Experimentation Still Required Even with Machine Learning
While the Design Lights Consortium has come out with efficiency standards that help guide the grow light industry (Ref: Coverage), much testing and experimentation is still required.
The DLC’s efficiency standards offer an important step for the grow light industry and can help sort out which products are efficient and which are not. However, users of grow lights still need to do much experimentation to find the best grow recipe for their particular crops and plants.
Vertical farming is a new field. One of the primary unknowns in vertical farming as well as with indoor farming and with grow lights specifically is the distance between the light and the plant. Further complicating this factor, the height of the plant tends to change over time. The faster the plant grows, the faster the height increases and the distance to the grow light decreases.
Another component of this variable is the amount of light exposure over an area. For a flat plane, such a calculation is easy. Yet, for a plant that grows continuously at varying rates, gets more and more leaves that are not the same size, flat, square, or all oriented the same direction, this calculation is incredibly complex. Of course, such calculations also vary by the type of plant and the specific variant. Machine learning would be useful for such calculations.
Completely Controlled Growth Environments
Also, in a completely controlled growth environment, numerous other factors need to be taken into account, such as water, nutrients, temperature, and humidity. Another part of the equation is the stages of plant growth and harvesting for a particular plant or crop. These stages and their timing again, of course, depend upon the plant or crop. In the different stages of growth, a plant can grow more leaves, more flowers or more fruit. Additionally, the size and the number of flowers, or fruit also varies.
So while defining an ideal growth recipe may at first glance seem like a simple calculation, the true optimization is a balancing act of complex calculations that humans would have a very difficult time attempting to solve.
Artificial intelligence and machine learning can help growers find a good result of this mathematical maze. Such a system requires constant monitoring, control, and management.
Therefore, growers that are experimenting can be hugely aided by a control and monitoring system, and the ideal technology to run such a system is a cloud-based IoT application.
Complexity of Profitability
Two other and perhaps the most important factors are evaluating costs, and determining the profitability of the overall system. Costs may not be too hard to calculate, including operating costs, such as electricity, water, supplies, and staffing. In fact, a good, old-fashioned accountant would likely be able to figure out operating costs. Naturally, the right growth recipe must take costs into account when determining profitability.
Profitability may be the most complex calculation of all because in addition to taking into account the number of harvests per year, the yield per harvest, such a determination must also look at the demand for such crops and how much they could sell for.
How much such crops sell for would also be an extremely complex calculation if you were trying to optimize profitability. The sale price for such a commodity depends on its quality. A measure of quality would look at the size, and weight, and the number of fruits, or flowers, or in the case of cannabis, the number of leaves. It would also have to look at the medicinal content of the leaves or the flavorfulness of the crop or fruit. Once again, these traits are also a function of the growth recipe.
What Would a Perfect, Automatic Growth and Testing System Look Like?
Even with the perfect (theoretical) system that could constantly monitor all of the factors and variables mentioned above, the machine learning to optimize the results would still require many, many data points to learn the patterns for each plant and crop before being able to completely optimize a growth recipe.
Not only would such as system have to count leaves, and monitor plant height, it would need to be able to automatically test crops for the medicinal content or even taste fruits for their flavor or even smell flowers to ascertain the quality of their smell. So right now such a system would not be possible now, it might not be technologically out of the question at some future point in time.
On the other hand, such a system would require far fewer data points to make a recommendation that would “likely” improve a growth recipe for a certain crop or plant. Again, this type of gradual improvement is a long-term process of experimentation that can be greatly accelerated with the right technology, and IoT with cloud computing and machine learning fits the bill.