Communications of the ACM

P H O T O B Y K Y O D O N E W S V I A G E T T Y I M A G E S

the side of the robot to provide a visual frame of reference.

The platform uses a customized convolutional neural network to detect objects of interest and label them with bounding boxes, which are used to train and build up the system’s knowledge. While rolling between rows of plants, the camera captures the location of each fruit or vegetable, while also measuring properties such as ripeness, size, and quality grading. The data capture is done in real time, on the robot itself, without requiring access to a data center or the cloud.

The robot also uses a soft gripper,
which looks like a pair of plastic salad
tongs, that can pick a fruit or vegetable
without damaging it. The idea is to al-
low the cultivation of these types of
plants continuously and more effec-
tively than humans can do, while aug-
AGRICULTURAL BUSINESSES usually have a massive number of trackable as- sets(plants,livestock,and machinery), often oper-
ate in wide geographic areas in which
these assets are located, and are sub-
ject to operational factors often be-
yond their control, such as the amount
of sunlight or rainfall they receive, or
temperature fluctuations. As such,
agriculture is ripe for the adoption of
new technologies to help monitor and
manage assets on a granular level, and
everything from Internet of Things
(Io T) sensors, robots, and drones are
being used by farms around the globe.

The U.S. Department of Agriculture’s National Institute of Food and Agriculture notes that the farms of today are avid users of agriculture technologies such as robots, temperature and moisture sensors, aerial imaging, and GPS technology, which are more precise and efficient than humans alone, and allow for safer, more efficient, and more profitable operations.

One example of how technology enables new farming techniques is the use of robotic harvesting on indoor farms, which today account for a tiny fraction of the 900 million acres of traditional farmland in the U.S. However, these indoor farms are well suited to the growth of vegetables such as tomatoes, lettuce, and other leafy greens, are highly sustainable, generally feature an average yield per acre more than 10 times higher than that of outdoor farms, and represent a continuation of the agricultural sector’s trend toward incorporating precision agriculture techniques to improve yields and become more sustainable.

“Whether it’s indoor or outdoor
farmers, finding technologies that
drive efficiencies is a big deal for
[farmers],” says Josh Lessing, co-
founder of Root AI, a company devel-
oping a robotic platform that allows
the inspection, analysis, and harvest-
ing of leafy vine plants grown indoors,
such as tomatoes. “[A lot] of work has
been done specifically in precision
agriculture. ‘How do I reduce the
amount of herbicide; how do I reduce
the amount of pesticide?’”
Lessing notes indoor agricultural
practices expands a farm’s margins, be-
cause less is spent on pesticides, since
insects can be kept out of the green-
house. Furthermore, reducing the use
of chemicals can also limit the environ-
mental impact of the operation.

Root AI’s robot uses multiple cameras to collect color images and three-dimensional (3D) depth information on growing plants. One camera is located in the arm of the robot itself, while a secondary camera is affixed to