Abstract. Road traffic modelling and simulation is currently well provided with a variety of packages dealing with the minute detail of road layouts from single isolated junction models to complete network simulations. There has also been much work in developing assignment models to optimise traffic signal sequences. The same is not true in the pedestrian modelling arena. With the exception of models dealing with railway and airport concourses and models of pedestrian movements around sports stadia there is very little support for the planner or designer of the pedestrian urban environment. The system discussed in this paper provides some insights as to the reasons for this and describes a highly parallel microscopic model called PEDFLOW (PEDestrian FLOW) which attempts to remedy the situation. The model operates on a grid size that is equivalent to the space occupied by a person at rest. The major difference between vehicular and pedestrian movement is that the former really has only one degree of freedom, forwards, whereas a pedestrian has unlimited two-dimensional degrees of freedom. Vehicular travel is governed by a large number of laws and regulations that make it much easier to model. Within the pedestrian urban environment there are very few laws and regulations and those that do apply are related to interactions with vehicles. The design of PEDFLOW is discussed and it is shown how the complex behavioural rules governing pedestrian movement are captured. The parallel architecture underlying the model is described and it shows how the maximum possible parallelism is achieved among all the moving pedestrians at any one time. The performance of the model is then presented and uses to which the model is being put are then briefly presented.