A Novel “Smart Mobility Behavior Sensor (SMBS): A Multimodal Edge-Enabled Framework for Real-Time Urban Mobility Intelligence”

Abstract

The rapid growth of urbanization and vehicular density has significantly increased the complexity of modern transportation systems, necessitating advanced, intelligent solutions for real-time mobility monitoring and management. This study presents the Moventra Smart Mobility Behavior Sensor (Moventra-SMBS), a novel, multimodal, edge-enabled sensing framework designed to capture, analyze, and predict urban mobility behavior with high accuracy and low latency. Unlike conventional traffic monitoring systems that rely on isolated data streams and static analytics, the proposed system integrates vision-based sensing, radar and thermal inputs, environmental sensors, and real-time telemetry into a unified architecture capable of behavior-driven intelligence.

The Moventra-SMBS system employs a hybrid computational pipeline combining edge computing, deep learning, and probabilistic modeling. At the hardware level, the device integrates a high-performance edge AI processor, multimodal sensor modules, and optimized power and communication subsystems. At the software level, the system utilizes advanced machine learning models, including Convolutional Neural Networks (CNNs) for object detection, Long Short-Term Memory (LSTM) and Temporal Convolutional Networks (TCNs) for sequence modeling, and unsupervised algorithms for anomaly detection. A sensor fusion layer combines heterogeneous data streams to generate high-dimensional behavioral feature representations, enabling robust detection of complex mobility patterns.

Experimental evaluation demonstrates that Moventra-SMBS achieves high detection accuracy (~96%), strong anomaly discrimination (AUC ≈ 0.96), and balanced precision–recall performance (F1-score ≈ 0.96). The system further exhibits reliable predictive capability through time-series forecasting with uncertainty estimation, allowing proactive identification of congestion trends and anomalous events. An ablation study confirms the critical contribution of multimodal fusion and temporal learning, while system-level analysis shows significant improvements in latency and computational efficiency, enabling real-time deployment.

The proposed framework represents a shift from traditional data-centric traffic monitoring to behavior-centric, predictive mobility intelligence. By integrating sensing, analytics, and adaptive learning into a unified platform, Moventra-SMBS provides a scalable and robust solution for smart city applications, including traffic optimization, safety monitoring, and urban planning. The findings establish the system as a promising approach for next-generation intelligent transportation systems, with potential for further enhancement through distributed learning and large-scale deployment.