Pedestrian flow in built environment has attracted considerable attention and been extensively studied, to research its characteristic is important for our life. Computational simulation is one of the main methods. In this thesis, lattice-gas model will be chosen to study pedestrian flow for two conditions. Based on the traditional lattice-gas model of biased random walkers, new improved models are established. In the new models the occupied cell of a pedestrian is divided into 8*8 small grids. This improvement brings some advantages: the discrete model is made to close to be continuous in space, and the walkers could not have to align in order, moreover for ConditionⅠ the width of exit could be more accurate, not must be an integer times of a pedestrian size. Also the influence from built environment is taken into account. Simulation model A and B are built up for ConditionⅠ, model C and D are for ConditionⅡ. ConditionⅠ is evacuation from the room. Firstly, some simulation results obtained by new improved models are compared with those by the social force model. It is found that some important phenomena which have been observed by social force model could also be reproduced by the new improved model with much less time. Then a serious of numerical study is done by the new improved model. The results show that with the width of exit and occupancy loading increasing, the evacuation time will decrease dramatically, and different steps in unit time do not change the general trend. However, when the width and occupancy loading reach a certain value, evacuation time will nearly have no change. The sensitivity of drift point, which is an important parameter for the new model, has been also investigated. It is found that with increase of the drift point forward, evacuation time will decrease and then tend to be smooth when drift point forward reaches a certain value. And the proportion of drift point forward and to go left (right) is also found to be mainly affected by the shape of the room, but slight impact from the width of the exit. ConditionⅡ is pedestrian counter flow. At first the data from a previous experimental study is used to compare with the simulation results by the new improved model with totally the same condition to found appreciate parameters. The results show that when the step in unit time is chosen as 1/2*cnt or 1/4*cnt and drift point is selected between 0.5 and 0.7, the simulation result will agree with the experimental data well. Then some numerical study for the jamming transition has been done for the pedestrian counter flow in a channel with constant density on two boundaries. The results show that the system size and ratio of total density of two boundaries have little influence on the general trend of velocity and occupancy against total density, and also the transition point has almost no change. However the value of total density corresponding to transition point is greatly affected by the step in unit time and drift point. With the decrease of the step in unit time, the transition point will decrease apparently. But with the increase of the drift point, the transition point will decrease dramatically.