A Tractor's Digital Shadow: An Exact Copy in Motion ✛
Context
Today’s farms increasingly deploy robotic vehicles that can plow fields for hours without a human. Testing new control algorithms on real machines is risky, expensive, and weather-dependent. A digital twin—a dynamic virtual shadow of a physical object—solves these problems by letting you run scenarios in the lab. Unlike the auto industry, where digital twins are standard, the technology is just beginning to emerge for tractors. The research tractor AMX G-trac, a modified Lindner Lintrac 130, served as the testbed for this virtual shadow.
Methods
The digital twin’s core was a numerical simulation environment, Mevea, specialized in multibody dynamics for heavy machinery. The tractor’s kinematic model was built by splitting the chassis, front axle, and rear axle into separate bodies. Masses and inertia tensors were assigned to each, partly from manufacturer specs, but the mass distribution was obtained in an unusual way: the 3D model from the game Farming Simulator 25 was imported into SolidWorks, and its properties estimated via Digital Image Processing. The bodies were linked by Dynamical Systems constraints with revolute joints. For signal processing, the CAN bus was used: the real tractor communicates via the ISOBUS protocol, while in the virtual version messages were passed through text files using Python scripts and a virtual Kvaser CanKing channel. Each body’s motion obeys Newton's second law, and rotation follows the equations of Leonhard Euler, forming a system of differential equations solved in real time.
Results
Several turning tests were conducted—from normal steering to crab mode, where all four wheels point in the same direction. The real tractor and its digital shadow traced circles on virtual and real stubble. The average deviation of the front axle steering angle stayed under 10%, while for the rear axle it reached 36% in some modes—partly due to delays from file-based data transfer. The real tractor’s data was noisier than the simulation’s because nonlinear dynamics of tires and ground irregularities are only approximate in Mevea. The minimum turning radius of the digital model was less than 6 meters—about a meter less than the spec sheet value; the difference was attributed to significant lateral tire slip in the simulation. The acceleration test exposed the main flaw: the real tractor has a hydrostatic CVT that smoothly keeps engine rpm in the peak torque band, while the twin’s simplified stepped transmission behaved differently. The Stochastic modeling nature of approximate inertia parameters added uncertainty to longitudinal dynamics.
Implications
The results confirm that even with incomplete input data, a digital twin can accurately replicate a tractor’s lateral behavior. This opens the door to pre-testing control theory algorithms—such as path planning or lane keeping—without ever leaving the lab.
Future development
The next step is to flesh out the shadow with details. Developers plan to incorporate hydraulics, implements, and a more realistic CVT transmission, possibly through co-simulation with MATLAB Simulink. Better sensors like lidar will enable testing of environmental perception.
Impact
The technology will influence the development of autonomous agricultural machinery, enabling faster, cheaper deployment of precision farming systems.
Next steps
The immediate task is to obtain accurate inertia tensors and CVT parameters, and to streamline CAN data exchange by eliminating the text-file middleman.
Key open problems
The digital twin touches on the fundamental problem of parameter identification in complex mechatronic systems: without reliable mass and friction data, even the most advanced physics model will remain inaccurate.
The knowledge machine's view
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors built a digital twin of the AMX G-trac tractor in the Mevea environment, importing geometry from Farming Simulator 25 to estimate inertial parameters. They conducted turning and acceleration tests, comparing the behavior of the real machine and the simulation: during turns, the front axle angle deviation did not exceed 10%, but acceleration revealed discrepancies due to the simplified transmission.
Where its strength is
A key strength is the practical implementation of a digital twin for agricultural machinery, with an original approach to obtaining inertial data from a game 3D model, paving the way for rapid prototyping.
Where it could go next
- One could consider adapting uncertainty modeling methods, akin to those used in drone simulators, to better reproduce the stochastic dynamics of tires and soil.based on: Imitation Drone Dodges Black Holes of Uncertainty arXiv:2607.13436
- It would be interesting to compare this approach with digital twin work in other fields, such as biomedical imaging, where model calibration from limited data is a challenge.based on: Digital Twin Predicts Patient Anatomy arXiv:2608.00831v1
- To improve the accuracy of longitudinal dynamics, one could draw on ideas from chaotic rule reading methods, which help uncover hidden patterns in nonlinear systems.based on: Stellar Pinball: An Algorithm That Reads the Rules of Chaos arXiv:2607.16763v1
Worth trying
- Test replacing the stepped transmission with a hydrostatic variator model with feedback on engine speed.
- Investigate how using a continuous environment representation, similar to a radar sculptor, affects the modeling of tire-soil interaction.
- Assess the feasibility of applying energy-efficient data exchange algorithms to reduce latency in the virtual CAN channel.
Why this may matter
The development of accurate digital twins for agricultural machinery will accelerate the adoption of autonomous systems in the agro-industrial sector, reducing testing costs and enhancing safety.
Nearby works in our archive
- Imitation Drone Dodges Black Holes of Uncertaintysimilarity 0.54
- Digital Twin Predicts Patient Anatomysimilarity 0.53
- Digital Twin: How People's Invisible Behavior Changes Planssimilarity 0.53
- Quantum Recipe for a Doublesimilarity 0.52
- Round Trip: How a Neural Network Checks Itself Without Hintssimilarity 0.52
- Stellar Pinball: An Algorithm That Reads the Rules of Chaossimilarity 0.52
The archive is sparse around this work: the closest review is only 0.54, against a usual 0.64. It looks like few have drilled nearby.
Original abstract
as written by the authors arXiv:2607.19912
In efforts to increase research efficiency and availability, a digital twin of our research tractor (AMX G-trac) is created, focusing especially on the CAN communication for data reading and actuation command following the ISOBUS protocol. Mevea Simulation Software is utilized as the foundation, providing the kinematic model and visuals, while Python is used to read and write CAN messages over a Kvaser CanKing virtual CAN channel. Various performance tests involving straight line and turning behavior are performed in both the digital twin simulation and in the real world to measure similarity. Results indicate that the Mevea model behaves very comparable in its lateral dynamics, often within 5-10 percent, but requires better data to fully capture the longitudinal aspects like acceleration. The final model described in this paper sets the table for a second iteration to include more tractor functions such as hydraulics and tractor-implement dynamics.
The tractor’s 3D body for simulation was borrowed from the computer game Farming Simulator 25—virtual farm equipment proved useful for scientific calculations.
Key numbers
- lateral dynamics deviation: 5–10%
- minimum turning radius in simulation: 5.8 m
- tractor power: 136 hp
- tractor mass: ~4000 kg