The Sweeper consortium was invited to hold the first live demonstration of its new sweet pepper harvesting robot at the De Tuindershoek greenhouse horticulture firm in IJsselmuiden. The so-called ‘Sweeper robot’ is the world’s first harvesting robot for sweet peppers to be demonstrated in a commercial greenhouse. An audience of over 40 interested parties watched the harvesting robot pick its first commercially-grown sweet peppers.
The Sweeper robot was designed to harvest sweet peppers in a cultivation system based on single plant stalks in a row, a crop without clusters and in little foliage near the fruits.
In earlier test set-ups in a commercial greenhouse with a V-type double-row cultivation system the harvesting robot achieved a harvesting percentage of 62%. Based on these test results, the Sweeper consortium expects to be able to bring the commercial sweet pepper harvesting robot to the market in about four or five years.
Further research required
Until then, further research will be needed to enable the robots to work faster and achieve a higher success percentage. Additionally, commercially viable cultivation systems must be developed that are more suitable to the robotic harvesting of crops. The test and research results are not only suitable for the automatic harvesting of sweet peppers; the data can also be used to robotise the harvesting of other crops.
International research partnership
Sweeper is a partnership between Wageningen University & Research (WUR), sweet pepper farm De Tuindershoek BV, the Umea University in Sweden, the Ben-Gurion University in Israel, the Research Station for Vegetable Cultivation and Bogaerts Greenhouse Logistics in Belgium. The study receives financial support from the EU’s Horizon 2020 programme and is also funded by the Dutch Horticulture and Propagation Materials Top Sector.
Successor of CROPS
The Sweeper robot is the successor of CROPS (Clever Robots for Crops), an EU project launched by WUR, in which WUR and the other participants developed a robot that can make a distinction between a sweet pepper plant’s fruit, leaves, stalks and main stems. As a result, the robot can harvest sweet peppers without damaging the fruit, leaves, stalks or stems.
Source and photo: www.sweeper-robot.eu. Video: Wageningen UR greenhouse horticulture.
Brabant start-up Crux Agribotics has developed the world's first fully automated cucumber harvesting robot. The robot is unique because it also harvests cucumbers deeply hidden between the leaves of cucumber plants. The robot uses a machine learning algorithm, which allows it to independently determine which cucumbers are ripe for cultivation, and which cucumbers are diseased.
The harvesting robot consists of a trolley with rotating cameras and a flexible grab arm. The cameras make 3D images of the crop from a variety of angles, after which the robot uses special software to build up a complete picture of where cucumbers are located. The robot then sends the coordinates of each cucumber to the grab arm, which clamps the cucumber by the stem.
Since the robot visualises the crop from different angles, it also sees cucumbers hidden deep between foliage. Special software converts various images into a whole, and controls the grab arm. According to Richard Vialle, co-founder and director of Crux Agribotics, the robot picks 96 percent of all ripe cucumbers. A limited number of employees follow the robot to pick any cucumbers which are missed.
Vialle says that the data collected by the robot is very important. The robot uses a machine learning algorithm, allowing it to assess more accurately which cucumbers are ripe for picking. The grower is partly responsible for this learning process. According to Vialle, "The robot sends the grower a picture when in doubt, and asks what it should do with the cucumber in question." The robot also learns to recognise diseased cucumbers in this way.
Harvesting, sorting and packing
A major advantage of the machine learning algorithm used by Crux Agribotics is that a robot can also sort and pack cucumbers. However, the advantages of the algorithm do not stop there. Vialle says, “The learning algorithm means that the robot can also make predictions, such as how the cucumber will grow, or how many days before a cucumber is ripe."
On the market
Crux Agribotics has tested the fully automated cucumber robot with a number of growers, and is currently working on developing an industrial product that can be used anywhere in the world. “We need another two years for this, and are we looking for strategic investors," says Vialle. Crux Agribotics is also developing similar robots for pepper and tomato crops.
Text: Leo Hoekstra. Photo: Crux Agribotics.
