Multi-modal Modeling of Urban Mobility

Design and efficacy of a routing protocol for mobile ad-hoc networks strongly depends on the underlying mobility pattern of the mobile nodes. Counter to the perception of mobility being random in time and space, we show that spatial modality is a key trait of human mobility. Most of the mobility of an individual can be captured by a small fraction of all locations, called spatial modes, visited by the person. Further, we show that there is strong predictability of spatial modes in time. We combine these properties into a mobility model that accurately captures more than 90 % of the mobility of a typical individual. Any realistic mobility model must either derive out of real traces or be compared back to them. In developing our mobility model, we extensively analyze publicly available cell tower connectivity traces of urban users. We apply the technique of multidimensional scaling to fill in missing location information in the traces with high reliability. We believe this approach will make available a wealth of connectivity traces for mobility analysis which are otherwise considered unfit for the task. Keywords-mobility modeling; trace analysis; ad-hoc networks; vehicular networks