Precision Agriculture Trends and Realities

World agriculture is confronting a major challenge: in order to sustain the population that will exist in 2050 and end chronic worldwide food insecurity, agriculture will need to produce 60% more food on essentially equivalent arable land. Producing high yields of a narrow suite of cereal grains using indiscriminate quantities of inputs and water as well as pesticides and heavy tillage can no longer be sustained. The needs of regenerative agriculture must be met by a confluence of sustainable energy, robotics, communication, analytics and field sensors now known collectively as precision agriculture, or AgTech. As part of the transformation, AgTech must confront old problems previously managed with techniques and products no longer acceptable for regenerative farming. One of those legacy problems is unwanted invasive plants, or weeds.

Farm robots take on an old headache for growers: weeds

Farm robots (technically Unmanned Surface Vehicles, or USV’s) are highly visible indicators of the revolution taking place in agriculture known as precision agriculture. Other sectors of precision agriculture may get more attention from venture capital, but do not fire the public’s imagination like a shiny robot in a field. Farm robots incorporate some of the latest technologies, such as artificial intelligence (AI), machine learning (ML), Internet of things (IoT) as well as advanced image acquisition/analysis and an amazing variety of sophisticated end effectors purpose-built for the farm task at hand. Development of these systems is complex, and can be expensive and time-consuming.

Many investors and others new to AgTech are often surprised at the number of farm robotics developers that are focused on the mundane task of weed abatement. In 2017, John Deere paid $305M, or about four times valuation to acquire Blue River, makers of a sophisticated system that could identify weeds from crops and deliver a precision dose of herbicide. There are many other systems now competing for market share. Why all this interest in robotic weeders? Do growers really need them enough to pay for them?

Why do growers need robotics weeders?

Many factors have come together to drive this need to kill a weed.

Regenerative Farming—Traditional farming methods with heavy tillage have left the soil vulnerable to erosion from wind and rain. Regenerative farming proposes a number of practices to “regenerate” the soil. Among those is “no-till” which recommends little or very light tilling. One issue for the grower who implements regenerative farming is the impact on weeds; tilling was an effective pre-emergent control method. Without that, the grower must rely on a heavy pre-emergent application of herbicide (called a “burndown” as it also removes remnants of the last crop). This is essentially swapping one environmental issue for another.

Herbicide-resistant weeds—In 1996, Monsanto introduced the herbicide Glyphosate (trade name Roundup®) concurrently with genetically modified soybean and corn seeds that were resistant to glyphosate. Growers had an herbicide that could be used post-emergent up to harvest, and the combination was credited with a reduction in famine worldwide. In the years since, several common weeds, notably waterhemp, feathertop Rhodes grass, sesbania and amaranth, have developed effective resistance to glyphosate, forcing growers to use other strong herbicides pre-emergent and in most cases simply deal with the weed growth or remove weeds by hand.

Drought—Drought over much of the Western United States including California has dramatically reduced the available supply of water. Weeds compete with crops for water very effectively, and may grow fast and tall enough to make a canopy over crops, depriving crop plants of essential sunlight.

Labor shortage—Immigration policies and pandemic travel restrictions have limited growers’ access to foreign labor. Growers who might have managed weeds with hand-weeding may not have access to the resources they once did.

These four factors come together to create an impossible situation for growers. They need fresh ideas. They are ready for robotic weeders.

Farm robotics and autonomous vehicle basics

Evaluating available offerings related to a specific use case or set of use cases can be complicated. It is helpful to understand some fundamentals about self-driving vehicles and some basics of robotic automation as it applies to a task like removing/killing weeds:

Self-driving (autonomous) capability—How does the robot know where to go? Many systems go with a straightforward “teach” mode, where a human operator sets the system to record, then manually pilots the drone/robot through the entire sequence. The robot can then repeat the sequence with excellent precision and repeatability. These systems still require sensors to detect unanticipated objects in the path and stop as needed. Other systems may use a combination of GPS, area transponders and image recognition/machine learning to plot a course through the crops.

Power source—Farm vehicles today are usually powered by diesel engines (about 75% of farm vehicles are diesel). While today’s Tier 4 diesels are very low emissions compared to older engines, they are relatively heavy, especially when the weight of the gas tank is included. Regular use in a field can lead to soil compaction. Some robotic systems offered today are all electric or solar-electric, which can save money and energy as well as offering a lower gross vehicle weight. The trade-off is that electric at this time does not offer the power or performance of diesel, so the end effectors for the system must be designed with this in mind to keep power consumption low.

End Effectors—There is no one ideal solution for robotic weed removal. Depending on the power source, the power provided to the tools abating the weeds can be electrical or hydraulic. Hydraulic systems are highly reliable and can have high response speeds, but they are usually associated with fossil fuel powered systems as they must be energized with compressed air from a compressor powered by a generator powered by a diesel or gas engine. Here are some common approaches:

  • Mechanical mowing—A series of rotating blades cuts off the weeds on either side of the crops. For some systems a set of scythes also cuts weeds between the plants if spacing allows. Weed roots are not removed and the cut off weeds may re-seed, but this method can be fast and relatively low power, lending itself to electric platforms. Because the system does not distinguish between plants and weeds, less compute is needed, which can make the system faster while using less electricity.
  • Mechanical pulling—Grippers that remove weeds by the roots require the system to distinguish between weed and plant, so they generally need to be programmed for each crop that is to be managed so that they can recognize crop plants at specific stages of growth. Everything else is a weed.
  • Precision herbicide—For this the system distinguishes weeds from crops similar to pulling, but then precisely applies a micro-dose of herbicide to the weed. Advantage here is that the weed is destroyed, and may not re-seed. Compute power can be quite high. It will use much less herbicide than broadcast, but may be victimized by overspray. Any home gardener will attest to weeds very cleverly growing right next to the crops.
  • Lasers—Similar to precision herbicide, except in this case the system burns each weed with a micro-laser. This avoids the overspray issue, and the weed is destroyed, but crop/weed differentiation is the same. This system will probably use the most power of the three types shown here.

For robotic weeders, an opportunity exists in Internet of Things (IoT). Several of these systems will “know” how many weeds they killed, and possibly what kind of weeds they killed. This could be useful in the selection of pre-emergent burndown herbicide. Some systems also collect data on crop plants at various growth stages, useful in phenotyping, yield estimation and basic crop management. All of these data might be transmitted in near real time or batch for post analysis and predictive analytics.

For artificial intelligence and machine learning, most of the systems available are learning systems; true AI is not yet a reality. That said, many systems are collecting excellent feature identification, which may one day lead to weed removal and high-speed diagnosis/treatment of crop plants for specific issues. As the world-wide need for more food from less land grows, robotic farm systems will be there to help meet the need.

The intelligent, automated future of farming

At AquaSpy, we believe in the intelligent, automated future of farming through AgTech. There are a lot of exciting innovations taking place. These new ideas and technologies plus the resulting shared data stems that from them will help create a better, more sustainable world for our future on this planet. That’s why we’ve joined the Global AgTech Alliance because it’s important that we collaborate to facilitate the adoption of precision and digital agriculture products and practices.