Conceptualising the multifaceted nature of urban road congestion

Abberley, Luke Francis James (2023) Conceptualising the multifaceted nature of urban road congestion. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (10MB) | Preview

Abstract

Urban road congestion is not a new phenomenon and remains an outstanding problem that continues to impact people around the world. Road congestion costs the European Union an estimated 1-2% of GDP each year and is responsible for 27% of deadly C02 emissions. In addition, it can cause life-threatening delays in the emergency services response time. Road congestion has a multifaceted nature and lacks a clear and explicit definition. This makes the problem of tackling it very subjective, time and context dependent. There have been several approaches to both modelling and predicting road congestion. From a physical perspective, road congestion has been modelled using speed, capacity, velocity, and journey time; relatively road congestion has been classified using terms such as non-recurrent and recurrent congestion which tend to be relative to each stakeholder; conceptual models such as the bathtub, traffic flow, and origin to the destination have been used to ascertain the impact of road congestion on a city scale. This research presented tackles the problem of defining what is meant by congestion within an urban road network through defining a conceptual model that captures the semantics of road traffic congestion and its causes. The model is validated through the construction of a real-world dataset and the development of a visual tool which can be used to identify and alleviate congestion. The final stage of the project uses both the model and the dataset to investigate and implement a series of fuzzy systems to classify three types of congestion (non-recurrent, recurrent, and semi-recurrent). The fuzzy system results are then validated against human methods of classifying congestion. The main contributions of this thesis to world knowledge can be summarised as follows: The design and development of a novel universal Urban Road Congestion Conceptual (URCC) model. The URCC model is broken down into two main components: Analogical conceptualisation which builds upon the famous ‘bathtub’ model and will integrate with other analogies to create ‘a raindrop hitting a leaf inside the bathtub with ever changing water temperatures’. The second component is an ontological approach to modelling congestion thus providing a better understanding for decision-makers through providing a formal and explicit explanation for concepts within the domain of urban road congestion. Another contribution is the development of a real-world spatiotemporal quasi-real-time big data dataset known as the Manchester Urban Congestion Data (MUCD) dataset which was used to validate the URCC. A visualisation graphical user interface called TIM (Transport Incident Manager) was developed with stakeholders TfGM (Transport for Greater Manchester). TIM has the ability to fill the void left by the clear lack of visualisation tools that are capable of visualising real-world big data datasets, such as the MUCD and models of urban road congestion. The final contribution to knowledge is the design and development of two fuzzy decision-making systems which are not only capable of predicting urban road congestion on a link but the type of congestion occurring on a network of links. Using a fuzzy decision-making system allows for explainable and interpretable decisions, and also provided useful and meaningful qualitative context back to the relevant TfGM stakeholders. The non-optimised multi-classification fuzzy system had slightly worst accuracy than the J48 decision tree algorithm, however, the fuzzy system is easier to interpret and provides meaningful context compared to the J48 algorithm due to only requiring 12 rules compared to the 1184 learned rules in the J48 decision tree. Furthermore, once the fuzzy system has been optimised (future work) it is likely to have similar if not better performance than the J48 decision tree.