A multiple ultrasonic neural network position sensing system was developed to automatically predict the lateral control position for row crops. This system used six ultrasonic transducers (3 on each side of the row crop) and a pre-trained neural network model to locate the transverse center of the crop row. Tests on chosen corn plants have shown that the prediction error ranges from -3.6 to 5.8 cm. At the maximum forward velocity of about 0.8 km/h, the system functioned accurately. The symmetric arrangement of transducers achieved temperature compensation.