In order to realize the emotional autonomy of emotional robot, a generative autonomous emotionally intelligent robot system (GAEIRS) is proposed. The structure of GAEIRS is composed of perception module, emotion recognition and understanding module, emotion generating module, emotion expression module and feedback adjustment module. By collecting, analyzing and predicting multi-modal fusion information through sensors, the system identifies and judges emotion type, intensity and duration, and generates autonomous emotion expression information for human-computer interaction. The unique contribution of the proposed GAEIRS lies in the emotionally robot can carry out emotional interaction independently according to external stimuli, and the emotional expression content is rich and natural, and the matching is accurate, so as to achieve the effect of human-machine empathy. Finally, the application prospect and security risks of GAEIRS are discussed, including the potential application of GAEIRS in various fields such as medical treatment, education and entertainment, as well as ethical issues.