Adometer: Push the Limit of Pedestrian Indoor Localization through Cooperation

Abstract

Existing work on indoor localization achieves accuracy by conducting site surveys or deploying additional infrastructures. In this paper, we study the feasibility of designing a low cost scheme, which is based on pedestrian dead reckoning and free from these requirements. Our method targets the scenarios where there are a large number of pedestrians in the public places holding their mobile phones while walking. First, we make one key observation that positioning errors brought by different pedestrians are diverse and can be mutually compensated. Second, we leverage the acoustic communication ability in mobile phone to enable users to exchange their locations when they meet with each other within a short distance. The received location information is crowdsourced to get a better location estimate. Finally, we use extensive experiments to evaluate the proposed scheme; the experiments are conducted in a 3,500 square-meter area with more than 18 hours (more than 100 km) real walking traces. Results show that our method is able to reduce the median localization error to 4 m when the user density of the area is more than 50/3,500 per square-meter.