Integrating perspectives of mobility for spatially-explicit Agent-Based Modelling

点击翻译按钮获取中文摘要

Agent--Based simulations of mobile objects are widespread, from modeling movement of people inside buildings to animals (and animal groups) in habitats. One integrating factor in simulating movement via Agent--Based Models (ABMs) are wayfinding algorithms, although implementation varies by disciplines. Given that there have been recent calls for integrated approach to mobility, this proposal suggests 3 multidisciplinary strategies to improve wayfinding in ABMs: (1) Utilize least cost -surfaces to generate 3D topology (2) Utilize high frequency movement data to extract probability surfaces guiding movement, and (3) Combine the methods developed above with visual models from pedestrian simulations (e.g., exosomatic visual architecture) to consider multiple factors governing motion. For step one, recent developments, including high resolution satellite imagery and 3D building models, afford the creation of least cost surfaces that can be used to develop detailed wayfinding algorithms with very high spatial and temporal accuracies which are also efficient. Moving in this direction, GIScience research has utilized 3D building models to develop 3D topology to support emergency evacuation planning, and making use of 3D topology for emergency evacuations from multi-storied buildings is a recurrent theme of several journal publications. They primarily work by identifying the shortest path (using standard algorithm such as Djisktra's) in this 3D network, with the goal of showing the user an efficient egress route from their current location to a building exit point. Least cost surfaces can be utilized to develop a global model of spatial distance between points of origin and destination, and then converted into a vector--based 3D topology. The next step is to devise a method of generating "probability surfaces", which utilizes high frequency movement data for both animals and humans. The volume of this data, combined with high resolution remote sensing data, provides an opportunity for the use of geospatial operators and machine learning to extract probabilities of locations where movement may occur (presented as a continous raster layer). The final step will be to link behavioural and visual concepts (e.g., exosomatic visual architecture) from pedestrian simulation and extend to movement ecology, for example, by deriving virtual 3D environments from high resolution satellite data and improving animal behavioural simulations based on a frontal perspective. It is expected that the new algorithms will prove useful in building mobile agents across a wide spectrum of applications. These include ABM models of building evacuation using highly detailed 3D building models, pedestrians in a detailed neighbourhood, and animal movement in their habitats. Combined with higher spatial resolution datasets and 3D building models, such improvements are almost a necessity to improve ABM simulation accuracy as well as reduce computational loads.