A nice example of cross-sectoral innovation. ACTTiVAte looks for similar projects.
A technology capable of predicting production without human intervention would be a dream come true for winegrowers across Europe. If everything goes according to expectations, this technology will be available in less than two years.
In a market where quality is becoming more and more important, the VINBOT project could boost the competitiveness of European wine producers. More accurate forecasts are more accurate decisions, and the autonomous robot created and its network of sensors can calculate the grape production and relevant characteristics of the vegetation cover, generate maps and disseminate the information derived through the cloud computing.
André Barriguinha, Agri-Ciência director and VINBOT post-project dissemination and marketing leader, talks about the results of the initiative and plans for the future.
1.- What specific challenges did VINBOT propose to address?
VINBOT can operate in the vineyard autonomously, without human intervention. It has a sensor system that provides you with the ability to move, locate and acquire data. It is equipped with a camera that captures images of the vegetation cover. Then, it applies algorithms that recognize the grapes and the clusters and calculate the production.
There is no such thing in the market. There is no device that calculates the productivity, and therefore the winegrowers have to resort to manual processes that are very slow and not precise. Thanks to VINBOT, you will have a new tool with which to make these calculations as soon as possible.
2.- How come there was no such technique before VINBOT?
Due to the difficulty of having a completely autonomous robot capable of moving around the vineyard and calculate the production. But that is just what VINBOT brings: an autonomous displacement thanks to the built-in GPS receiver and a 2D rangefinder; An HMI that defines a series of reference points and the characteristics of the data acquisition task; Components for measuring grapes; A cloud-based software that processes data from the robot’s sensors to extract relevant information and generate production maps; And a web application for the end user to consult the maps.
One of the biggest obstacles has been the recognition of the clusters in the vineyard, especially those that are not in sight of the camera because they are hidden behind leaves or other clusters.
In fact we continue in it; We use models based on a three-dimensional reconstruction of the vegetal cover facilitated by the range finder. Our results showed that the VINBOT platform is able to calculate foliage characteristics and production with a respectable precision. However, it is necessary to investigate to a greater extent the underestimation of the actual production, caused mainly by the hiding of the clusters, to improve the precision of the algorithm. We are confident of our ability to increase the pressure with new research into both artificial vision algorithms and models that calculate hidden clusters. We are planning a second project that will do that.
3 .- What arguments would you offer to growers to convince them of the advantages of using VINBOT?
The margin of error of the manual procedure is immense, of about 30%. So a technology that reduced that margin to 10% would be a huge advantage. VINBOT is able to calculate its productivity; Generate maps autonomously and almost in real time; Indicate the need to prune clusters to prevent excessive production that depletes the quality of the wine; And improve planning and organizational decisions.
Finally, you can help plan purchases and sales of grapes, decide on prices and management of wine cellars, schedule investments and develop market strategies.
4. Do you already have an approximate idea of what it will cost the growers to acquire VINOBOT technology?
For most producers, it would not make sense to buy a VINBOT; Its purpose is to calculate the production, so if they bought a copy they would have it parked practically all the year. That is why we intend to offer VINBOT through service providers, but also directly to large producers who manage large areas of cultivation.
On the other hand, VINBOT is more than the robot. It requires a server for the post-processing of the images, so it would be easier and cheaper for producers to turn to a company that provides the service.
As for the price, the final version of VINOBOT would cost about 30,000 euros in its full version, although this price could be reduced as we adjust the technology.
5. By the way, VINBOT connects to the cloud. How important is that role?
Given the huge amount of data that needs to be processed, it is easier and cheaper to use the cloud. Therefore, the algorithm with which the images are treated is housed in a processor located in the cloud. Thus, wine growers do not need more than a username and a password to access their results.
6.- They also expect VINBOT to allow producers to sell their wine at a higher price. How do you intend to achieve it?
It is not a direct consequence. But if I use VINBOT, I can make better management decisions and, indirectly, improve wine quality. Optimization of production management and harvesting logistics, fruit quality and homogeneity, foliage management, bunches pruning and differential harvesting make it possible to plan the production, marketing and distribution of the wine with Greater efficiency.
In theory, all this allows the producer to aspire to a higher market price, although this may not always be realistic, given the great competition in the market. On the other hand, VINBOT can contribute to reduce the general costs of production, and this would increase the margin of benefit.
7.- What conclusions have you drawn from the results of the field tests?
Overall, we are satisfied with the overall performance of this robotic platform. We have had several problems with the traction of the wheels in plowed field and with the fact that the system moves practically like a tank, but already we are thinking of installing a set of wheels able to turn independently to avoid this problem.
The next challenge has to do with software and algorithms. We need a more in-depth field validation that is not limited to the acquisition of data from the images, but allows to perfect the algorithms of artificial vision and the modeling in the treatment of the data. In this way we will know perfectly what we have to achieve to achieve a margin of error of less than 10-15% in the calculation of production.
8.- Assuming that they obtain another subsidy, when do you believe that this technology can be commercialized?
Currently, VINBOT has a TRL (technological readiness level) of 7. We will seek funding from Horizon 2020, and if we succeed and the process of improvement and validation fits our plans, we may be able to market VINBOT in between two And four years.
In addition, we hope that this technology will be used in other areas, and not only in vineyards; For example, in raspberry greenhouses in Portugal, where it is interesting to calculate the production from the analysis of images. We also maintain contacts with entities in the United States and we intend to try VINBOT there. Finally, we want to integrate more sensors, some of environmental type.
Do you have an innovative project with a focus on technology transfer between Agri-Food and ICT, Health or AeroSpace sectors?
ACTTiVAte offers direct funding and business support services to innovative SME projects. Open call for proposals. Apply now!