Estimation of Mobility Model Using Limited Mobility Data

Abstract

In Japan rural area conventional public transport does not meet the needs of non-driving population, especially the elderly. To solve this problem, it is essential to obtain precise mobility data. Mobility data refers to measurements of the movements of individuals, including where they come from and where they go, as well as how long they stay at their destinations. However, for privacy reasons, in many cases, projects that measure mobility data do not aim to track specific individuals; instead, they simply describe how many people are present within a larger geographic area during a specific time frame. In addition to that, it is expensive to obtain precise measurements for an entire area. Therefore, this study proposed a method to estimate an area-wide probability model of mobility from precise but limited data. The proposed method generates the origin and destination of a trip in the following two steps. First, the starting point is generated from population density distribution. Next, a conditional probability distribution for the destination at the starting point is derived from the mobility data in the vicinity of the starting point, and the destination is generated. In a concrete manner, the destination is generated by resampling the destination of the mobility data nearby the starting point. The proposed method was demonstrated with GPS mobility data of 106 people. Using the proposed method with geographical distributions of several factor such as population, commercial and public facilities and more mobility data with attributes, more precise area-wide mobility models would be derived.