Road network recovery from concurrent capacity-reducing incidents : model development and optimisation

Local and regional economies are highly dependent on the road network. The concurrent closure of multiple sections of the network following a hazardous event is likely to have significant negative consequences for those using the network. In situations such as these, infrastructure managers must decide how best to restore the network to protect users, maximise connectivity and minimise overall disruption. Furthermore, many hazardous events are forecast to become more frequent and extreme in the future as a result of climate change. Extensive research has been undertaken to understand how to improve the resilience of degraded transport networks. Whilst network robustness (that is, the ability of a network to withstand stress) has been considered in numerous studies, the recovery of the network has captured less attention among researchers. Methodologies developed to date are overly simplistic, especially when simulating the dynamics of traffic demand and drivers' decision-making in multi-day situations where there is considerable interplay between actual and perceived network states and behaviour. This thesis presents a decision-support tool that optimises the recovery of road transport networks after major day-to-day disruptions, maximising network connectivity and minimising total travel costs. This work expands upon previous efforts by introducing a new approach that models the damage-capacity-time relationship and improves the existing reinforcement-learning traffic-assignment models to be applicable to disrupted scenarios. An efficient metaheuristic approach (NSGA-II) is proposed to find optimal solutions for the recovery problem. The model is also applied to a real-world scenario based on the Scottish road network. Results from this case study clearly highlight the potential applicability of this model to evaluate different recovery strategies and optimise the recovery of road networks after multi-day major disruptions.