The contour error (CE) of machining processes is defined as the difference between the desired and actual produced shape. Two major factors contributing to CE are axis position error and tool deflection. A large amount of research work formulates the CE in convenient locally defined task coordinate frames that are subject to significant approximation error. The more accurate global task coordinate frame (GTCF) can be used, but transforming the control problem to the GTCF leads to a highly nonlinear control problem. An adaptive robust control (ARC) approach is designed to control machine position in the GTCF, while additionally accounting for tool deflection, to minimize the CE. The combined GTCF/ARC approach is experimentally validated by applying the control to circular contours on a three axis milling machine. The results show that the proposed approach reduces CE in all cases tested.