Distributed Classification of Urban Congestion Using VANET

Al Mallah Ranwa,Farooq Bilal,Quintero Alejandro

DOI: https://doi.org/10.1109/TITS.2016.2641903

2020-12-05

Abstract:Vehicular Ad-hoc NETworks (VANET) can efficiently detect traffic congestion, but detection is not enough because congestion can be further classified as recurrent and non-recurrent congestion (NRC). In particular, NRC in an urban network is mainly caused by incidents, workzones, special events and adverse weather. We propose a framework for the real-time distributed classification of congestion into its components on a heterogeneous urban road network using VANET. We present models built on an understanding of the spatial and temporal causality measures and trained on synthetic data extended from a real case study of Cologne. Our performance evaluation shows a predictive accuracy of 87.63\% for the deterministic Classification Tree (CT), 88.83\% for the Naive Bayesian classifier (NB), 89.51\% for Random Forest (RF) and 89.17\% for the boosting technique. This framework can assist transportation agencies in reducing urban congestion by developing effective congestion mitigation strategies knowing the root causes of congestion.

Machine Learning

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the real - time classification and monitoring of non - recurrent congestion (NRC) in urban traffic. Specifically, the paper focuses on how to use Vehicular Ad - hoc Networks (VANET) to distinguish non - recurrent congestion caused by accidents, construction, special events or bad weather, etc., and to distinguish it from periodic congestion. Traditional monitoring means such as induction loops, cameras and floating car mechanisms are costly and difficult to achieve real - time monitoring when the coverage area is large, so a more flexible and cost - effective solution is needed.

### Main Problems
1. **Highly variable time, location and duration of non - recurrent congestion (NRC)**: This makes it difficult for traditional monitoring methods to conduct real - time or continuous monitoring.
2. **Data requirements and non - real - time nature of existing NRC detection methods**: These methods not only require a large amount of data sets, but also cannot be deployed in real - time. And real - time information is crucial for taking timely measures to restore normal traffic conditions.
3. **Insufficient characterization of NRC events**: Existing methods only quantify the spatial and temporal impacts of detected NRC, but fail to classify their root causes.

### Solutions
The paper proposes a VANET - based distributed framework for real - time classification of congestion components in urban road networks. The framework classifies the root causes of NRC by extracting spatio - temporal causal relationship features in traffic flow data and using machine - learning models for classification.

### Main Contributions of the Framework
- **Evaluating machine - learning methods**: Used to classify congestion components according to traffic characteristics.
- **Algorithm development**: Able to detect, identify and spread the causes of NRC through VANET.
- **Verifying inference methods**: Verified based on an extended real - world Cologne scenario simulation.

### Formula Representation
The formulas involved in the paper include:
- Observed travel time (oTT) consists of periodic delay (Drec) and non - periodic delay (Dn - rec):
\[
oTT = D_{\text{rec}}+(D_i\vee D_{wo}\vee D_{we}\vee D_{se})
\]
- Periodic delay (Drec) is the historical expected travel time (TTh) at a specific location and time:
\[
D_{\text{rec}} = TTh
\]
- If the observed travel time exceeds the threshold, the travel time is considered excessive:
\[
oTT>(1 + c)\times TTh\Rightarrow oTT\text{ is excessive}
\]

Through this method, the paper aims to provide effective congestion mitigation strategies for traffic management departments, helping them better understand and deal with non - recurrent congestion problems in urban traffic.