Toggle Main Menu Toggle Search

Open Access padlockePrints

Using real-time aggregated data sets to continually improve prediction by neural networks

Lookup NU author(s): Professor Margaret Carol Bell CBE, Dr Fabio Galatioto, Dr Graeme Hill



This paper describes a new approach, which is designed to be able to evaluate the impacts of TDMS (Traffic Demand Management Strategies) on congestion and the environment. Performance measures will be based on a combination of data from pervasive sensors, namely motes measuring pollution, carbon monoxide [CO] and nitrogen dioxide [NO2], and noise, and legacy systems including SCOOT (Split Cycle Offset Optimisation Technique), traffic loop detectors, AURN (Automatic Urban and Rural Network) and meteorological conditions. The proposed framework has been evaluated using a case study area in Leicester. Analysis of pollutant levels measured by inexpensive static (located on street furniture) pervasive sensors known as motes, will be presented in the paper. Next the process by which the SCOOT data is used to validate parameters of traffic simulation models (congestion states, flows, origindestinations, etc) will be shown. The microsimulation model is used to predict tailpipe emission, taking into account the second by second drive cycles and the canyon model OSPM is then used to predict pollutant concentrations in the canyon at positions that coincide with the motes. Using the SCOOT data for flow and a speed estimate based on delay and independent estimate of emissions and concentrations is produced and compared with that derived by AiMSUn and measured by the motes. In this way an in-depth understanding of the spatial and temporal changes in the congestion and associated carbon monoxide across an area due to recurrent congestion occurring at the shoulders of a football event was made possible. The limitations of the both the AIMSUN and SCOOT derived pollutant estimators were explained. The results of this work showed that both AIMSUN and the SCOOT based estimates emissions estimates along with the Ospm model and measured meteorological conditions can provide estimates of roadside concentrations in a Canyon with a level of statistical confidence reflected by the regression coefficient R2=0.70. Given that this is a first attempt at developing a real-time Canyon model and there is scope to address the shortfalls identified in this work the results presented in this paper show much promise. The paper has highlighted the benefits of pervasive sensors and how they can compliment legacy systems through their flexibility in covering detection gaps in existing urban networks.

Publication metadata

Author(s): Bell MC, Galatioto F, Hill G

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Unpublished

Conference Name: 42nd Annual UTSG Conference

Year of Conference: 2010

Date deposited: 27/02/2010