Forrester robot uses artificial intelligence to help diagnose sick trees

Left
to right: Maksim Mikhailov and his team: Daniil Nechaev, Igor
Lositsky and Gleb Zagarskikh, as they accept the gold medal at
the 2017 World Robot Olympiad.
ITMO
University

After hearing a radio program
describe the labor-intensive work of forest pathologists —
basically, tree doctors — Maksim Mikhailov had an idea: what if a
robot helped collect their data?

Mikhailov is a 16-year-old
student at

ITMO University

, the renowned science and technology
institution in St. Petersburg, Russia. As a member of the
school’s Youth Robotics Lab, he was perfectly positioned to bring
his idea to life.


With a full team working on the the
project

, the robot won
the gold medal at last year’s World Robot Olympiad; it can record
tree locations within a forest, identify their species, measure
the widths of their trunks, and even identify if a tree is
healthy or not.

Its name is Forester, and most of
its job is to explore forests and hit trees with its mallet. It’s
a robotic adaptation of a technique that human tree experts often
use, called “sounding,” to help their appraisal of a tree’s
health.

“The robot hits a tree and its
microphone records the sound,” Mikhailov explained. “Since sick
trees have cavities or low wood density in their trunk, they make
a sound with a lower overall frequency than that of a healthy
tree.” The robot makes use of an algorithm that analyzes the
recorded sound to determine if it came from a healthy
tree. 

A closer look at the
Forrester robot.
ITMO
University

Forester also takes a photograph
of the tree and feeds the image to a neural network, identifying
12 different species of trees with accuracy better than 90
percent.

This student invention is “a
great idea,” says Lee Dean,

lead arborist at Cornell
University

,
simultaneously cautioning that “trees are living, dynamic
systems.” He identifies the robot as a tool for the human
arborist, not an automated solution that will render their work
useless. “Tree risk assessment is qualifiable, not quantifiable.
This can give indicators about a tree’s health, but can’t make
the diagnosis.”

This jibes with Mikhailov’s own
perception of his team’s creation. “While the robot can collect
data about the trees, it cannot analyze that data to decide what
needs to be done in order to preserve forests,” he said. That
task falls to the human experts, whose jobs are perhaps made
simpler by a data-collecting robot.

An arborist’s work appears to be
safe from automation — at least, for now.