Probabilistic routing using multimodal data

Human tracking and prediction are pervasive nowadays. It is possible to integrate multi-source human tracking data and location based social media data, which includes spatial data, temporal data, and textual data, to make human-mobility prediction and over-crowded station detection, and then to plan convenient bus/subway routes for passengers. This study is useful in many real applications, including convenient travel route recommendation and location based services in general. We face two challenges in this study: (1) how to use multi-source human tracking data to model probabilistic transfer cost between different bus/subway lines practically, and (2) how to compute convenient bus/subway routes efficiently. To overcome these challenges, we define a set of probabilistic spatial metrics and propose a travel-time threshold and a transfer-cost threshold convenient route planning queries. A series of optimization techniques are developed to enhance the query efficiency. We also conduct extensive experiments to verify the performance of the proposed algorithms. (C) 2017 Elsevier B.V. All rights reserved.