Virtual field robot contest uses NASA technology

The International Field Robot Event was launched in 2003 by Wageningen University in the Netherlands. The field robot contest tests pioneering robotics and precision farming technologies under real-world conditions, and gives young scientists the opportunity to exchange and develop their ideas alongside their peers.

The FRE event has been held at the DLG-Feldtage exhibition since 2014. DLG stands for Deutsche Landwirtschafts-Gesellschaft – German Agricultural Society. This year’s contest will be held live on DLG’s digital platform due to the Covid-19 pandemic and promises to provide demanding virtual tests for the robots and their inventors.

14 international student teams

Automation and digitization, optimized crop production as well as sustainable agriculture are among the agricultural challenges that have been selected for the virtual field robot contest. Already the event has attracted 14 international student teams from Denmark, Germany, the Netherlands, Romania and Slovenia, who will present their virtual robots in action, performing complex field maneuvers as part of the simulation tasks.

Offering conditions corresponding closely to those that would be found in real-life farming situations, like the presence of obstacles and weed, the simulation environment produces high-quality lighting, shadows and textures, making it not only realistic but entertaining as well. The 3D robotic simulator “ROS / Gazebo” offers the simulated environment in which the field robots and their algorithms can compete.

_{Text continues underneath image}

Simulated maize fields

Similarly to the real-world field event, the teams and their robots have to solve several compulsory tasks. Contestants receive the test environment, which consists of simulated maize fields, shortly before the event starts, requiring that the robot already possesses a range of skills that have been programmed as control algorithms in advance.

Recognizing and removing weeds

Navigating the robot through a curved crop row, recognizing objects that are weed or simply rubbish and mapping the objects using geo-referencing are part of the challenges that are solved using the algorithms. Interpreting the information, like determining what is weed using a camera is complex, and requires many months of prior development work.

A further task involves removing the weed, where accurate software-control of the actuators that operate the tool is needed. The challenge is to make intelligent sense of the data on the day, which involves the robot interpreting the given information and make optimal decisions.

Simulation software also used by NASA

The simulation environment ROS / Gazebo is open-source, supporting software code developed for sensor and actuator simulation. ROS / Gazebo is a simulation environment employed for a number of technological competitions such as DARPA Robotics Challenge, Virtual RobotX Competition and NASA Space Robotics Challenge.

Students can use sensor models within Gazebo to “see” the simulated environment. A laser scanner typically measures the distance to locate the various objects while a camera will interpret the scene, for example watching out for something that might be weed.

Challenges and fun

“When a robot is designed, a simulation model with algorithms is developed long before the field robot is turned into hardware,” says Prof. Hans W. Griepentrog, Director, Department of Technology in crop production, University of Hohenheim, Germany. “Software simulation is therefore part and parcel of the field robot development, which means that the student teams participating in the Field Robot Event are very much at home in this simulation environment. The challenge for them here is to bring all the required algorithms into a whole,” adds Prof Griepentrog.

_{Text continues underneath image}

“The agricultural tasks are always a challenge for the teams and their robots, as technical and agricultural process knowledge is essential. However, the fun of the competition should not be neglected,” comments Prof. Griepentrog.

A virtual event, the contest requires field robots to perform the compulsory tasks in a simulated field environment. In the freestyle optional task, the teams can live-stream their physical robots performing an agricultural task of their choice.