Development of a Proof-of-Concept Data Acquisition and Analysis Framework for Agricultural Machinery Operations in a Virtual Simulation Environment

Nowadays, agricultural machinery operation requires combined mechanical, digital, and data-interpretation skills. This creates a need for tools for operator training and performance assessment in order to assess operator behaviour. This study developed and evaluated a proof-of-concept integrated telemetry and controller-logging framework for Farming Simulator 25. The system combined a custom game-side telemetry modification (mod) with an external recorder that captured machine, implement, spatial, work-state, and operator-input data within a synchronised 1 Hz dataset. Inputs from a Thrustmaster T128 steering wheel and pedals and one SimTask FarmStick P were recorded through DirectInput, while controller devices were polled at a higher frequency before aggregation. The framework was tested in four simulated operations (cultivation with a pulled cultivator, ploughing, power harrowing, and wheat harvesting with a combine harvester). The logger reconstructed trajectories, speed profiles, working distance, estimated treated area, work-state duration, state transitions, steering, pedal use, FarmStick activity, and button records. Mean speed ranged from 3.10 km h−1 for ploughing to 10.93 km h−1 for cultivation, while estimated treated area ranged from 0.086 to 1.258 ha. Mean absolute steering input ranged from 0.164 to 0.204. The four operations produced 4–17 work-state transitions, demonstrating sensitivity to differences in implement activation, lowering, working, and idle states. A relevant data-quality limitation was that approximately 35–36% of integrated rows contained stale telemetry, indicating that the recorder maintained the 1 Hz timeline but did not receive a fresh data stream at every sampling instant. The results demonstrate the technical feasibility of transforming a commercial agricultural simulator into an instrumented research environment. Further validation with multiple operators, repeated trials, expert assessments, and real GNSS, CAN-bus, and ISOBUS data are required before the framework can be used for formal learning-outcome or operator-competence assessment.