A Typical Day for the Road Network
Lead Research Organisation:
University of Southampton
Department Name: Faculty of Engineering & the Environment
Abstract
The Transportation Research Group at the University of Southampton (www.southampton.ac.uk/trg) and Siemens Mobility (Poole, UK) are pleased to be able to offer a fully-funded PhD studentship focussing on using 'big-data' and 'data-mining' approaches to analyse real-time and historical traffic data. The project would suit a student with a strong mathematical/engineering/computing background, prior experience of working with large/diverse datasets would be an advantage, but not essential.
In the road traffic industry real-time data is collected from a diverse variety of sources, including sensor data from individual vehicles, meteorological data and time context data (e.g. school/university terms). Effective traffic control increasingly depends upon the way in which these data sets are combined/analysed, to predict future traffic and enable traffic managers to act proactively to prevent congestion. While there is a general consensus that different types of day (for example a 'wet Tuesday during the school term') generally follow the same underlying traffic patterns, as there is always something atypical going on in any road network, it is a non-trivial problem to define these underlying patterns.
This project therefore seeks to use historical traffic data archives to quantify underlying patterns and trends, to develop algorithms to compare real-time data to the patterns, to increase the speed of detection of incidents within the traffic network and ultimately enable reduction of congestion in an increasingly busy world.
In the road traffic industry real-time data is collected from a diverse variety of sources, including sensor data from individual vehicles, meteorological data and time context data (e.g. school/university terms). Effective traffic control increasingly depends upon the way in which these data sets are combined/analysed, to predict future traffic and enable traffic managers to act proactively to prevent congestion. While there is a general consensus that different types of day (for example a 'wet Tuesday during the school term') generally follow the same underlying traffic patterns, as there is always something atypical going on in any road network, it is a non-trivial problem to define these underlying patterns.
This project therefore seeks to use historical traffic data archives to quantify underlying patterns and trends, to develop algorithms to compare real-time data to the patterns, to increase the speed of detection of incidents within the traffic network and ultimately enable reduction of congestion in an increasingly busy world.
People |
ORCID iD |
| Benedict Waterson (Primary Supervisor) | |
| Jonny Evans (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/N509358/1 | 01/10/2015 | 31/03/2021 | |||
| 1734072 | Studentship | EP/N509358/1 | 12/07/2016 | 30/06/2019 | Jonny Evans |
| Description | The main discovery in this research project is that road traffic incidents can be detected more easily by algorithms when they incorporate contextual data. Road traffic incidents are undexpected events that disrupt traffic flow, eg accidents, breakdowns, illegal parking etc. Contextual data is data onevents that can be expected to cause disruption in traffic conditions, eg weather, school dates, sporting events, public holidays etc. A key part of the methodology of the developed incident detection algorithm was a traffic forecasting algorithm names RoadCast. This algorithm improved on the state of the art by incorporating contextual data to improve forecast accuracy. The original objective of this research project was to understand what a 'typical day in the road network' was, and to use this insight to improve incident detection. This project has achieved that by using RoadCast to forecast traffic in whatever particular day it is, and then detect incidents by looking for discrepencies with real-time traffic data. |
| Exploitation Route | Both the traffic forecasting and incident detection algorithms are planned to be implemented by Siemens, this project's industrial sponsor. In an academic sense, it has demonstrated that incident detection algorithms require contextual data in order to account for certain variations in traffic data, and that improvements in performance can be gained by incorporating them. Incident detection algorithms are needed more and more by transport management centres to detect incidents. Existing manual methods such as monitoring CCTV are becoming unsuitable as transport networks increas in size, and funding decreases. Effective incident detection algorithms allow traffic management centres cover larger transport networks with fewer resources by helping operators detect incidents more quickly and effectively. This improvement in incident detection algorithm performance will be a step towards implementing more effective algorithms in traffic mangement centres, allowing incidents to be detected more quickly and effectively. Ultimately, this will result in less disruption from incidents on road networks. |
| Sectors | Transport |
| Description | The two main algorithms developed in this research are a traffic forecasting algorithm and incident detection algorithm. Both are planned to be implemented in Siemens systems, this project's sponsor. Firstly, Siemens won funding to use the traffic forecasting algorithm to better understand congestion. It is also planned that this traffic forecasting algorithm will be implemented within Siemens traffic management software, which will be used in traffic management centers across the UK. Finally, a trial of the incident detection algorithm in a traffic management center is on the way. The outcome of this trial will shape whether the incident deetection algorithm is also to be implemented in Siemens traffic management software. |
| First Year Of Impact | 2017 |
| Sector | Transport |
| Impact Types | Societal,Economic |