A calibrating technique is provided for the position/attitude or only the position of an arm tip of a robot, such as an articulated type of robot. At plural positions, respective n pieces of errors (Δφn, wherein n is a positive integer larger than a value obtained by dividing the number of unknown parameters by 6 or 3) are calculated. Each error is a difference between a position of the arm tip measured and a position commanded by control. An inter-error difference (Δεy (1≦y≦n−1)) between a reference error (Δφm (1≦m≦n)) arbitrarily selected from the n-piece errors (Δφn) and other errors (Δφx (x≦n, except for m)) other than the reference error (Δφm) is calculated. A parameter, which is a basis for calculating the inter-error differences (Δεy), is made to converge until a sum of absolute values of the inter-error differences (Δεy) becomes within a given threshold (ε0 (ε0>0)).