Polybot's Tomato Harvesting Robot: Greenhouse Automation

Harvesting with precision

In a German greenhouse, a robot's arms carefully pluck trusses of tomatoes. Polybot GmbH, a startup from Tübingen, is automating the delicate task of tomato harvesting with a robot that learns by imitation. Although robots are common on farms, most are programmed with specific rules about what to pick and where to cut. Polybot's robot, however, learns by observing human workers. Martin Kiefel, the startup's founder, explains that Polybot shows the robot how to harvest the tomatoes instead of detailing the fruit or its stem. The robot's speed is currently comparable to a human's, but Polybot aims to increase it by 50%. This boost, combined with automation of other greenhouse tasks, will enable farms to rethink crop rotations.

A new model for agricultural automation

While greenhouse automation is not a new concept, Kiefel’s approach is unusual. Humans continue to be needed for picking crops and processing delicate items because these tasks require adaptability and precision. Companies have been developing more sophisticated robots to address labor shortages, and Polybot's focus on learning by demonstration is one promising direction. The learning approach that Polybot has chosen is typically seen in humanoid robots that pick up tasks by observing humans. This growing trend in technology is driven by labor shortages and an increased appetite for fresh, locally grown produce. According to the U.S. Department of Agriculture's National Agricultural Statistics Service, greenhouse and nursery production in America increased dramatically in the early 2020s, from $18.6 billion in 2020 to $23.4 billion in 2021.

How polybot's robot learns by imitation

Polybot's robot learns by human demonstration. Therefore the precision of each human demonstration makes a significant difference. The robot gradually learns to select the correct tomatoes and cut their stems without damaging the plants. Over time, this consistent training helps the robot achieve the same precise movements as the human demonstrator. The method of learning is called imitation learning. It is a form of machine learning where the robot learns by observing demonstrations of a task. It is designed to handle tasks that are too complex for traditional rule-based programming. This approach is particularly useful in situations where tasks vary and cannot be easily described by a set of rules, such as harvesting tomatoes from different plants.

Riding the rails

Polybot's robot moves efficiently along a track within the greenhouse. One critical part of Polybot is a tightly controlled system of trolley rails. The robot uses these rails to move along the rows of tomato plants. Thanks to its precise navigation, it picks an entire truss of tomatoes in one go. Moreover, it moves along the track at a speed comparable to a human worker. This efficient navigation and precision in movement are essential for the robot to keep pace with the output necessary for a commercial greenhouse operation. Speed is essential for efficient operations, especially harvesting tomatoes, which require careful handling. The robot’s arms are designed to perform the complex tasks involved in precision cutting and stem separation while navigating the greenhouse space.

Inching toward broad adoption

Polybot's automated harvesting system will enter full-scale production in the summer of 2025, with the first systems expected in 2027. This could revolutionize farming by cutting down the cost of delicate tasks. The startup will use this pilot project to gather data and refine the technology, addressing issues like speed and efficiency. Polybot sees the future of farming as one where human workers and robots work hand-in-hand. The startup hopes to make delicate tasks cheap to automate. This technology will allow farms to grow mixed crops again, a practice that has declined due to the high labor costs of manual harvesting.

Keep one eye on the market

Farmers considering Polybot’s system should assess their specific needs and the current state of their operations. Is there a labor shortage or high turnover rate? Would automation help with harvesting timing and consistency? Here are a few practical tips for those thinking about automating their tomato harvest:

• Evaluate Existing Infrastructure: Polybot's system uses trolley rails growers already have, so assess your greenhouse layout and ensure it’s compatible.
• Pilot Testing: Polybot is offering a full pilot this summer, which can be a cost-effective way to test the technology before committing to a full-scale rollout.
• Data Gathering: Use this pilot to gather data on the robot's performance and identify areas for improvement, and how much more efficient it is than human labor.
• Cost-Benefit Analysis: Compare the cost of the automation system to the current labor costs and potential savings from increased efficiency and reduced labor needs.