Gray Failures Detection for Shared Bicycles

Abstract

Today bicycle sharing system has been widely deployed around the world to provide a convenient transit connectivity for short trips. However, due to the careless usage or incident, it is inevitable to have some shared bicycles in the bad working condition. Rather than the fail-stop failure, e.g., a broken tire, which can be easily detected by users, the gray failure, i.e., subtle faults, can only be discovered when riding the bike. Such gray failure causes bad user experience and affects the reputation of bicycle sharing system provider. Unfortunately, it is not easy to detect these gray failures. Current solution to this problem mainly relies on users’ self-reporting. However, due to the lack of incentive mechanism and the knowledge on bicycle, cyclists may submit inaccurate report or ignore the subtle faults. To provide users the working condition of bicycles in advance and reduce the maintaining cost of system providers, this paper introduces an approach to automatically detect the gray failures of the shared bicycles. To this end, we leverage the smartphone to collect and analyze the data collected from the sensor-equipped bicycle. We extract patterns from sensing signals of broken bikes, and identify the problematic components of shared bicycles. Compared with the conventional user-reporting, the proposed method is cost-effective, automatic, and easy-to-deploy. Extensive experiments based on real bicycle sharing systems show that the proposed method achieves high accuracy in determining gray failures for a variety of scenarios.