This paper studies the adaptive asymptotic tracking control problem of robot manipulators with unknown parameters and external disturbances. An novel adaptive command filtered backstepping strategy is given, in which the command filters are used to avoid the differential calculations of virtual control functions in conventional backstepping, and the error compensation mechanism is established to compensate for filtering errors. Compared with the existing command filtered backstepping tracking control for robot manipulators, the tracking error can asymptotically converge to the origin, and only one adaptive parameter needs to be designed. Finally, the simulation example is given to verify the effectiveness of the proposed scheme.