Authors: Arianna Rana, Antonio Petitti, Angelo Ugenti, Rocco Galati, Giulio Reina, Annalisa Milella
If you’ve ever tried to model a real agricultural robot inside a simulator, you already know the struggle: high-detail mechanical dynamics rarely play nicely with real-time performance, and outdoor environments introduce messy, unpredictable terrain. This new study by Rana, Petitti, Ugenti, Galati, Reina, and Milella digs precisely into that challenge.
The team explored how to build a reliable digital twin of off-road agricultural robots by comparing two powerful simulation worlds: Gazebo, known for its strong ROS integration, sensor realism, and real-time behavior, and MSC Adams, respected for high-fidelity mechanical modeling and detailed track–soil interaction.
And they didn’t keep it theoretical. They modeled two real platforms—a tracked robot (Polibot) and a four-wheel-drive/steering rover—equipped with GNSS, IMU, RGB-D cameras, and LiDAR, and tested them in a simulated vineyard row to evaluate row-following performance.
The results reveal clear strengths and trade-offs. Gazebo shines when real-time, sensor-driven digital twins are needed, making it ideal for closed-loop control development. Adams, meanwhile, captures terramechanics with impressive detail—pressure distribution, wheel sinkage, and track behavior—but its computational load makes real-time DT updates impractical. By applying the same control law across both frameworks, the authors deliver a true apples-to-apples comparison, showing how different terrains (sand, loam, hard ground) affect stability, convergence, and robot behavior. Their findings point toward Gazebo as the more practical DT foundation, especially if paired with future AI-based soil interaction models.
For AgRibot, this research offers a valuable roadmap: choosing the right simulation tools, understanding when high-fidelity physics matter, and building digital twins ready to handle the unpredictable realities of agricultural fields.
Read more on Zenodo.