Preprint
Multimodal AI
Featured

A Multi-Modal Distributed Real-Time IoT System for Urban Traffic Control (Invited Paper)

Khanam, Zeba(BT Research), Achari, Vejey Pradeep Suresh(Keele University), Boukhennoufa, Issam(University of Essex), Jindal, Anish(Durham University), Singh, Amit Kumar(University of Essex)
January 1, 2024Leibniz-Zentrum für Informatik (Schloss Dagstuhl)14,341 citations
PDFDOI

14k

Citations

0

Influential Citations

Leibniz-Zentrum für Informatik (Schloss Dagstuhl)

Venue

2024

Year

Abstract

Traffic congestion is one of the growing urban problem with associated problems like fuel wastage, loss of lives, and slow productivity. The existing traffic system uses programming logic control (PLC) with round-robin scheduling algorithm. Recent works have proposed IoT-based frameworks that use traffic density of each lane to control traffic movement, but they suffer from low accuracy due to lack of emergency vehicle image datasets for training deep neural networks. In this paper, we propose a novel distributed IoT framework that is based on two observations. The first observation is major structural changes to road are rare. This observation is exploited by proposing a novel two stage vehicle detector that is able to achieve 77% vehicle detection accuracy on UA-DETRAC dataset. The second observation is emergency vehicle have distinct siren sound that is detected using a novel acoustic detection algorithm on an edge device. The proposed system is able to detect emergency vehicles with an average accuracy of 99.4%.

Analysis

Why This Paper Matters

Urban traffic congestion is a persistent global problem, causing fuel waste, lost productivity, and delayed emergency responses. Traditional PLC-based systems with round-robin scheduling are inefficient. Recent IoT-based approaches using traffic density have low accuracy due to lack of emergency vehicle image datasets. This paper tackles both issues by proposing a distributed IoT framework that leverages two key observations: rare structural road changes and distinct emergency vehicle sirens. The multimodal approach (visual + acoustic) is practical and deployable on edge devices, making it relevant for smart city applications.

Technical Contributions

The paper's main innovations include:

  • Two-stage vehicle detector: Exploits rare road changes to achieve 77% accuracy on UA-DETRAC, a challenging benchmark.
  • Acoustic emergency vehicle detection: A novel algorithm on edge devices that detects siren sounds with 99.4% accuracy, overcoming dataset limitations.
  • Distributed IoT architecture: Processes data locally on edge nodes, reducing latency and bandwidth usage for real-time traffic control.

Results

The system achieves 77% vehicle detection accuracy on the UA-DETRAC dataset, which is competitive given the dataset's complexity. More impressively, emergency vehicle detection via acoustic signals reaches 99.4% accuracy, demonstrating robustness. These metrics are reported without direct comparison to baselines, but the high acoustic accuracy is a clear strength.

Significance

This work advances multimodal AI for IoT by combining vision and audio in a resource-constrained edge setting. It directly addresses a real-world problem—traffic management—with a deployable solution. The high acoustic detection accuracy could inspire further research in siren-based systems, while the two-stage detector offers a practical trade-off between accuracy and computational cost. The framework's scalability and real-time capability make it a strong candidate for smart city infrastructure.

iottraffic-controlemergency-vehicle-detectionacoustic-detectionedge-computingvehicle-detectionmultimodal-ai
Neural Networks

Related Papers