Human-Machine Interaction (HMI) has an impact on passenger’s trust. A structural equation model was constructed to study the influence path with the Automated Vehicles system as the object. The six latent variables of vehicle cognition, performance expectation, perceived ease of use, perceived risk, trust repair, trust and 20 HMI observation indicators are used to describe passengers' expectations for HMI and measure trust. The network SP questionnaire is used to obtain data. Finally, the model results are analyzed according to the observed index influence path coefficient. The research results show that the above-mentioned latent variables have significant effects on trust. Except for the negative correlation between perceived risk and trust (-0.18), the influence path coefficients of the other latent variables on trust are in the order of vehicle cognition (0.44)> performance expectation (0.27)> perceived ease of use (0.14)> trust repair (0.10), perceived risk and performance expectation are positively correlated (0.26), trust repair can reduce the distrust of perceived risk by 24%, and perceived ease of use is helpful to carry out trust repair (0.68).