Tackling the challenges of greenhouse robotics

As with many things, CEA and greenhouses find their roots in ancient Rome. First century Roman Emperor Tiberius was documented to have demanded a fresh cucumber to eat every day, as prescribed by his royal physicians. To provide this all year round, the gardeners grew the crops in wheeled carts that could be left in the sun during the day and brought indoors at night, with translucent sheet-covered frames to permit light without exposing the fruit to the elements.

Today’s greenhouses are decidedly more advanced, with complex heating, ventilation and cooling (HVAC) systems in place to ensure that temperatures and humidity are regulated and kept within controlled parameters. Embedded lighting controllers are often used to control LED lighting throughout evenings and nights to increase yield of certain crops. These systems, alongside tightly controlled irrigation, fertiliser application and carbon dioxide enrichment, work together to maintain the optimum growth environment for plant and vegetation.

It's therefore no surprise that greenhouse robots are growing in popularity. In such a carefully controlled environment, especially with carbon dioxide enriched airflow, the presence of human workers can disturb the balance. Greenhouse robots can manage the crops with minimal environmental change.

Similarly, plants need various amounts of space to grow properly, which requires careful spacing of seeds. The space requirements change as the plant grows, which leads to either frequent manual adjustment or wasted space from the beginning. Greenhouse robots take this repetitive task and allow for much more adaptive spacing, which leads to greater yields and reduced loss.

However, greenhouse robots have been slow in development due to several operational challenges. Whereas an industrial robot in a manufacturing plant can handle items with more force, fruit and vegetables must be handled delicately to avoid damage. This requires sophisticated grippers equipped with sensors to ensure precise yet gentle handling of produce.

An equally great challenge for greenhouse robots comes from the computing technology needed. Not only do the robots require an embedded processor that can run increasingly complex AI algorithms, but they must also be compatible with and able to quickly process video input from the machine vision system. Most importantly of all, these embedded computing systems must be able to withstand the high temperatures and humidity.

In these regards, the embedded computing requirements of greenhouse robots are not too dissimilar from those of surveillance and reconnaissance units in security and defence, which Recab UK routinely works with. Therefore, the same latest rugged embedded technologies can offer significant benefits to greenhouse robot OEMs.

For example, one option is to base a system on a Computer-on-Module (COM) that features an 11th generation Intel Core processor, such as congatec’s conga-TC570 COM Express Type 6 module. In addition to the low-power high-performance CPU, the module also features a high-performance Intel Iris Xe GPU and extensive I/O interfaces, including support for PCIe Gen 4 and USB4. These features make it ideally suited to graphics-intensive processing.

Computing tasks can run smoothly across all four CPU cores and up to 96 graphics execution units on the GPU, ensuring reliably high performance. The module also features Intel Time Coordinated Computing technologies, in-band error correction and virtualisation to effectively support real-time computing. Crucially, the module can operate in extended temperatures up to +85 degrees Celsius and in environments with up to 90% relative humidity.

Another option that ensures high video processing capabilities is to use a general-purpose GPU (GPGPU). The challenge is in developing a GPGPU solution that meets the temperature and humidity requirements of a greenhouse, but this is a challenge that Recab UK’s specialist team is well-positioned to help overcome. Drawing on our experience of developing GPGPU solutions for harsh environments, we can develop a rugged system based on Diamond Systems’ Elton carrier board.

The Elton board is an ideal platform for the NVIDIA Jetson AGX Xavier module to provide high compute capabilities for high-performance applications such as AI, machine learning and image processing. The module can provide up to 30 tera operations per second (TOPS) in performance at 30W, or 32 TOPS on a 512-core Volta GPU.

The computing power is coupled with ruggedisation features that make it a prime choice for demanding environments, with a thicker PCB, conduction cooling, latching connectors and a rugged PCIe/104 expansion connector. The board’s AGX Xavier compatibility means it can also support the newer industrial version of the module, for an even more reliable solution in harsh environments.

To make the most of the capabilities of the Xavier module, the Elton board features several sockets to expand I/O. The camera module socket can support up to eight CSI cameras, or four cameras with 4K resolution. There is also an M.2 socket for PCIe x4 NVMe flash storage and an LTE socket that supports cellular modem connectivity. Each of these sockets, along with the core computing performance afforded by the Xavier module, make this board well-suited as a platform for greenhouse robots.

In addition to the board’s ruggedised design making it an ideal base for a rugged GPGPU, Recab UK’s expertise can ensure a bespoke system and ingress protected enclosure that can withstand even the harshest environmental conditions.

Both the congatec conga-TC570 COM and Diamond Systems’ Elton GPGPU carrier board are new technologies, building on decades of computing experience. It is this that allows them to combine ruggedisation with high performance to offer an ideal platform for greenhouse robots.

It may have taken some time for embedded technologies to arrive at this point, but they allow greenhouse robots to at last overcome the environmental conditions and ensure that healthy produce grows all year round.