Heimdall Roadcast: Road Traffic Observation and Forecasting

Description

An alternative method of forecasting traffic conditions explored in a PHD research project with Southampton University, sponsored by Siemens. Could Congestion be predicted using a machine learning approach that fused data from traffic detectors and selected contextual sources? This research programme developed a machine learning algorithm called Roadcast which has been found to be more accurate that historical average predictors by being able to learn the effects of different contexts (events, roadwork's etc.) on traffic behaviour. The contextual data sources used in the Roadcast research project were collected manually, so an important step towards commercial upscaling is the automation contextual data collection. The feasibility project will propose a method for automatically delivering contextual data for use in improving traffic condition forecasting.