Traffic assignment, the process by which vehicle flows are loaded on to paths traversing a road network for the purpose of spatial demand forecasting, has been traditionally approached as a mathematical optimisation problem. However, this assumes typical highway network conditions, yielding ‘average day’ traffic forecasts only. Such approaches fail to account for time-dependent variability caused by infrequent events such as traffic accidents, vehicle breakdowns or road works which result in sub-optimal network performance. On any day, especially when incidents cause abnormal congestion patterns, drivers can only choose routes and diversion strategies according to the best of their own subjective knowledge and experience which is unique to each traveller. Ensuring that knowledge, both within-day and between days, is represented adequately and with realistic assumptions within models is key to forecasting traffic flows in all situations and their resulting network phenomena accurately. This thesis explores how drivers react under these irregular conditions, termed ‘states’, with a goal of understanding route choice and consequently advancing demand forecasting techniques. To this end, a simulator based survey is used in order to gain further knowledge of driver learning and diversion behaviour, then an agent based simulation modelling approach is developed using this insight to explore the network and traffic flow effects of drivers reacting to uncertain network conditions by altering their route choices. This work argues that representing driver knowledge and choices from a disaggregate agent based perspective, rather than a traditional aggregate approach, is more appropriate for modelling the impact of variable travel conditions. Results demonstrate that the possibility of incidents occurring and the potential for diverting can have a significant effect on network characteristics and the decisions of drivers, even on incident-free ‘clear’ days. Importantly, results show that drivers diverting can temporarily alleviate congestion but ultimately cause more delays and suboptimal network performance. These results have significant implications for demand forecasting practitioners and policy makers who try to minimise disruption through traffic management systems or effective network design.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:680735 |
| Date | January 2015 |
| Creators | Snowdon, James |
| Contributors | Waterson, Benedict |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/388114/ |
Page generated in 0.0102 seconds