In order to effectively depict the characteristics of bidirectional pedestrian flow, a novel pedestrian flow model is proposed based on cellular automata. At first, according to direction gain, velocity gain and herding gain, the calculation formula of target position is defined, and the walking rules by combining overtaking behavior and herding behavior are given in the model. Meanwhile, the actual channel is used for experiments, where the self-organizing effect formed by pedestrian flow is observed. The simulation platform is established to study the key factors influencing pedestrian flow characteristics. The numerical analysis results showed that when the pedestrian density in the channel reached to the critical degree, the overtaking behavior can easily produce jamming. Moreover, pedestrians' rational choice is good for relieving jamming.