Abstract: A short-term traffic flow prediction model is formulated through data fusion algorithm of multisensor dynamic system based on step-by-step filtering，and comparing complexity of the new algorithm with traditional Kalman Filtering（KF） algorithm is done.Some real time traffic flow data sampled from an Australian freeway are used to prediction and simulation experiments.The results from prediction experiments indicate that the new algorithm may ensure estimate accuracy and reduce the calculations.It can also be used to predict real time traffic flow.