April 20, 2024

The promise of cloud, fog, edge and 5G for developing hands-free greenhouse farms

Author profile photo

Advances in robotics and artificial intelligence in greenhouse agriculture are advancing faster than ever. The entrepreneur has more and more options to help him in repetitive jobs such as harvesting, support in cultivation in the form of models to configure his climate computer or take measurements of his crop “digitally”. With all of these new applications, there is more and more data available on a farm. The first challenge here is to intelligently link all the data from the numerous digital services and products. More on this interoperability challenge in a future column. The question for now is: “How do we get all that data, mainly from the greenhouse, to the place where it is stored and processed?”

Cloud Computing

In 2006, Amazon launched its Amazon Web Services (AWS). The service allowed users to rent virtual machines as infrastructure for their data and applications. Salesforce began offering software as a service (SaaS) over the Internet around 1999, enabling cloud computing.

There are several definitions of Cloud Computing, according to Gartner:
Cloud computing is a style of computing in which scalable and elastic capabilities enabled by IT are provided as a service using Internet technologies..” In addition to SaaS, it can be Paas (platform as a service) or IaaS (infrastructure as a service). Currently, it is estimated that 90% of data worldwide is stored in the cloud.

Storing data in the Cloud has advantages. No major hardware investments are required, and scaling up and down computing and storage capacity is relatively easy. However, there are also disadvantages. Just think about the regulations in many countries regarding data ownership, storage and processing.

For real-time applications, there is another major disadvantage: latency or delay in data transfer over the network. For some applications, that delay is not a problem; Think about sending the total liters of water given in the greenhouse for those 24 hours once a day. But what if a robot needs to use vision to determine whether a rose needs to be harvested and then also needs to immediately stop and point at the rose in question with a laser light?

Or, more importantly, if a logistics system in the greenhouse needs to determine in real time if it is in danger of running over a person or if it has reached the end of its road. Failure to stop in time can have serious consequences! At these times, data will typically not be processed in the cloud but rather “nearby” or even in the system.

Edge and fog computing

Edge Computing is about bringing computing closer to the data source. It is based on processing data at the “edge” of the network. This reduces the amount of data that must be sent to the cloud for processing, reducing network latency and improving overall system performance.

Fog computing is a distributed computing model designed to complement edge computing. It extends the capabilities of edge computing by providing a layer of computing infrastructure between edge devices and the cloud. This infrastructure, called the fog layer, provides additional data and application services to edge devices.

However, there are quite a few challenges with fog computing. Edge applications in a greenhouse, for example, involve many different applications/devices, all with their own protocol, authentication, and security issues. Therefore, new standards are essential!

source: https://www.enisa.europa.eu/publications/fog-and-edge-computing-in-5g


Without edge computing, 5G is a fast network technology developed to transport large amounts of data with low latency and high speed. When many sensors with different types of data use the network, even a 5G network can quickly become overloaded. Therefore, it is important that, where possible, (part of) the data is processed “at the edge”. Furthermore, using the ‘slicing’ technique, it is possible to divide the network into ‘slices’, each of which processes a different part of the data stream. In this way, a difference can be made in the speed, latency and priority of the data in the different “slices”.

In greenhouses, 5G offers a solution to transmit the increasing data collected by sensors and robots. Wi-Fi is often absent throughout the greenhouse or does not work well due to all the steel, glass, water, and the dense crop of a crop like a tomato. That’s why technology developers now often develop their solutions with beacons or radio links. The disadvantage is that a horticultural entrepreneur often uses systems from different suppliers. Therefore, each supplier may have to install a different system, resulting in high costs and maintenance.

Meanwhile, since the end of 2023, a semi-practical-scale practical 5G test site has been installed in the field laboratory’s tomato greenhouse, offering data-driven cultivation at Tomatoworld in Honselersdijk. Together with researchers from TU Delft, TNO and the company MCS, a private 5G network has been created that companies can use to prepare their sensors, robots and other systems in the greenhouse for the commercial introduction of 5G.

Towards a hands-free greenhouse

With the technology available, developers can increasingly make trade-offs about what data to process and where to store. Where that will be depends on practicality, risk and cost.

For a robot where real-time actions are important, processing in the robot itself is now often chosen. High computing and storage capacity combined with low latency requires, for example, an industrial PC integrated into the system. A PC of this type increases the cost and weight of the system. Integrated into an autonomous robot platform, larger batteries are needed if the system can still operate for a good number of hours without recharging. At the moment when a lot of data can be sent very quickly with a 5G connection in the greenhouse to a server located, for example, in the company’s barn, the calculation is carried out there and the result returns to the robot almost without delay, the system can become much lighter and cheaper!

Furthermore, sensors or actuators, which are always essential to remain functional, can easily become wireless with a stable and fast connection. With a wireless temperature sensor, it is less problematic if you ever miss a measurement due to a poor data connection. For this reason, wireless temperature sensors for greenhouses have been available for years. Wireless control of faucets in a tomato greenhouse is a different story. It’s often technically possible, but with greenhouses constantly growing, you don’t want to run the risk of thinking that a wireless water poisoning tap has been turned off but hasn’t occurred due to a poor network connection.

With all the new developments, there are more and more opportunities to get a hands-free greenhouse! How long do you think it will be until we pick a tomato with a robot?

Leave a Reply

Your email address will not be published. Required fields are marked *