Over the last few decades, there were numerous model developments in the field of pedestrian flow simulation. Most of these models try to simulate the pedestrian's evacuation in buildings in case of fire or emergency situations. They also look at capacity issues at bottlenecks such as rail interchanges and bi-directional flow. A number of models for pedestrian movement have been developed in a variety of disciplines. There are two distinct groups of models in macroscopic and microscopic perspectives. The macroscopic models focus on the system as a whole, while microscopic models study the behaviour and decisions of individual pedestrians, their effect on other pedestrians around them, and the system as a whole. There is a need to model the pedestrian behaviour for a range of applications including event planning, resource usage, and urban planning. For instance, the organizers of a large event in an exhibition hall require information on what areas are likely to be congested so that management strategies can be developed and tested before starting the event. Similarly, the designers of a shopping mall might be interested in how people move around their intended design so that they can place shop entrances and seating in useful locations. Recently, computer based analysis for pedestrians movement in buildings is widely adopted by the fire researchers or engineers in performance-based fire engineering study. However, the pedestrians exhibit different behaviours depending on their knowledge of the environment and other personal characteristics. Unlike the rules that govern vehicular traffic, there are few formal procedures or rules that govern the pedestrian movement, resulting in often complex and chaotic movements. Pedestrians are not restricted to lanes or specific routes. In general, they are restricted by the physical boundaries around them such as the width of doorways or presence of walkways and also the movements of their neighbours. As a consequence, the modelling of pedestrian movements presents some specific problems not encountered in other forms of transport modelling. A full understanding of crowd behaviours normally requires exposing real people to the specific environment for obtaining empirical data, which is difficult because of such environments are often dangerous in nature especially in emergency situations. Moreover, the major deficiency of the existing pedestrians modelling is the adaptation of crowd behaviour that is extremely difficult to be described by mathematics. In addition to studying the crowd behaviour based on the observations and the historical records, computer simulation may be a useful alternative that can provide valuable information to evaluate a building design, to help planning process, and for dealing with emergencies. In this thesis, a modified computing pedestrian model, namely Improved Social Force Model, has been developed to simulate the perception and the cognition of a pedestrian in case of emergency evacuation. The algorithm of the model is implemented based on the Social Force Model introduced by Helbing and Molnar (1995). This model examines pedestrian movements as either positive or negative social fields, in which a pedestrian behaves as if acted upon by external forces. However, the decisions and interaction between pedestrians is an extremely flexible and intelligent process. To provide more accurate results in pedestrian behaviours, the physical features of pedestrian movements such as walking speeds, acceleration, queuing, and herding behaviour must be accurately reproduced. The motivation of this research comes from the need to understanding pedestrian psychology and modelling pedestrian behaviour accurately. Some parameters involving human behaviours will be introduced into the original Social Force Model in order to improve the accuracy for the computing modelling. In summary, two human behaviours will be added into an individual's walking performance: herding behaviour and visual angle. By adding a herding parameter into a pedestrian in an evacuation simulation is proved to be obtained a more accurate result. However, herding behaviour could not be the same in all age of pedestrians. In an Improved Social Force Model, the younger pedestrian, the larger herding behaviour; as younger pedestrians have lower judgement in the wayfinding (searching for a suitable escape route) so that they will follow the actions of others as a guide to determine how they might act. In addition, under evacuation situation, pedestrians will more concentrate on finding their destination (e.g. an exit) in order to leave the building as soon as possible. Therefore, applying a visual angle of ±85° to the pedestrians can achieve a more accurate simulation result. Applying the concept into the algorithm of the proposed model, the predicted values for each pedestrian and time step are calculated. The model consists mainly of three terms. These terms are the desired velocity of motion of a pedestrian, the interactions between pedestrians, and the interactions between pedestrian and boundaries. Human behaviours, herding parameter and visual angle, are applied to the second term when processing the algorithm. Then the last step in the algorithm is to update of the position, velocity and acceleration of an individual for the next time step. In proofing the performance of an Improved Social Force Model, two other computational pedestrian movement models, Social Force Model and Simulation of Transient Evacuation and Pedestrian MovementS (STEPS), were applied. Social Force Model is a well-known pedestrian movement modelling and it was published in the Journal of Nature in 2000 by Helbing et al.; Improved Social Force Model is proposed by the author in order to improve the prediction of pedestrian movement modelling which is based on the Social Force Model; STEPS is a commercial application of simulation in pedestrian movement and it is well-validated in academic and well-adopted in industry. Therefore, STEPS is being a benchmarking model in the research so as to evaluate the performance of the Improved Social Force Model. From the computational results of the pedestrian simulation for various scenarios, it was found that the simulated escaping time of pedestrians in the proposed model from the fire room is longer than the original Social Force Model by comparing to the results from STEPS. The agreement of the proposed model is not expected to be perfect since there are important variations among collected by different authors under different layouts, situations, and cultures. However, an improvement is clearly visible.