Abstract
This thesis focusses on evaluating the design of an Elevated Pedestrian Network (EPN) using space-centric human-sensing methods. Speci?cally, a sensor network installed throughout said EPN collects information about pedestrian ?ows. This data is then input into a pedestrian simulation for visualisation and analysis of said ?ows. A preliminary analysis using pedestrian simulations in AnyLogic was conducted using fabricated data to achieve the following purposes: (1) to clarify the data required to create said simulations and to be collected by the sensors, and (2) to identify potentially congested areas in the network, in order to direct research resources to these areas. The Control Experiment was performed to establish a point of reference in order for e?ective comparisons to be made with Experiments 1 and 2, and assumed pedestrian arrivals of 10,000 pedestrians per hour, with a 70:20:10 ratio of pedestrians leaving, entering and remaining in the EPN in the morning, and a corresponding 20:70:10 ratio in the evening. In comparison, Experiment 1 increased expected pedestrian arrivals to 15,000 pedestrians per hour. On the other hand, Experiment 2 altered the pedestrian arrivals at each origin point while keeping the ratios of pedestrians leaving, entering and remaining in the EPN constant. In all three experiments, 100 iterations of the simulation were performed. For each iteration, pedestrian densities at three locations (chosen to represent critical conditions in the EPN) were obtained from the simulation every minute, and compared against Fruin’s (1971) Level-of-Service (LOS) criteria. Using pedestrian LOS, a set of evaluation criteria was formulated and used to determine the likelihood that pedestrian ?ows would satisfy requirements set out by Singapore’s Land Transport Authority (LTA) for adequate pedestrian facility design. In the morning, congestion levels for all experiments were likely to be satisfactory. Decreases in pedestrian LOS were experienced in Experiments 1 and 2, as compared to that of the Control Experiment. However, these were well within the limits set by LTA. In the evening, congestion levels were higher in general than that in the morning, with the most critical conditions expe-rienced in Experiment 2. In conclusion, the AnyLogic models created over the course of this thesis form a promising foundation for future development, and can be improved in their ability to model actual conditions at the EPN. Some suggestions for extending these models include, but are not limited to, (1) extending the network to include buildings which are soon to be completed, (2) di?erentiating pedestrians’ speeds by age and gender group, and (3) testing the in?uence of alternative scenarios on pedestrian LOS experienced.