In the European SWEEPER project, efforts are being made to create a sweet pepper harvesting robot. Part of this project is the development of fruit and obstacle recognition based on colour and depth images.
For this purpose, an artificial 3D model of a pepper crop was recently produced based on plant measurements in the greenhouse, including all the distributions of angles and geometry. This digital model can be used to grow random, unique digital plants in a simulated greenhouse environment. Once the digital crop is created, synthetic colour and depth images can be rendered on a supercomputer and used as training material for artificial intelligent learning systems. This enables us to localise parts of the plants such as the fruits, leaves and stems from the robot’s perspective.
The recognised parts of stems can then be used to map out obstacles for the robot to avoid during harvesting. Previous research has shown that the angle of approach during harvesting is very important in achieving a high harvesting percentage.
Sweeper was launched at the beginning of 2015. The goal of this project is to develop a harvesting robot for sweet peppers. The robot is being tested at De Tuindershoek, a sweet pepper farm in IJsselmuiden run by two brothers, André en Paul Kaashoek.
What made you decide to join in the development of Sweeper?
‘As a member of the Dutch National Sweet Pepper Committee I was involved in Crops, the predecessor of Sweeper. As Sweeper had to be tested in practice, we proposed our farm as a testing ground. Also, if you are among the first to collaborate on projects like these, you will be able to benefit from innovation grants. That doesn’t mean that funding is our only incentive: all six partners in the project are expected to put a lot of effort into the project.’
What is being researched at your farm?
‘The prototype should be ready in February 2018; we are currently still looking into a number of different aspects. One of these was testing several types of cameras. The goal was to choose the best camera, while also exploring the possibilities of what a camera can capture in terms of data. It may even lead to us cultivating another variety to accommodate the research, for example. That would be quite a drastic change, but like I said, we have very high standards in terms of what we want to accomplish with this project.’
‘Labour is currently our biggest cost item. A robot would help us achieve tremendous savings in this.’
What advantages would a robot have for your farm?
‘Our biggest cost item is labour, for harvesting the green peppers. A robot would help us achieve tremendous savings in this. We would also need fewer seasonal workers. On the other hand, we would need more technical staff for maintenance on the robots. That would mean a big change in the composition of our workforce. Of course, we would have to make sure that the venture is economically viable: a robot has to give you something extra.’
Aren’t you apprehensive about making this investment; by being the first to test a harvesting robot?
‘Nobody is forcing us to join in, or adapt our greenhouses. This is our very own decision. And a decision we will continue to support.’
Are you enjoying your contribution to the development of a robot?
‘Absolutely. It is also very exciting and quite challenging at times.’
Text: Tuinbouwteksten.nl/Mario van Vliet. Photo: Mario Bentvelsen.
Robots appear to be the best solution to combat the rising cost of labour. After all, they can work 24 hours a day, without uttering a single complaint. So, what is keeping the entire horticulture industry from embracing the robot? Are they too expensive, has the technology not yet been developed to a high enough level, or are humans simply better workers in the horticulture industry?
The first robots have been spotted in the horticulture industry. Even better: there are plant growers with entire ‘streets’ of thirty robots whose job is to take and root cuttings. Still, most of the work being carried out in this industry involves manual labour. But aren’t robots cheaper as well as better than humans? Or, in other words: what’s preventing the large-scale introduction of the horticulture robot?
Big chances for success
The robots we are most familiar with are those used in the car manufacturing industry. In the horticulture industry robots are also very welcome alternative, with big chances of success. According to a whitepaper on robotisation (Dutch only) by a team of researchers from Wageningen University and Research Centre (WUR), this is mainly because work in the horticulture industry is often monotonous and physically demanding, and the sector subsequently attracts increasingly fewer workers. The nature of the work is, however, not the only factor. Robots perform their work with a more consistent level of reliability than people, say the researchers. ‘Thanks to modern sensors robots are able to pick, grade and inspect food products with increasingly greater precision.’
