Abstract: In this work, we develop a control system that learns from human driving behavior, utilizing FIFO buffering and LSTM networks to process a series of inputs, including previously buffered data, current sensor readings, and future pitch predictions, for axle torque and deceleration values estimations. Validation tests with real sensor data from a Chevy Bolt EV confirmed the system's predictive accuracy, comparing favorably with ground truth measurements on rural roads. Furthermore, its capability for future pitch prediction includes the removal of additive noise with expectation values of zero, making both measured and predicted pitch values less prone to additive noises. Statistical analyses, including central distribution measurements, residual computations, and correlation evaluations, further validated our approach, demonstrating the system's ability to replicate human-like decision-making in driving on rural roads. This underscores the proposed method’s potential in enhancing autonomous driving by ensuring safety, comfort, and efficiency, closely mirroring the intuitive responsiveness of a human driver.