There is an overwhelming problem that many companies are currently laser focused on: how will we feed the world by 2050? With the world’s population expected to increase to nearly 2 billion more people by then, it only makes sense that we’ll need to increase agriculture production by at least 60%. But to do this within the next 27 years is a daunting task and will require multiple solutions to get us there. Startups and 150+ year-old agriculture equipment companies like John Deere are looking to full and semi-automated processes to reduce labor and other input costs with the goal of making our food supply more resilient to shocks.
We see many companies laying the groundwork to reach fully automated farming operations, which will be a key part in achieving global food security. The building blocks of automating farm operations lie in the ability to use machines to supplement, and in some cases, replace humans in traditional farming jobs, allowing humans to do more value-add work. Before we explore the possibilities into what the future could look like to prevent a global food crisis, let’s look at how robotics, machine vision, and machine learning are being used today.
Autonomous farming
From Bear Flag Robotics’ retrofit autonomous driving technology to CNH Industrial’s Auto Guidance capability, tractors are driving themselves all over the world. A farmer can have 10 tractors working in the fields all at once using GPS, machine vision, sensor networks, and connectivity without any drivers. All this tech is put in place to use less manpower directly in the fields.
Mineral, an X company, is using autonomous robotics to collect terabytes of data in fields, from which they are performing a range of experiments. For example, the data being collected include images of strawberry plants in various stages of growth and will be used to develop powerful algorithms to help predict yields, identify diseases, and determine ripeness. Field data, combined with satellite imagery, drone-collected measurements, and weather information will help inform growth models and provide the key to scaling agricultural production.
Aigen’s Element is a robotic solar powered vehicle that travels through row crops, weeding, and observing plants as it moves along. This data is passed to the grower who uses it to manage water rates and other aspects of growing throughout the season. Ultimately, allows growers to eliminate chemical use during the growing season.
Autonomous driving systems unlock the ability to precisely plant, grow and harvest crops. A person gets tired over time and can make mistakes — planting too much seed in one place, not enough in another. Precision planting ensures the right seeds at the right depth, every time, day or night.
AI-powered weed killers
In 2021 Blue River Technologies and its parent company John Deere launched a new technology that applies herbicides equivalent to the work of 100 people each armed with a spray bottle. Powered by machine vision and artificial intelligence, hundreds of valves enable self-propelled sprayers to only focus on the weeds – just like how we’d do in our backyard attacking dandelions. Is this new sprayer a robot? You bet.
Carbon Robotics, an AI-powered robotics company, has developed a unique way to get rid of weeds in farms. The LaserWeeder focuses on chemical reduction using machine vision and artificial intelligence to identify weeds. Once located, weeds are then eliminated with lasers. Yes, lasers.
These technologies help significantly reduce the amount of chemicals used in a field compared with traditional spraying. The growers save money by using fewer resources and the reduction of runoff protects groundwater. It makes crops healthier and minimizes impact by herbicides.
Seeing fields covered in robots in the form of self-driving tractors, automated weed fighting machines, self- propelled sprayers, and lasers show that we can make massive improvements to our farming capabilities. However, we still have more to do to get to where we need to go.
Using technology to harvest specialty produce
Picking fruit seems fairly straightforward to most of us – we can do it pretty easily. If we break down the process, though, it requires identifying the individual piece of fruit, making sure the fruit is ripe, and picking the fruit off the plant without damaging it. A machine that can accomplish these tasks will have a smart vision system to “see” and a picking mechanism to “pick.”
Teaching a vision system the characteristics of ripe fruit requires images of fruit when it should be picked. The variability of each individual berry drives the number of images needed to train the algorithms to identify what ripe fruit looks like – something a human can do through sight and experience.
Once the system can “see” the fruit and know where it is, the picking mechanism can move to the berry. It needs to know that it is close to the berry, and how hard it needs to squeeze the fruit – hard enough to grasp the berry without dropping it, but not too hard to damage the fruit. And the mechanism has to adjust for the size of the berry, requiring the “size” of the grasp be separated from the “force” applied to pick but not drop. Separately these actions are robotically straightforward, taken together they are much more complicated.
All this to say robots are fantastic machines when they are purpose built for doing one or two tasks really well. As they become more complicated, the development costs increase significantly.
Robotic development and automation affordability
The cost to design, engineer, train the algorithms, and commercialize robotic farm equipment is much higher than most expect. According to each company’s respective annual reports, the R&D budget in 2022 was $1.9 billion for John Deere, $866 million for CNH, and $1.8 billion for Caterpillar. On the flip side, Blue River Technology raised $31 million prior to its acquisition by John Deere for $305 million in 2017. In 2023 their technology is finally becoming more widely available, demonstrating the significant effort required to bring new technology to market.
Development costs aside, farmers, growers, and commercial farming operations operate on thin margins. Investments in new equipment are hard to justify if the business case isn’t a slam dunk. Co-ops and equipment as a service can ease the investment, but the adoption of this tech is still slow. Getting traction in a traditional, proud, generational industry like agriculture means meeting the needs of many different use cases. Smaller farms, which arguably benefit more from robotics and automation, simply can’t afford the cost to acquire the next big thing.
Close the gaps by investing more than just money in robotics
Fully autonomous farming operations are the holy grail, but just like fully autonomous vehicles on our roads, building this capability is costly and time consuming. How do we attract more talent and shine a light on what an amazing challenge we face? Do we incentivize companies to deliver it? The auto industry is hastily chasing fully electric vehicles in part due to carbon emissions regulations. What about tax breaks or economic incentives for startups and established companies alike to invent in farming? John Deere was the keynote speaker at the Consumer Electronics Show 2022, a smart move to immerse themselves in the innovation taking place in other industries.
Compare the money and effort being put into electric vehicles to that of farming equipment. Federal incentives provide some relief. We need to ask ourselves why and what can be done about it. Regulations force this in the auto industry.
A familiar story to most is the launch of Tesla’s first electric vehicle, the Roadster, in 2008. The Roadster borrowed heavily from the Lotus Elise — a small, two seater, built in the UK. Granted they were heavily modified but starting with a platform like the Elise allowed Tesla to focus on the motor, batteries and unique features of the new car. They didn’t try to do it all at once.
This approach should be considered by those inventing robotic harvesting equipment and other farm machinery. Take one or two steps forward instead of ten when possible. It helps with adoption, funding, speed, and costs. As mentioned before, more affordable and robust robotic machines are exceptionally good at one or two things.
Another example is a Swiss Army Knife compared to a good chef’s knife. Sure, the Swiss Army Knife has a knife blade that can cut, dice, and chop just like the chef’s knife. But the chef’s knife is so much better at those tasks because it is purpose built.
The future of robotics in farming is bright. New talent, new ideas, more focus and investment will help us move towards food security. People were shocked by how much work could get done when the world was introduced to gasoline powered tractors in 1892. We are living in a new era of farming in 2023 — one that takes robotics, automation, machine vision, artificial intelligence and mashes them all together to solve the single most important societal problem: feeding the world.
Insight
Podcast
News
