Abstract: Clouds are closely associated with high-impact weather. Satellite infrared radiance data provide a lot of information in cloudy areas. However, the large observation operator error caused by the initial cloud parameter errors and the nonlinear influence of clouds on the radiation process and the non-Gaussian error distribution brings challenges to the direct assimilation of satellite infrared data in cloudy conditions. Aiming at the key technical issues of direct assimilation of satellite infrared radiance data in cloudy conditions, this study reviews and summarizes the research progress of assimilation approaches, radiation transfer mode, control variables, background errors, cloud detection, observation errors setting, and bias correction at home and abroad in last 20 years. The findings demonstrate that the trend of direct assimilation of satellite radiance data is to supplement the cloud and rain information and enhance the corresponding technology based on clear sky radiance assimilation technology, to achieve the direct assimilation of satellite data under “all-weather” conditions. It is noted that there are difficulties in constructing the control variables related to cloud parameters and their background errors, as well as in removing the nonlinear influence of cloud on observation and observation operators when attempting to directly assimilate satellite infrared radiances in cloudy conditions. With the common development of observation, assimilation, and model technology, satellite infrared radiance data will inevitably play a significant role in numerical weather prediction.