LocateMyBus: IoT-Driven Smart Bus Transit

Abstract

Uncertainty of traffic in cities makes it difficult for metropolitan buses to adhere to predetermined schedules, making it strenuous for commuters to plan travel reliably. The proposed LocateMyBus system leverages Internet of Things(IoT) set-ups at bus stops and buses, and Machine Learning(ML) to assuage this uncertainty by allowing commuters to track live-runningstatus of buses, disseminate tentative and live-status to commuters through Public Announcement(PA) systems at bus-stops and a web-application interface. The schedule prediction module provides a tentative schedule of buses with stop-wise arrival times estimated using ML based on historic and real-time route data. Arrival times of two bus-routes in the Massachusetts Bay Area were collected for a period of four months by periodically querying its real-time General Transit Feed Systems(GTFS). This dataset was used to train and validate the proposed ML methods. The IoT system was modeled on Proteus, and validated with a miniature prototype. LocateMyBus is proposed as a step forward toward minimal intervention algorithmic set-ups to ease the uncertainty associated with bus commute in cities. It enables commuters to track live running status and avail ML-predicted tentative schedules. Furthermore, it eradicates the computation requirements of GPS-based systems, whilst ensuring stop-level tracking granularity. LocateMyBus's ability to log bus arrival times at each stop paves the way to building real-time GTFSs.

Keywords

• internet of things
• machine learning
• smart bus transit
• transit feed systems
• deep learning
• schedule prediction