The Impact of Machine Learning in Agriculture

Machine learning is now a big part of every single one of our lives. When you use Netflix, recommended shows are presented based on an AI algorithm. Your order history on Amazon is run through a program to create a list of potential products that are uniquely suited to your tastes. Marketers use automation as a way to reach potential prospects and keep current customers engaged with their company. Believe it or not, 79 percent of the top businesses are currently using AI in their business model in some way.

At a glance, it seems like everything related to machine learning involves relatively new products and technology. A new study by the Journal of Field Robotics at the University of Cambridge is challenging our preconceptions about AI and how it ought to impact our lives. Their research and experiments resulted in a tremendous breakthrough in agriculture. We are going to take a look at the challenge, their new robot, and what this means for the future.

The Challenge

The article, which is summed up by the title, “A field‐tested robotic harvesting system for iceberg lettuce” takes a look at one of the challenges the agricultural industry is facing — harvesting crops. Beyond wheat, there are only a tiny handful of products that could be collected by a machine.

Produce like strawberries, cucumbers, kiwi, and lettuce were nearly impossible to harvest using machines because they are extraordinarily sensitive. If the robot isn’t calibrated just right, it would destroy the fruit, and wasting money in the process.

Pressure sensitivity is just the start of the problems when working with machines while harvesting. Another huge issue is that it’s nearly impossible for the robot to decide which produce is considered ready to collect, and what’s not.

A New Robotic Harvester

The team behind the study developed a robot that they call Vegebot. The machine is comprised of a laptop, which contains the control software, a six-degree-of-freedom robot arm, two cameras, and an end effector. Their goal was to use the software to teach and program Vegebot to harvest iceberg lettuce when it’s ready for harvest.

Vegebot uses both cameras as a way to solve two of the biggest problems associated with AI harvesting. One camera is mounted to the top and is used to identify which heads of lettuce are ready to get picked. They provided countless photo samples of lettuce that’s appropriate for harvesting, and the program uses the top camera to analyze the field and determine when it will harvest the iceberg.

The bottom camera is connected to the end effector. The purpose of this camera is to help Vegebot judge the distance when cutting lettuce. Finally, the robot arm is calibrated and tested to ensure that when it squeezes the lettuce for harvesting, it doesn’t damage the produce. All of this is controlled through the software on the laptop.

The Results

Countless hours went into training Vegebot to successfully harvest iceberg lettuce. The team created controlled experiments inside the lab and proceeded to retrain the bot after every session. As the accuracy improved, so did the quality standard. Data points were added regularly until Vegebot was ready for its first field run.

The first couple of attempts in the field proved to be challenging. There are certain conditions that the people conducting the study had to adjust for as they made the transition. Two of the biggest challenges they faced was getting Vegebot to harvest correctly during different types of weather, and determining where to cut the lettuce on uneven ground.

When the results were released, many saw a bright glimmer of hope for the potential use of AI for harvesting crops. The concluded that Vegebot was able to determine lettuce that was ready to harvest with a 91.0 percent accuracy. The error rate for false positives was 1.5 percent.

So, what does this mean?


The results show there are plenty of promising uses for AI as it relates to agriculture. We already know that AI has completely disrupted industries live customer support chat and music. Another study revealed that by 2020, a staggering 80 percent of businesses hope to have chatbot technology on their website. Couple this statistic with the fact that countless music and tv applications are running on AI, and it’s not hard to see the bright future we have with this technology.

If we can master things like machine learning in entertainment, it’s a safe bet that we will eventually implement AI in agriculture on a large scale. This study was just a taste of things to come. We may still need humans for other agricultural tasks, but the need for living people to harvest crops is coming to a close.

This UrIoTNews article is syndicated fromDzone