Why the agricultural sector has had to wait so patiently for the large-scale introduction of robots can be easily explained. ‘Circumstances in this sector are, by the nature, subject to constant fluctuation: a bell pepper will never be suspended at exactly the same spot for two days in a row, and no two udders are identical.’ Developments are, however, coming along rapidly. It is inevitable that the robot will soon be making a breakthrough in the horticulture industry, say the researchers. ‘The sector will be facing a gigantic challenge worldwide in 2050: having to feed nine billion hungry mouths. Without the extensive application of robots in food production we do not expect the industry to be capable of meeting this challenge.’
Tomato packaging line
Nevertheless, there are plenty of developments being launched, albeit it fits and starts. In the kick-off to the 1996 Dutch Mushroom Days the presentation of a grand total of three robots was announced. In the end, only one robot was actually presented at the fair; the other two never left the testing area - and that one robot was demonstrated for only a limited number of hours during the fair in Grubbenvorst due to technical problems. Now, almost a decade later, there are still no commercial harvesting robots for mushrooms.
Researcher Rick van de Zedde of WUR knows more examples of horticulture robots that failed to make it to the market. This can seldom be attributed to technology or the quality of the robot, but is more often than not related to a lack of confidence among suppliers and growers. ‘You may be able to develop the most amazing robot in the world, but you need to find someone willing to dig into his pockets to pay for the first one.’
The European Union found it time for a breakthrough. Politicians and policy-makers felt Europe would lag behind in terms of competitive power if not more robots were to be developed. This is why the EU is putting a lot of money into the development of robots. WUR and several of its partners are currently applying an EU grant to design a new, fully automatic tomato packaging line: the PicknPack.
The bar is quite high in the development of the PicknPack project, says Van de Zedde. ‘We will be testing the line on vine tomatoes. This has made it a complex project: a vine tomato is a composite product, after all. Any other fresh product would have been easier.’ The project is currently at a halfway mark. The line is being constructed in Wageningen and will be demonstrated to the public in 2016. PicknPack will be able to grade vine tomatoes (based on camera images), and pick up move and package bunches of vine tomatoes.
Attempts to promote the development of the robot for the horticulture industry aren’t limited to the European Union; several years ago the Japanese government funded the development of a horticulture robot, too. This strawberry-picking robot was able to harvest 60 per cent of the red strawberries, at a rate of only 9 seconds apiece. The Spanish Agrobot has been equipped with several picking arms, further increasing its productivity.
The EU aims to ensure that more and better robots are developed. However, this does not mean that there are no horticulturists already making use of robots. Rose grower Leo van der Harg purchased the Rombomatic a decade ago to clip cuttings from rose stems and subsequently plant (or root) them. The robot was designed by the firm of Jentjens, which was taken over by Irmato in 2013. This company was also closely involved in the Crops Project , a project that focused on the development of harvesting robots for various types of crops, such as apples and sweet peppers.
The Crops Project was a good starting point for the further development of harvesting robots, according to WUR researcher Jan Bontsema. Crops showed that greenhouses will need to be organised differently to facilitate the successful implementation of robots. Robots, for example, see crops only from one side. They therefore tend to overlook fruits that are hidden or even partly concealed behind leaves. This means that they would have to be able to approach a crop from both sides. Greenhouses will need to be prepared for this.
Such aspects are taken along for consideration in the follow-up to Crops: Sweeper. This research project was launched at the beginning of 2015 and receives European funding. ‘The Sweeper project explicitly takes the crop into consideration,’ continues Bontsema, the project’s leader. ‘This is why it is so important that growers are also involved in the development of Sweeper.’ A study will be conducted at the De Tuindershoek sweet pepper farm in IJsselmuiden during the next few years to discover how a greenhouse should be reorganised to make Sweeper a success.
