In process control industries, the control of a nonlinear conical tank is a benchmark problem. The dynamics of the plant changes as the level of the tank changes due to the shape of the tank. A trade-off exists between robustness and tracking performance with 1-DOF controllers; we can get either a robust controller or controller with good transient performance but not both at the same time. 2-DOF controllers were introduced which solve this problem and deliver both robustness and transient performance. Three 2-DOF PI robust schemes were compared Kappa-Tau, model-reference robust tuning (MoReRT) and Butterworth rules with genetic algorithm. Butterworth gave high degree of robustness to varying plant dynamics and a transient performance which was close to that of the PID, while PID becomes unstable for changing plant dynamics, thus showing poor robustness.