A control system wherein a reverse characteristic having a reverse input/output relationship to a target input/output characteristic of a control object, is preset, and a target value for an output of the control object is calculated. A command signal corresponding to the target value then is determined based on the preset reverse characteristic, and the command signal is output to the control object. The control system includes first means for calculating a command signal corresponding to a given target value for the output of the control object based on the preset reverse characteristic, outputting the command signal to the control object, measuring an output of the control object and estimating a deviation between an actual input/output characteristic of the control object and the target input/output characteristic therefor, and storing data on the deviation between the input/output characteristics as learning correction data. The control system further includes second means for reading out the stored learning correction data and correcting the command signal using the learning correction data.