

The company found its benchmark in labor costs. Based on field research showing that hiring one foreign worker costs 35 million to 40 million won a year, it reverse-calculated a price at which the investment could be recovered within 1.5 to 2 years. Adding the “Picky Hub,” which automatically replaces trays, for 30 million won brings the total to around 100 million won, enabling an unmanned nighttime circulating work setup. In this case, the payback period increases to around 2.5 years.

To meet this price, sensors were used only minimally, and end effectors, including the harvesting gripper, were designed and manufactured in-house to reduce costs. However, considering the risk of fire, the company did not compromise on battery quality. Vice President Jung Young-hoon said, “The strategy was to solve the limitations of low-spec, low-cost hardware through software technology.”

A greenhouse without even Wi-Fi… large AI models were not the answer

The recent buzzword in the robotics industry is “physical AI,” which combines large language models with vision models, but Beyond Robotics did not simply follow that trend. CEO Byun explained, “We are a mobile robot, and AI has to run on-device in the field, where there is not even a Wi-Fi connection. We concluded that large models could not be operated there in the first place.”
 

Instead, the company sequentially deployed several lightweight small models at the points needed by the robot and combined them with the founding team’s strengths in 3D vision and precision control technologies. As a result, Picky runs on batteries alone without external power and independently determines which strawberries to harvest and which route to take to approach them. Variables such as changes in lighting or fruit hidden by leaves are also handled during this decision-making process.

14 to 16 per minute, 85 to 95% harvest rate… It picks all night and even sorts by size

CEO Byun stated, “In a favorable environment for the robot to operate, when it achieved maximum performance, it harvested 14 to 16 strawberries per minute.” According to the company, under the same conditions, a skilled worker harvests about 20 strawberries per minute. In ordinary environments, it shows a 95% harvest rate, but in environments where many fruits are hidden by leaves or harvesting is otherwise difficult, the harvest rate is around 85%.

Care is needed in interpreting the figures. The company’s “harvest rate” does not mean recognition rate, but how many of 100 ripe strawberries that should be picked are actually harvested.

CEO Byun said, “If it attempts too aggressively, the strawberries can be damaged, so we set it to operate defensively.” Because it is designed to prioritize preventing damage, the more complex the environment, the lower the harvest rate becomes. The battery is designed for 8 hours of continuous work and can operate for up to 10 hours.

The noteworthy point is unmanned nighttime work. Strawberries should be harvested at dawn to have high firmness and sugar content, but work at this time is the greatest burden for farms. CEO Byun said, “When farmers come out in the morning, dozens of trays are stacked up, so instead of everyone having to come out at 5 or 6 a.m., it becomes possible to come to work around 9 a.m., sort them, and ship them in the afternoon.” Because the trays are divided into sections, large, medium, and small fruits are classified from the moment they are placed, while deformed fruit or fruit damaged by disease or pests is placed in a separate disposal bin.

Vice President Jung pointed to the problem of failing to harvest at the right time. He said, “If you need to pick 100 today but can only pick 70 to 80, there are many situations where the remaining 20 to 30 have to be discarded the next day.” The company expects that, based on 8 to 10 hours of nighttime operation, it will be able to replace one worker. However, Vice President Jung prefaced this by saying, “This is an ideal figure and may differ in actual field conditions.”

Technology cooperation agreement with Gyeongbuk Agricultural Research and Extension Services… 1.25 billion won in letters of intent to purchase

The turning point came in Gyeongbuk. The Seongju Korean Melon and Fruit-Vegetable Research Institute under the Gyeongsangbuk-do Agricultural Research and Extension Services signed a rental contract, used Picky for three months, and then committed to purchasing one unit. Including the robot and Picky Hub, the deal is worth 100 million won and is booked as revenue for the first quarter of 2027. On June 17 this year, Beyond Robotics signed a memorandum of understanding (MOU) with the Gyeongbuk Agricultural Research and Extension Services for technology cooperation on strawberry harvesting robots. The company aims to connect this agreement to next year’s distribution program.

About 10 farms are currently discussing contracts. Based on letters of intent (LOIs), the amount exceeds 1.25 billion won, and the company says that if actual contracts and deliveries are made, more than 1 billion won in revenue will be possible. Vice President Jung said, “Because there have been no cases of people actually using harvesting robots, many farms feel burdened by the idea of becoming ‘the first case.’ If the public sector creates references and the market is laid through distribution programs, the pace of adoption will accelerate.” The company is also pursuing registration on the Korea ON-line E-Procurement System’s comprehensive shopping mall and designation as an innovative product by the Public Procurement Service.

After strawberries, tomatoes… A structure that deploys robots and earns from data

In 2024, Beyond Robotics completed a proof of concept (PoC) for recognizing and harvesting cherry tomatoes by partially modifying its strawberry harvesting technology.
CEO Byun explained, “Strawberries bear fruit at waist height, but tomatoes, paprika, and cucumbers are trained upward as they grow, so the hardware has to become larger. Aside from that, almost all of the software can be shared.” The next model is a platform that handles both strawberries and tomatoes. However, Vice President Jung drew a line at an all-purpose robot, saying, “There is no reason for strawberry farms to buy a more expensive platform just because it can also handle cucumbers and tomatoes.”

The long-term revenue model is data. Once video captured by robots during harvesting, along with information on yield, time, and fruit condition, accumulates, farms can choose to subscribe to control or disease and pest detection functions based on that data. Vice President Jung said, “The ultimate vision is a ‘dark farm’ where people do not have to enter.”

Although collaboration and investment proposals have continued to come in from Europe and Japan, the company is targeting late 2027 to 2028 for overseas expansion. Vice President Jung said, “We want to first establish the robot we made for domestic farmers, and then go abroad with the references created here.”

The long-term goal presented by CEO Byun is to replace 60 to 70% of farm work with robots within five years. He said, “In a greenhouse, legs are not necessary. Our direction is to run on wheels, while mounting on top of them precise arms, sensors, and AI models that can replace human hands.”