The present invention provides a method of building a smart vehicle control model, and a method and apparatus for controlling a smart vehicle, wherein the method of building a smart vehicle control model comprises: acquiring sample data which comprise corresponding steering wheel turning angles under driving environments; extracting vehicle state features and road condition features from the sample data; using the extracted features to train a neural network model to obtain the smart vehicle control model. The method of controlling smart vehicle comprises: extracting vehicle state features and road condition features of a vehicle to be controlled; inputting the extracted features into the smart vehicle control model to obtain a steering wheel turning angle; controlling the vehicle to be controlled using the steering wheel turning angle. The present invention builds the smart vehicle control model in a machine learning manner, does not require manual parameter regulation and reduces man power costs caused by parameter regulation.