‘This is the discovery process in its purest form’

“We spent a lot of time testing possible issues, configuring the software, and learning how it works.”

-Adrian Ramirez ’25

Farmers have acres of fields to look after, and it can be hard to spot an insect infestation before crops endure significant damage. What if a robot could help?

That’s a question computer science Professor Gary Holness and students are tackling in Clark’s robotics lab, the Laboratory for Intelligent Perceptual Systems.

Over the summer, Holness, Adrian Ramirez ’25, and Richard Geli, M.A. ’25, worked on an experimental prototype consisting of a team of four mobile robots networked together to model collaborative foraging in fields scanning for pests.   Spending countless hours designing and building the system from scratch in the lab in Clark’s Center for Media Arts, Computing, and Design, Holness is building an experimental prototype in which to expand upon the research and test out various ideas.

Holness is interested in owning a farm someday, inspired by his parents, who grew up on farms in an agricultural community.  He’s also drawn to this project because farms provide an interesting setting for machine learning, his area of expertise.    “A robot operates quite successfully in a structured environment like a factory floor or even semi-structured environments such as a warehouse where much is controlled but some aspects involve uncertainty.  Structure means that everything is engineered specifically,” says Holness. A farm does not provide strict structure, as terrain varies, crops occur in various shapes and sizes growing at different paces, and it abounds in the activity of laborers, equipment, and animals.  “A field is unstructured, so it’s a challenging environment,” he adds. “There are a lot of pressing needs specifically for digital farm management.”

The robots will focus on surveying fields using an information-driven probabilistic algorithm, video cameras, specialized sensors, and laser scanning to identify areas exhibiting indicators of crop damage within the local environment. This mechanism could help farmers locate and map areas with pest infestations.

Ramirez and Geli had to configure, implement, and modify several aspects of the physical robot platforms, compute infrastructure,  as well as all of the underlying software.  “It’s actually a lot harder than it seems,” says Ramirez, an economics and data science major. “We spent a lot of time testing possible issues, configuring the software, and learning how it works. It’s a great feeling to know that you can understand how to control something like this.”

Geli, who is pursuing a master’s in information technology, was excited to learn more about it.  “It created a lot more confidence in me,” he says. “I’ve developed more interest in software for such complex systems and now I have a stronger foundation I want to build upon.”

Holness says this kind of system development (robotics) involves several custom technologies.  “Robotics is very wonderfully messy,” Holness says. “There are many opportunities for things to go wrong, and they often do.” Ramirez and Geli learned about operating systems and networking using the Linux system and robotics middleware software.   They also learned about several topics in robotics.

After graduation, Geli hopes to focus on research in a lab setting. Ramirez would be happy to continue working in robotics.  Holness says the students worked on concepts drawing upon material covered in at least five different courses at the undergraduate and graduate levels.   Moreover, much of the specific work does not exist in textbooks.   “This is hands-on learning and the discovery process in its purest form,” he says.