The invention discloses a road slope real-time estimation method for an automobile, which aims to solve the shortcomings of the prior art, such as many quantities of calibration, heavy influence of automobile acceleration and deceleration, large use of sensors, high cost, much dependence on automobile parameters, complex models and poor robustness. The real-time estimation method comprises the following steps: 1, initializing signal, initializing sensors after each power supply, and correcting the original errors of data; 2, pre-processing input signal, including 1) threshold limit, 2) slope limit, 3) low-pass filtering and 4) sliding average filtering; 3, correcting longitudinal acceleration signal; 4, based on Kalman filtering algorithm, calculating the parameters needed for weighted fusion and conducting data fusion by determining the reliable degree of acceleration signals and the situation of the automobile going up and down slope; and 5, correcting the slope and outputting the optimal estimate of the road slope.