The accurate prediction of the pedestrian trajectory is necessary to endow automatic guided vehicle with the capabilities to adjust velocity and path dynamically for the navigation in real pedestrian scenes. For this purpose, this study presents a social conscious prediction model considering two main factors that affect the pedestrians’ walking in the crowd – relative distance and moving direction. To form an effective model, the authors’ conscious pooling layer is added to the Long Shot Term Memory network (LTSM) model to build the relationship between pedestrians, learning the current position m and movement trend. The experiments are conducted to compare the proposed model with the previous state-of-the-art model on several public datasets. The experimental results show that the proposed model predicts pedestrian trajectories more accurately.