Gould, Nicholas and Abberley, Luke (2017) The semantics of road congestion. In: 49th Annual UTSG Conference, 04 January 2017 - 06 January 2017, Dublin, Republic of Ireland. (Unpublished)
|
Accepted Version
Available under License In Copyright. Download (774kB) | Preview |
Abstract
Most live road traffic information systems, such as Google Traffic, do not provide the user with the context of congestion. To usefully support decision making, by drivers and network managers, such systems need to provide information such as the probable cause of the congestion and its likely time span. The focus of this work is on non-recurrent congestion. We aim to develop a system that captures the semantics of road congestion by interpreting sensor data collected in the Greater Manchester region. This data consists of journey time data (collected by Bluetooth sensors) and volume, or count, data collected by induction loops. Rather than supplying information such as the current journey time on a particular road link, which is meaningless without context, we aim to provide context sensitive information such as increasing, abnormal, journey times near the football stadium, in the direction of the football stadium. Clusters of anomalous sensor readings are identified using an agglomerative hierarchical clustering algorithm in R. The main challenge is in determining which readings are anomalous. The characteristics of the largest clusters are then taken as typical of that kind of congestion causing event. Initial work has involved identifying the journey time and volume patterns of a known attractor, a football match and we aim to extend the work to automatically identify unplanned events such as road accidents, using the sensor data.
Impact and Reach
Statistics
Additional statistics for this dataset are available via IRStats2.