Harvesting asparagus by hand or using a harvesting robot?

Teboza BV in the Dutch village of Helden operates an autonomous asparagus harvesting robot on its farm. Both the qualitative and quantitative harvesting performances of the AVL Compact S9000 asparagus harvesting robot are being compared with manual harvesting as part of the NPPL-R Horticulture programme (the Dutch National Fieldlab for Precision Agriculture). “Clear guidelines on whether investing in a harvesting robot is profitable are still lacking.”

For asparagus growers, it remains difficult to make a well-founded decision on whether it is more profitable to continue harvesting by hand or to invest in a harvesting robot. That is why WUR Open Crops at the Vredepeel location is independently validating the performance of the AVL Compact S9000 asparagus harvesting robot from AVL Motion in comparison to conventional manual harvesting, within the NPPL-R Horticulture programme. The National Fieldlab for Precision Agriculture (NPPL) is evaluating four different robotic systems in this project: three harvesting robots and one packaging robot. These are being tested in practice on Dutch farms.

Robot versus manual harvesting

Project manager and agri-robotics researcher Toon Tielen from Wageningen Plant Research explains how the study is conducted. In a specially designated trial plot, four rows of asparagus are harvested manually using conventional methods. For comparison, the machine also harvests four rows. The task times/harvest rates of both methods are logged and compared. Additionally, quality assessment and comparison are carried out for samples of both manually and robotically harvested asparagus. These validation tests are conducted randomly throughout the asparagus season—over one week in April, two weeks in May, and another week in June—to provide a good impression of the entire season.

Teboza director Rik Kursten: “It is a challenge for robots to perform outdoors in difficult and changing conditions.”

Technical performance

Another element of the NPPL project at Teboza is examining how this new technology integrates with current operational processes and identifying possible areas for improvement. To better understand how the machine is functioning, an additional camera is mounted on the AVL Compact S9000 harvesting robot, providing feedback on the robot’s performance. The camera checks each module to see whether an asparagus spear is in the gripper and can also measure quality characteristics such as width and length. This provides earlier feedback to the grower if a module underperforms due to a mechanical obstruction, for instance. It also gives the manufacturer a clearer insight into potential areas for improvement. This helps determine whether the detection system occasionally makes mistakes or whether mechanical issues are preventing proper harvesting under certain conditions.

WUR: positively surprised

Agri-robotics researcher Toon Tielen from Wageningen Plant Research is the overall project manager for the asparagus harvesting robot project. “I am positively surprised by what the AVL Compact S9000 is already capable of. It is quite advanced.” However, he observes that it is still difficult for asparagus growers to assess whether investing in a harvesting robot is financially viable. There are no clear tools available for this.

Even seemingly minor variables, such as small differences in quality sorting ratios, can significantly impact overall profitability. Independent validation of harvesting robot performance—both technical and in terms of working speed and the quality of harvested asparagus—will help in making purchase decisions. Tielen also hopes that the attention surrounding the NPPL project will contribute to broader awareness among growers about robots.

“There is still a lot of unfamiliarity and uncertainty around new technology. You have to learn to look at it differently than conventional methods. Humans often make certain mistakes, but growers and staff have learned to deal with these over time. A robot makes different mistakes. You should not immediately declare that it is malfunctioning and dismiss it. Many factors can be logically explained. You need to look beyond them. That is how solutions are found.”

Integration into the operational process

Within the project, researchers are also looking at how the technology of the harvesting robot could be improved to better integrate with the operational process and further increase its added value. During robotic harvesting, data is collected and recorded in combination with exact GPS locations. The aim is to make this data accessible in a farm management system, with computer-generated visualisations of observations displayed on a field map (heatmap). Based on this data, growers could make highly targeted decisions about specific cultivation measures at particular locations within a field. Until now, they have often only had a general idea of local needs at the field level.

Less dependent on labour

Teboza in Helden is active throughout the asparagus chain: from supplying asparagus plants to cultivation, processing and sales of fresh asparagus. The company has been involved in robotics initiatives in asparagus cultivation for about 10 years, with extensive experience using the AVL Compact S9000 asparagus harvesting robot.

Director Rik Kursten: “We have seen opportunities and potential for a long time, especially in terms of operational reliability and reducing dependence on seasonal labour. But we also recognise the challenges for robots to perform outdoors, in all weather conditions and with variable field circumstances.” A further complication is the limited development time available for the robot. As asparagus is a highly seasonal crop, there are only a few weeks each year to test it under real conditions. Kursten: “Still, I am positively surprised by how fast development of the AVL Motion harvesting robot is progressing—especially in terms of asparagus detection, how quickly the vision system can learn this information, how accurately the harvest modules are positioned, and how the asparagus is transported from the harvesting bed to the operator.”

Project manager Toon Tielen (WUR): “Positively surprised by what the harvesting robot is already capable of.”

Confidence in deployability

One challenge that remains is ensuring the reliability of the robot’s deployment, so that it is always available during the harvesting season. Any technical failure can be a real problem when harvesting becomes fully dependent on robots. “There is no longer any room to fall back on an alternative solution, such as hiring temporary replacements if a few manual harvesters are unavailable. The robot has to keep working.” Kursten is looking forward to the results of the ongoing NPPL project. “We will then see how the harvesting robot performs, with clear insight into how many kilos it can harvest per hour and what the quality of that robotically harvested asparagus is—compared to manual harvesting.”

Increasingly better

AVL Motion has been developing the autonomous selective asparagus harvesting robot since 2018. Co-founder and CFO Raymond van den Bos says of the current version of the AVL Compact S9000 harvesting robot: “This robot can already harvest just as well—and in our view even better—than a human. Trained harvesting staff do not keep improving. Also, skilled harvesters do not automatically increase the standards and quality of the underperformers. But the robot, with its self-learning algorithm, will keep improving in terms of working quality. Each machine performs reliably at the same highest level. Even the automatic acceleration based on the amount of asparagus on a bed can still be optimised, so that the robot covers more ground in the same time. Although the constant adjustment of speed is already high.” This results in a capacity of about one third of a hectare per hour.

Co-founder and CFO Raymond van den Bos of AVL Motion: “The robot can already harvest just as well—and in our view even better—than a human.”

Cautious approach

Nevertheless, asparagus growers remain cautious about investing in robotics. Last year was particularly challenging, with excessive rainfall and limited sunshine, resulting in lower asparagus yields. This left uncertainty during robot demonstrations about whether such a harvesting robot would be able to deliver sufficient capacity and work quality under typical, larger harvesting volumes.

That is why Van den Bos welcomes the NPPL project, which now compares robot performance with manual harvesting in a validated, independent way—not only in harvesting standards but especially in quality results. “Because very little is actually known about what hand harvesters leave behind. With a harvesting robot, all damage is immediately visible to growers, as the robot also collects the waste. But what and how much manual workers leave behind in terms of waste and damage is not seen.” Furthermore, the end of the season will reveal the impact of robotic versus manual harvesting on the asparagus plants.