Aiming at the dynamical parameter identification for robot manipulator, the artificial bee colony algorithm for identification was proposed. Considering the friction characteristics that the friction model was unable to reappear the behavior of complex dynamic friction at low speeds, the nonlinear robot model contained the nonlinear friction model was deduced by introducing the nonlinear Daemi-Heimann model. Then, the five order Fourier series was designed as exciting trajectory to collect experimental data, which satisfied velocity and acceleration boundary conditions. With the artificial bee colony algorithm, the colony bee was employed as search unit to identify unknown parameters which included 15 minimum inertia parameters and 12 friction parameters in the model through exchanging the information and retaining the superior individual. Finally, the established model was validated and analyzed, and all the results demonstrated that the proposed identification algorithm can accurately identify the dynamical parameters, and it also had highspeed convergence, strong search capability and can achieve the accurate prediction of robot driving torque. Compared with the linear dynamic model, the established nonlinear dynamical model can effectively improve the condition of sudden change about friction torque at the moment of joints commutation and can well reflect the dynamical characteristics of robot.