A method for controlling a plurality of manipulators, such as multiaxial or multiaxle industrial robots. At least one manipulator functions as the reference manipulator and is moved in a plurality of preset poses within its working area at which internal position values are determined as first desired poses. For each desired pose, subsequently a first actual pose of the reference manipulator is determined by an external measuring system. Subsequently at least one further manipulator moves up to specific poses of the reference manipulator as second desired poses and for each of these poses an actual pose of the further manipulator is determined by an external measuring system. On the basis of actual-desired deviations between the thus determined desired and actual poses of the two manipulators, subsequently a parameter model for the further manipulator is established and with it it is possible to compensate simultaneously both its own errors and those of the reference manipulator. The method can be used in conjunction with a simplified, improved cooperation between manipulators.