Several position identification methods have been used for mobile robots. Dead reckoning is a popular method, but is not reliable for measurement over long distances especially on uneven surfaces due to the accumulation error of wheel diameter and slippage. The landmark method, which estimates current position relative to landmarks, cannot be used in an uncharted environment. We have proposed a new method called"Cooperative Positioning System (CPS) ."For CPS, we divide the robots into two groups, A and B. One group, A, remains stationary and acts as a landmark while group B moves. Group B then stops and acts as a landmark for group A. This "dance" is repeated until the target position is reached. CPS has a far lower accumulation of positioning error than dead reckoning, and can work in three-dimensions which is not possible with dead reckoning. Also, CPS has inherent landmarks and therefore works in uncharted environments. In previous papers, we have introduced the prototype CPS machine models, CPS-I and CPS-II and demonstrated high performance as a positioning system in an unknown and uneven environment. In this paper, we report on the third prototype CPS model, CPS-III, that is designed specificly as an automatic floor-cleaning robot system, and its experimental result. In this system, we categorize these robots for two tasks, that is, a accurate positioning task that is achieved with 3 robots using the CPS strategy, and a floor-cleaning task that uses an omni-directional vehicle, so as to improve the efficiency of the floor-cleaning system. Experiments show that these robots can perform a floor-cleaning task in a corridor within a positioning error of 140.8 [mm] even after robots move over a distance of 101.7 [m] .