The robot uses GPS for navigation, and there are two small cameras mounted on a frame on top of the machine to give the robot depth perception, just like a human, said Lei Tian, agricultural engineer at the U of I. "If he sees a weed, he can actually tell how far away it is."
An on-board computer offers access to information that provides the morphological features of plants, to help the robot determine just what is and isn't a weed. Once a weed is identified, a robotic arm attached to the front of the machine engages a device the researcher calls "a custom-designed end effector."
There are two layers to the device, according to Tian. One layer cuts the weed, while the second layer applies herbicide to the cut weed.
"This type of application is extremely effective," said Tian, "because it applies herbicide directly to the plant, instead of broadcasting uniform rates across a field."
With this level of precision, Tian says the system has clear environmental benefits. In addition to cutting herbicide use, chemicals do not drift off-target when placed directly on the plants.
The original inspiration for this robot came several years ago, when Tian and two graduate students were working on remote sensing systems for a CFAR (Council on Food and Agricultural Research) project.
"We were collecting field data from satellite imagery, such as soil moisture and plant conditions, but we needed to have ground reference data to validate that information," said Tian.
"But that kind of data is tedious to collect," he continued, "and it's very hot work. The grad students who collected this information stayed in the field m
Source:University of Illinois at Urbana-Champaign