The rotary inverted pendulum is a nonlinear non-minimum phase system and its identification is a challenging problem due to its complexity and unpredictability. The modified particle swarm optimization algorithm (PSO) is utilized in this paper for identification of the dynamics of the system. The main goal is to estimate the accurate mathematical model of the experimental single inverted pendulum with friction and minimize the estimation error between theoretical dynamics and practical data. The proposed model and exact identification can be adopted to design the desired controller for the system. The proposed modified PSO can overcome the parameter convergence that exists in the conventional evolutionary algorithms such as Genetic Algorithms (GA). Simulation and experimental results demonstrate acceptable mach between predicted and actual outputs and verify the efficiency of the method.