To improve the accuracy of data fusion system in modern air combat, an improved method based on reinforcement learning technique is developed in this paper. Firstly, the cubic spline interpolation is used for time alignment of multi-source data. Then, a reinforcement learning based data fusion method is proposed. The fusion accuracy reinforcement is realized by the error between observations and actual value. With an example, the simulation results indicate that the developed method is effective and feasible for the multi-sensor data fusion.