This paper outlines a multivariable control system design approach for industrial robots based on computer-aided frequency-response techniques. A configuration-dependent multivariable controller is derived which is shown to be related closely to various nonlinear decoupling controllers proposed in the literature. Simulation of the robot and controller is used as a design aid to assess whether typical performance criteria are met in the face of constraints. It is noted that, in the nonlinear regime of actuator saturation, the multivariable controller can be suitably modified to maintain performance. The stability and robustness of the total system are discussed. Implementation experience reveals the necessity of incorporating integral action into controller designs on final convergence to a setpoint. The final controller is demonstrated to work well in simulation and in hardware implementation. Studies on the hardware system yield further robustness conclusions.