In order to describe traffic operation characteristics of bi-directional pedestrian flow in urban pedestrian walkways and work for walking safe, authors proposed an improved cellular automaton bi-directional pedestrian flow model. This model takes into account many basic pedestrian behaviors, forward move, side stepping, position exchange, lateral-move and step back, and, using MOORE Neighborhood method, provides rules of pedestrian movement combining with the characteristics of walking process, for instance, intelligent, flexible and complicated. With computer simulation, authors studied the influence of position exchange and lateral-move on the model, the parameters of bi-directional pedestrian flow such as velocity and volume were discussed also. The results demonstrated that position exchange and lateral-move make the pedestrian behavior more comprehensive; the model could be better applied to simulate the high-density bi-directional pedestrian flow and provided reasonable traffic operation parameters of bi-directional pedestrian flow.