With the continuous development of intelligent driving technology, the demand for high-precision vehicle status information is becoming increasingly urgent. Road gradient is a key parameter for vehicle operation, having a significant impact on the vehicle's dynamics control. High-precision and low-latency road gradient estimation is a prerequisite for precise control, which can effectively enhance the intelligence level of the vehicle. Adaptive extended Kalman filter (AEKF) is a commonly used algorithm for road gradient estimation, but it has certain limitations in complex operating conditions with different noise levels. This paper proposes an improved adaptive Kalman filter algorithm, which enhances the estimation accuracy of road gradients in complex conditions by setting dynamic noise scaling factors. Through simulation tests under double lane change conditions and steady-state circular motion conditions, the effectiveness of the proposed method is verified, achieving a road gradient estimation accuracy with a root mean square error (RMSE) of less than 2°.