Exjobb/Thesis - Real-time Simulation and Control of a Tractor–Implement System in AGX Dynamics with ROS2 Integration

Autonomous agricultural machinery requires advanced control and rigorous testing in safe, controlled environments. Simulation provides a cost-effective path to evaluate systems prior to field deployment, but it demands real-time capability, sensor integration, realistic disturbance modelling, and the ability to validate controllers in closed loop. AGX Dynamics is well suited for vehicle/terrain and contact-rich scenarios; together with ROS2 it enables a modular research platform.

Väderstad AB is a Swedish agricultural equipment manufacturer specializing in highperformance tillage and seeding machinery. Their products are used worldwide to improve efficiency and precision in farming, and the company is heavily involved in developing the next generation of smart and autonomous implements. By including a tractor–implement system in this thesis project, the simulation work directly aligns with Väderstad’s longterm vision of integrating advanced control, automation, and AI technologies into future agricultural operations.

Objective

Develop a real-time capable simulation environment in AGX Dynamics where a tractor (towing vehicle) with an implement (e.g., cultivator) follows a predefined path across a field under realistic disturbances, integrate it with ROS2, and design and evaluate baseline control strategies for tractor path following and implement regulation. The platform shall remain extensible for future work in system identification and AI-based methods.

Proposed Goals

• Model a tractor and rear implement in AGX Dynamics (kinematics, masses, contacts, basic soil/ground interaction).

• Integrate ROS2 for control inputs, sensor outputs, and logging with low latency.

• Identify and implement the most relevant disturbances (terrain slope/incline, varying soil conditions, implement load).

• Control development (baseline): 1

– Tractor path following: lateral (heading/cross-track) and longitudinal (speed)

control.

– Implement control: maintain target working depth and tool force under varyingfield conditions.

– Provide at least one model-based strategy (e.g., linearized LQR or cascaded

PID with feedforward).

• Demonstrate closed-loop performance under disturbances and report quantitative metrics.

Expected Results

• A documented, executable AGX + ROS2 simulation environment (code & manual).

• Baseline controllers for tractor path following and implement control.

• Evaluation under disturbances: path tracking error, depth/force control error and energy consumption.

• Clear recommendations for future extensions (system identification, domain randomization, advanced/learning-based control).

Application
Please submit your application along with your CV no later than December 1, 2025.

Selection is ongoing, and the position may be filled before the application deadline.

Please note that we do not accept applications via email.

Contact

Academic Supervisor: Viktor Uvesten, Industrial PhD student, Division of Automatic Control

Linköping University / Väderstad AB

Viktor.Uvesten@vaderstad.com

Industrial Supervisor: Tomas Röjder, Algorithm Developer

Väderstad AB

tomas.rojder@vaderstad.com

Vi drivs av att göra saker lite bättre. Lite enklare. Lite mer hållbart. Att med våra lösningar möjliggöra en hållbar livsmedelsproduktion så att världens lantbrukare kan förse vår växande befolkning med mat idag och imorgon. Det är ingen liten sak. Vad drivs du av?