Model-Adaptive Event-triggering for Efficient Public Transportation Tracking

Abstract

Accurate arrival-time predictions in public transportation systems can improve the perceived quality-of-service offered and increase usage of these systems. To date, predictions were primarily based on periodic updates of mobility information that however exhibit a tradeoff between deterministic performance and system efficiency. To further improve on this tradeoff, event-triggering is emerging as a promising operations paradigm. This work describes an innovative design of a public transportation tracking system within which arrival-time predictions are made utilizing an event-triggering framework. As a first step towards this direction, behavior models are derived through extensive analysis of real mobility data. Thereafter, an event-triggering algorithm is developed to detect changes in the model in an online fashion. The efficiency and applicability of the proposed data-driven event-triggering paradigm is demonstrated through a real-world transit scenario that compares event triggering updating techniques.