A Hybrid Framework for Real-Time Validation of Public Transport AVL and APC Data Streams

Abstract

The increasing adoption of telematics technologies in public transport systems has enabled the continuous collection of operational data through Automatic Vehicle Location (AVL) and Automated Passenger Counting (APC) systems. These data streams support real-time monitoring, service planning, and performance evaluation. However, raw operational data are frequently affected by positioning inaccuracies, communication delays, asynchronous updates, sensor malfunctions, and missing observations, reducing their reliability for operational decision-making. Consequently, the transformation of raw telemetry into trustworthy operational information remains a significant challenge for transport authorities and operators.

This paper proposes a hybrid validation framework for the real-time processing of AVL and APC data streams in urban public transport systems. The framework combines multiple complementary validation mechanisms, including GTFS-based map matching, geospatial geofencing, edge-based sensor event detection, and stateful heuristic validation logic. Unlike approaches that rely exclusively on spatial filtering or statistical anomaly detection, the proposed framework integrates spatial, temporal, and operational context to improve data consistency and reduce the impact of noise and hardware anomalies.

The framework was implemented and evaluated using operational data from a metropolitan public transport network. The proposed approach enables the identification of unreliable vehicle positions, exclusion of inactive vehicles located in depots, validation of stop-related events, and stabilization of vehicle tracking under conditions of GPS drift and asynchronous updates. The results demonstrate that combining multiple validation layers improves the robustness of operational analytics and provides a reliable foundation for real-time performance monitoring. The proposed methodology is independent of specific hardware vendors and can be integrated into existing GTFS-based transport analytics platforms.