The invention discloses a method for designing a model predictive controller of a driverless automobile, which comprises the following steps: (1) establishing a tire sideslip angle model alpha; (2) designing a model predictive controller based on lateral deflection angle compensation and arc prediction; Step (3), predicting the output of the next time according to the speed of the vehicle and thecenter and radius of the arc; Step (4), constructing an optimization problem according to the predicted output and the given reference output; Step (5), solving the optimization problem (8), obtainingthe optimal control quantity U* (k) at each sampling time in the control time domain, and applying the first component of U* (k) to the system. The front wheel side deflection angle compensation of the invention rapidly obtains the front wheel deflection angle of the vehicle, and greatly improves the accuracy of the kinematics model under the bad weather and the high-speed driving state; Arc prediction can compensate the state error caused by the linearization of the model in the prediction process, so as to improve the prediction accuracy of MPC based on kinematics model, thereby improving the control accuracy and expanding the application scope.