ROBOCROP Grows from Seniors' Computer Engineering Studies

Computer engineering seniors Adam Calabrigo, Brian Holland and Jeremy Kerfs were looking for a culminating project that incorporated what they learned throughout their studies. They found that Robocrop, a robotic crop monitoring project, fit the bill perfectly. Representing the quintessential combination of hardware and software, the project required them to apply their knowledge of embedded systems, electronics and programming. Focusing on agricultural engineering, they felt they could make an impact because the use of automated robotics in farming is still developing. "We were excited by the prospect of making a significant contribution,” said Kerfs. "After all, agriculture is one of the industries driving California’s economy."

The team was asked to create an autonomous system that could collect field crop data — in this instance, from strawberries. From that information, farmers could calculate yield and better determine when to harvest. “The robot basically navigates the crop field on its own to count strawberries,” said Kerfs. To provide the farming background needed on the project, they collaborated with Professor Bo Liu and a team of students from the Bioresource and Agricultural Engineering Department. Each team member worked on a different component of the project, bringing their three-fold talents to fruition. Holland worked on the computer that ran the Robot Operating System (ROS) and all the software that controlled the robot. To aid in navigation and positioning, Calabrigo developed a Graphical User Interface (GUI) program using C# and .NET framework. The application also allowed farmers to select GPS waypoints that the robot could navigate between and configure when sensors were activated.

“That way the farmer has complete control when the robot takes crop readings,” explained Kerfs.

Kerfs applied machine learning concepts to enable the robot to operate autonomously. First, he built a test rig from wood scraps and bike tires to create a sturdy camera mount. More than 40,000 crop field images were collected to “train” the machine to safely move down crop rows and count accurately. 

Read the full article beginning on PAGE 6-7 in CPE's Fall 2016 Hello World pdf edition