Generalized Adaptive Smoothing Method for State Estimation of Generic Two-Dimensional Flows

In big cities, the proportion of slow-mode (such as pedestrian) flows in total trip demand is steadily growing every year. Along with this trend, many concerns arise about accessibility and safety. The monitoring and the management of pedestrians serve as a potential solution to maintain the resilience of the transport network. Monitoring and state estimation of pedestrian flows are crucial as a foundation for a successful crowd management support system. This paper focuses on the development of pedestrian state estimation. A two-dimensional (2-D) generalized adaptive smoothing method (2D-GASM) is presented to estimate the full state of an area on the basis of an increasing amount of available pedestrian observations in practice. The 2D-GASM method was developed on the basis of similar concepts in the adaptive smoothing method for motorway traffic, which was based on the characteristic that traffic travels forward in free flow and backward in congestion. The same mechanism is assumed for pedestrian flows. This extension accommodates the 2-D nature of the pedestrian flow and allows for the fusion and filtering of multisource data (e.g., data from counting cameras, data from wireless fidelity sensors, and GPS samples). Although focused on pedestrian flow, the approach is applicable to any generic 2-D flows, including bicyclist or mixed flows. This newly developed method is validated on the basis of trajectory data from a walking experiment at a narrow bottleneck. The test results present promising estimation performance, and possible extensions for future applications are suggested.