With the growth of the world’s population showing no sign of slowing down, George Kantor of Carnegie Mellon University hopes robots can produce the food we need.
Food and robotics are being intertwined in such a way now that our very future could depend on this partnership as our world heads towards a possible global population of just under 10bn by the year 2050, according to the UN.
This, of course, means more food is needed and, based on our production and available arable land, this could prove to be an insurmountable problem with catastrophic consequences.
One person hoping to use robotics to solve that problem is George Kantor, a senior systems scientist at Carnegie Mellon University (CMU) where he leads projects that bring research ideas from multiple disciplines to develop new robotic systems that solve interesting, practical problems.
After receiving his PhD in electrical engineering from the University of Maryland, College Park, he moved to CMU as a postdoctoral student. He has spent more than 15 years working in the area of field robotics (literally), through agriculture, mining and scientific exploration.
What inspired you to become a researcher?
I have been interested in electronics for as long as I remember. The concept of ‘closing the loop’ by using sensor feedback to control automated systems grabbed my attention as an undergraduate, where I was inspired seeing feedback controllers used to do things like balance an inverted pendulum, or having a vehicle automatically follow a line.
Feedback seemed to me to be the basis for machine intelligence, so I spent years as a graduate student learning how to analyse the behaviour of systems under feedback, and design effective feedback controllers.
This involved learning a lot of maths, and when I came to CMU I was surprised to discover that the same kind of maths used in feedback control is good for the problem of estimating things through indirect observation.
Can you tell us about the research you’re currently working on?
I started working in agriculture around 2000 with a project that was using wireless sensor networks to collect environmental information in container nurseries.
The network measured soil moisture, temperature, sunlight and several other parameters, and we developed methods to use that information to help make better irrigation scheduling decisions.
That work eventually led to a large project sponsored by the US department of agriculture to develop intelligent irrigation.
Borrowing ideas and technologies from CMU’s successful DARPA Urban Challenge self-driving car, we developed an autonomous utility vehicle that could drive up and down the rows of an orchard to perform various tasks, such as mowing and harvest assist.
This led to even more advanced information collection and sensing, such as our ongoing sorghum breeding work with Clemson University, where we are using robots to collect phenotype information about plant growth during breeding field trials.
In your opinion, why is your research important?
World population is growing faster than our capacity to produce food, which is leading us toward a global food crisis in the middle of this century.
We are already using most of our available arable land and fresh water for agriculture. We need to produce more food on less land, with less energy and water.
Robotics and artificial intelligence (AI) can help this problem in two ways.
First, by collecting new kinds of information to help plant breeders speed up the rate at which they are increasing plant yield while also selecting for factors such as drought tolerance, pest resistance and nutrition.
Second, management techniques that use high-resolution crop information in real time will allow us to make better use of the resources we currently have.
What commercial applications do you foresee for your research?
Precision agriculture tools for high-value crops such as apples and grapes is an obvious near-term market.
Plant phenotyping tools to support breeding programmes for large agronomic crops is another.
What are some of the biggest challenges you face as a researcher in your field?
It is hard to get funding to try new and unproven things. It is difficult to find problems with the right degree of difficulty. It is easy to come up with problems that are too difficult to solve, or to come up with problems that are solvable but won’t have a meaningful impact.
We want problems that are solvable and also will have a big impact if solved, and you need to talk to a lot of people with different perspectives to find that kind of problem.
Finally, it is hard to find good collaborators; people who are willing to work outside of the comfort zone of their particular domain, are open to trying new things that may threaten the status quo and people who have the patience to work together to communicate across disciplines.
Are there any common misconceptions about this area of research?
Many people are afraid that robots are going to take jobs in agriculture – this is not the goal of my research.
We are seeking ways of collecting information that there is currently no other way of collecting, providing growers with new insights into what is going on in their fields, so that they can make better decisions that result in increased yield and quality.
We are making people more effective, not replacing them.
What are some of the areas of research you’d like to see tackled in the years ahead?
The final frontier of this line of research is intelligent manipulation: the ability to carefully handle plants in the field to do delicate tasks such as harvesting and pruning. We are currently working on some simple manipulation tasks, such as grasping corn stalks to apply contact sensors.
Eventually, I would like to see AI for agriculture advance to the point where small-scale farmers growing polycultural crops can be economically competitive with the existing large-scale monoculture approach.
This will result in a more stable and diverse food system, while also providing a means of making a living for small-scale agricultural entrepreneurs.