A walking-type robot device and its learning method are disclosed, wherein the robot device is caused to perform walking that accords with the parameters for controlling the walking, the walking is evaluated, and the parameters are updated so that the very evaluation is enhanced. Besides, a walking-type robot device is provided with a controlling means for controlling the robot so as to cause it to perform walking that accords with parameters which prescribe the driving phase of each leg of the time of walking, an evaluating means for evaluating the velocity of the walking, and a parameter updating means for updating the parameters so that the evaluation of the walking by the evaluating means is enhanced.