<p>This paper proposes a neuroadaptive controller for guaranteed prescribed performance tracking of dual-arm robot in the presence of dynamic uncertainties and unknown object. Considering the existence of unknown terms in both the robot and the object being grasped, a comprehensive dynamic model for the center of the object was established; Subsequently, an adaptive neural controller based on backstepping technology was designed by employing a performance function that characterizes the convergence rate, the maximum overshot and the steady-state error. The significant advantage of this article is that, 1) this method makes full use of the known parameters in the model of dual-arm cooperative system, and adaptively estimates the unknown items respectively; 2) the tracking error of the dual-arm cooperative system is restricted within the prescribed boundary, and the system performance and security are guaranteed.</p>