A tower crane is a nonlinear mechatronic system with complicated underactuated characteristics, which is widely used in modern construction sites. At present, most existing methods for tower cranes are proposed by linearizing the original nonlinear dynamics near equilibrium points, which are, thus, prone to suffering from unexpected steady errors due to such factors as unmodeled dynamics, imperfect friction compensation, etc., since they have not included integral terms in either controller design or stability analysis. Therefore, in this paper, an improved feedback controller with an elaborately constructed integral term is proposed for 3-D tower cranes without linearization, which can achieve both antiswing and positioning control while being able to effectively reduce steady errors in the presence of, e.g., inaccurate friction compensation. Furthermore, asymptotic stability results are proven through rigorous theoretical analysis. Owing to no linearization, the proposed controller is applicable when state variables (e.g., cargo swing angles) are not close enough to the equilibrium points, which makes it suitable for complicated working conditions. Hardware experimental results are included to verify the effectiveness of the proposed controller. Note to Practitioners —This paper is motivated by the requirement of effective control methods for tower cranes. Tower cranes are widely applied in modern construction sites to fulfill cargo transportation tasks. For such systems, the jib slew motion not only enlarges the workspace but also brings more difficulties to suppress unexpected cargo swing during the transportation process. Up until now, most existing methods use simplified system models or need exact model knowledge, which are difficult to reflect real dynamics in many practical situations. To handle these existing problems, in this paper, a novel feedback control approach embedded with an elaborately constructed integral term is presented without model linearization. By introducing the integral term, even when frictions are inaccurately compensated, steady errors can be reduced effectively and, hence, the positioning accuracy can be improved. Also, the proposed controller can handle parametric uncertainties. By applying the proposed controller, the closed-loop system achieves asymptotic results, which is rigorously proven theoretically. Finally, the effectiveness of the proposed controller is verified by implementing several groups of hardware experiments. In the future studies, we will apply the proposed method in practical applications.