"In a vertical farm, the major operating costs are attributed to energy use, of which artificial lighting is one of the main components. Reducing this cost is critical if controlled environment agriculture is to become a more profitable production method, and this is the main problem we are trying to solve with our technology," says Dr. Rakibul Islam, Researcher in Biomedical Sciences at Photosynthetic, a subsidiary of Rift Labs.
He explains that the Photosynthetic R&D lab provides a platform to perform scientific tests by comparing environmental variabilities side by side while monitoring plant growth. It provides valuable insight to reduce the waste of resources through adjusting, for example, the photoperiod, wavelength, temperature, etc., for optimum plant growth. It provides vertical farms with a recipe for energy-efficient plant growth, which they can implement in the production environment.
As a researcher, Rakibul has always had a huge interest in AI. He has been working in a leading AI research group, where they developed and evaluated the application of Artificial Intelligence in cancer diagnosis and prognosis of the disease. Though, at Photosynthetic, Rakibul's role is to innovate AI solutions for growers.
"It was so rewarding to see the great potential of using my AI knowledge to help CEA become more efficient by creating intelligent processes and workflows. Working on this project satisfies both my hobby of growing plants and my intellectual curiosity."
Growing along with the grower
Photosynthetic's expertise is within data generation and precise control of plant growth inputs using its proprietary software and patented technology for mixing light of different wavelengths. For example, through the self-contained R&D lab, growers can better understand the environmental needs of a candidate crop, allowing for experimenting to achieve efficiency without compromising plant physiology.
By controlling environmental conditions in this R&D lab, one can recreate the 'problem situation' for the growers to generate highly relevant data necessary for training machine learning models. The data stored for optimal growth can be used to simulate growth and predict harvesting time for the crop computationally.
"We do not just make a product for the growers—we do it with them. Through Photosynthetic, our aim is to establish ourselves as a research partner and technology supplier for growers around the world. We have a data-centric approach where we want to make use of our expertise within the science of light and offer smart solutions that help our customers to continuously improve their production process," Rakibul concludes.
Since AI is a big field; usually, it is interchangeably used for Machine Learning (ML), Deep Learning (DL), etc. It consists of a set of tools that allow computers to learn without being explicitly programmed, explains Rakibul.
"Traditional software is rule-based, where exact instructions are provided to a computer to solve a problem, which means we need input and rules to provide an output. In machine learning, we need Input and Previous Output to train a model so that it can predict the output when it gets a new input. Machine learning applications are all around us today; most importantly, they are useful," says Rakibul.
For example, the machine learning model works behind the screen when you unlock your phone using a face recognition system. Spam filters in your e-mail, search engine results, and the movie recommendation system on Netflix; are also machine learning models. There are applications of AI in flying drones and talking robots, so it is not entirely wrong to imagine them when you think about AI.
"We must be mindful when designing technology to help productivity so that it does not just sound good but rather adds value to the users by reducing unnecessary work and complexity. The best AI products and services are often invisible and seamlessly integrated into workflows."
AI's actual role in CEA
According to Rakibul, CEA is one of the ideal candidates for harnessing the power of AI. This is because CEAs are designed to control and monitor growth conditions to improve the sustainability of agricultural production. The growers make their decisions according to the monitored variabilities for improved resource use and increased productivity, quality, profitability, and sustainability.
AI may play a role in these areas; for example, it can potentially optimize electricity consumption, which is one of the biggest cost drivers in CEA; automate processes to increase productivity; optimize plant growth and predict yield to reduce uncertainty; detect inferior products to ensure quality.
"The technological capacity is there. However, the thing is that there are no off-the-shelf solutions. These solutions are data-dependent, and therefore, need to be built in collaboration with growers, based on what kind of optimization they need," he notes.
Improving production with AI
In fact, AI solutions in indoor plant production require a remarkably diverse set of expertise. Coding a machine learning model is, of course, a crucial part but not the only part of it and not even the biggest challenge. Developing AI implies many different aspects of a complex infrastructure surrounding the model development, such as data collection, feature extraction, serving infrastructure, monitoring, etc.
"Yet, I consider AI to be one of several interesting tools we can use to increase yield and reduce production costs in indoor plant production. In my opinion, it has tremendous potential, which has yet to be fully explored. We work with AI through our multi-disciplinary team, consisting of plant scientists, AI product managers, and mechatronics, electronics, and software engineers, to support our customers by helping scope and frame their challenges as machine learning problems to be solved. AI can help reduce routine manual tasks, support decision-making, and ultimately optimize the production process."