Sensing data fusion is one of the most important technologies in autonomous driving. Its performance depends on advance communication technology. Cellular-Vehicle to Everything (C-V2X) initially defined as LTE V2X in 3GPP Release 14 is a solution for vehicle communication that includes Vehicle-to-Infrastructure (V2I), Vehicle-to-person (V2P), and Vehicle-to-Vehicle (V2V). Although 4G LTE and 5G provides high-speed transmission, packet loss and delay are still inevitable. Packet loss and delay affect the safety of autonomous driving, especially for the judgment of emergency. In this paper, we compare the accuracy of data fusion under different rate of packet loss and broadcast frequency on the simulated platform CALAR. And we propose a skill to improve accuracy. Experiments show that the proposed skill significantly alleviates the effect of communication packet loss and delay on the accuracy of V2X data fusion.