The electronic mechanical braking system directly affects the driving safety of automobiles, but there are auxiliary control methods. Therefore, a fault-tolerant control algorithm for automotive electronic mechanical braking ramp assistance is proposed. The faults of the electronic mechanical braking ramp assistance system and select sensor faults are analyzed as the main research object. By introducing a BP neural network, the longitudinal speed of the vehicle, the wheel speed, the difference between the target slip rate and the current slip rate, and the rate of change of the difference are taken as the network input. Online learning is used to adjust and update the weight coefficients of each layer of the network, output fault diagnosis results and optimized PI controller parameters, and achieve fault-tolerant control. Three different types of sensor fault signals were artificially added in the experimental testing. After the proposed fault-tolerant control method, the controlled signals were used to replace the fault signals, ensuring the normal driving of the car.