I. Introduction
Obtaining urban-scale vehicle trajectories is of great importance to understand the urban mobility, which benefits a wide range of applications such as vehicle tracking [1], usage-based electronic toll collection [2], intelligent traffic signal control [3], traffic prediction [4], and urban planning [5]. For instance, Google collects real-time mobility status from users of Google Maps to estimate traffic conditions for navigation services [6]. However, this approach may introduce biased observations due to the limited number of users and the diverse driving behaviors they exhibit, which may in turn cause unreliable navigation results. Even some users can manipulate the observation of road conditions and thereby artificially alter the navigation results (detailed in Section VII). Therefore, by adopting urban-scale vehicle trajectories, the impact caused by such biased observations on navigation services can be effectively mitigated.