Present on-line control schemes for robotic manipulators require high computing power to perform real-time estimation and adaptation and/or an exact model of the manipulator. These requirements result in impractical control schemes for real industrial manipulators. This paper presents a new robust tracking control technique for industrial manipulators in the presence of various uncertainties. It does not require an exact model of the manipulator and it compensates for uncertainties in the system dynamics, such as friction, and uncertain inputs including load variations. The controller consists of two portions: one for the nominal part of the system, and the other portion for uncertainties compensation. This control scheme is simulated for a General Electric P50 robot and the results are presented.