Phenotyping is the process of observing the physical characteristics of a plant in order to assess its health. Although a necessary part of crop breeding, it's also a very tedious and time-consuming task, particularly when you've got a whole field full of plants to inspect … it's exactly the sort of job that it would be good to get a robot to do. Led by Prof. Stephen P. Long, a team at the University of Illinois has created just such a beast.

The semi-autonomous device is guided by a combination of GPS and a human-operated laptop, as it moves between rows of plants on its all-terrain tank-like treads.

It's equipped with sensors including hyperspectral, high-definition and thermal cameras, along with weather monitors and pulsed laser scanners. These allow it to collect phenotypic data such as stem diameter, plant height and leaf area, along with information on environmental conditions like temperature and moisture content of the soil.

That data is stored on the robot's own integrated computer, plus it's transmitted to the user's laptop. They can then use the information to create a 3D computer model of each plant, to create predictive models of its growth and development, and to estimate the biomass yield both for that individual plant and for the whole crop.

So far the robot has only been tested on energy sorghum, which is grown as a source of biofuel. It is believed, however, that it should perform just as well with other tall-growing crops such as corn or wheat.

Prof. Girish Chowdhary, who is collaborating on the project, is now working on making the robot skinnier, so it can fit between plant rows more easily. He also plans to equip it with an obstacle-avoidance system. Ultimately, the team hopes to have a product on the market by 2021, in the US$5,000 price range.

"What we need is to be able to describe a plant as it grows," say Long. "You could do that perhaps with an army of people, but now the robot can do all of that for you."

The robot can be seen in use, in the video below. And for something that's sort of half-way between robotic and manual phenotyping, check out Kansas State University's Phenocart.

Source: University of Illinois