Context-supported local crowd mapping via collaborative sensing with mobile phones

Positioning and tracking of mobile devices have been a fundamental building block of people-centric mobile applications. As well as global (or absolute) positioning, recognizing one's current position with respect to surrounding crowd of people at a busy station or in an event place is increasingly needed for emerging services like mobile social navigation. In this paper, we propose a novel positioning system that provides a local map of surrounding people based on sensing data gathered from smartphones in the crowd, without relying on any infrastructure or exhaustive fingerprinting. To cope with large position errors due to sensor noise and other environmental factors, we introduce a heuristic error correction algorithm based on collective activity context of mobile phone users. Analyzing recent history of the sensing data, it detects “groups” of people who move together and then corrects deviation of estimated traces of individual users by harmonizing with the traces of other group members. Through a field experiment using Android smartphones, we have shown that our error correction mechanism successfully enhances positioning accuracy by 28% (from 4.16m to 3.01m). Furthermore, we have analyzed the performance of our method in detail through extensive simulations.