The harvesting robot itself will also be subject to redevelopment. Bontsema provides an example: ‘In the Crops Project, sweet peppers are harvested using two ‘fingers’ and a pair of scissors. The robot therefore needs to know exactly where the stem is located, a task which it doesn’t always perform successfully. Sweeper is currently testing a ring to capture the sweet pepper. This ring will make it much easier for the robot to find the stem.’ The Sweeper robot should be ready to market in a few years. Bontsema: ‘Our goal is to build a fully operational robot within the next three years. The next step will then be for our partners to bring it to the market.’
The crux of the matter is: will such a robot be a financially attractive alternative for growers? This is quite a difficult question. Van der Harg: ‘Thanks to this robot, I need fewer people to make and transplant the cuttings.’ Saving on labour costs is, of course, the foremost argument in favour of buying a robot. For this purpose Wageningen UR developed MARVIN™, a robot that can inspect and grade 19,000 tomato cuttings per hour, thus replacing a work force of 27. Besides MARVIN™, the Enschede-based firm of Demcon recently announced a new harvesting robot. The newspaper header read ‘Asparagus robot at DEMCON Enschede does the work of ten people’
Van de Zedde says that it is still only in a few cases that the investment costs of a robot can actually be earned back by cutting labour costs. ‘That will only be possible if your product is available for longer than just a season, and if the robot can work for 24 hours a day. An apple-harvesting machine, for example, can only be deployed for a few weeks every year. An investment in a robot like that will only be profitable if it can perform other tasks for the remainder of the year.’ This is precisely what makes the idea behind the Sweeper robot: it can perform its tasks on almost a year-round basis.
What is at least as important is the uniform quality of work performed by a robot. After all, if you tell a robot to cut roses 70 cm in length, it will keep on cutting 70 cm roses until there are none left, night and day. When you tell a person to perform the same task their enthusiasm for the task is bound to wane after some time. However, there are tasks that a human can execute far better than any robot. Van de Zedde: ‘People are very sensitive to deviations. This is why so much of the labour carried out in the horticulture industry is manual: people are very good at that.’ Piet Oomen of the ISO Group: ‘One of customers decided at one point to outsource his production to Africa, where labour is much cheaper. However, he recently returned to the Netherlands and has started using robots.
Additionally, a robot enables you to take an entirely new approach to growing fruit and vegetables. A mobile cultivation system with the deployment of a robot would be a very sensible idea, for instance: after all, a robot that has to continually move about from plant to plant (instead of the other way around) is, of course, much more susceptible to malfunctions than a stationary one.
Also, using robots will allow for a different business model. Van de Zedde refers to the Japanese market, where strawberries are sold as if the were bonbons: completely uniform in size and ripeness, and sold in posh packaging, Japanese Grade A strawberries can easily be harvested by robots. Those robots do not have to harvest 100% of the strawberries that match the Japanese criteria; whatever remains on the plant can be picked by people and will be marketed through other channels.
Opportunities abound. And there will be even more in the future. Van de Zedde has high hopes for plant phenotyping, for example: the assessment of the quality of a crop or product, based on its external traits for selection. Wageningen University and Research Centre is currently conducting a study into its possible applications, e.g. to monitor quality in the chain.
It is self-evident that many of the current developments feature the tomato as their test crop: it is a popular vegetable fruit that is produced by a relatively large number of growers. It also comes as no surprise that a relatively large amount of research is conducted on field produce. Thousands of hectares in Europe are dedicated to the cultivation of broccoli, for instance. A broccoli-harvesting robot would therefore also have reasonable chances of success.
Van de Zedde proposes that the successful implementation of robots in the horticulture industry depends on a handful of entrepeneurs: are they willing to rearrange their greenhouses in such a way to make the deployment of robots a profitable venture? Large-scale growers will be the first to consider a robot. ‘If you are a front-runner in a particular industry, your chances of profitability following the introduction of new technology will be greater. The ones to introduce new technology to the market will therefore always be its pioneers. I expect it to take another five to ten years before robots become the standard in the horticulture industry.’
Text: Tuinbouwteksten.nl/Mario van Vliet | Photos: Wageningen UR/Mario Bentvelsen.