Robust slip-ratio regulation for agricultural tractors with transformer-enhanced speed estimation

In heavy-draft field operations such as plowing, spatially varying soil properties and traction-load fluctuations can cause excessive tractor wheel slip, complicating longitudinal-speed estimation and slip-ratio regulation. To address this issue, this study proposes a robust slip-ratio regulation method integrating Transformer-enhanced extended Kalman filtering (EKF) with boundary-layer adaptive super-twisting sliding mode control (ASTSMC). First, a Transformer network adaptively scales the EKF process-noise covariance using causal temporal features, thereby adjusting the fusion between the model prediction and the non-ideal onboard speed observation to improve longitudinal-speed estimation. Second, the estimated speed is used to calculate the slip-ratio feedback for a boundary-layer ASTSMC, which combines adaptive gains with a continuous saturation function to improve robustness and reduce torque chattering. Finally, the Transformer-EKF was evaluated using full-scale tractor field data, and the slip-ratio controller was validated through hardware-in-the-loop testing on an NI-PXI real-time platform. Evaluation results showed that, during multi-gear-shift transport, the Transformer-EKF reduced MAE and RMSE from 0.1814 and 0.2403 km/h to 0.1286 and 0.1668 km/h, respectively; during steady operation, they decreased from 0.2764 and 0.3153 km/h to 0.1579 and 0.2167 km/h. HIL tests showed that ASTSMC reduced front- and rear-wheel slip-ratio settling times to 0.386 and 0.830 s, respectively, versus 0.834 and 1.672 s for PID, 0.756 and 1.098 s for conventional SMC. ASTSMC also reduced the corresponding steady-state standard deviations by 32.8% and 30.2% relative to PID. Taken together, the field-data estimation results and HIL control results demonstrate the potential of the proposed framework for adaptive tractor slip-ratio control in unstructured farmland environments.