Abstract Manager

Planning and designing efficient road freight networks requires a comprehensive understanding of their spatial characteristics and demand patterns. The complexity of road networks presents significant challenges for effectively representing and analyzing these patterns while managing computational efficiency. Additionally, public agencies responsible for these networks face challenges in terms of traffic monitoring, safety analyses, and asset management at a broader level. These challenges are compounded by the existence of multiple versions of Linear Referencing Systems, e.g., provincially owned systems, national-level datasets available for public use, and data readily available through third parties. Focusing on the Prairie region of Canada, this study proposes a methodology to abstract the network using a multi-model traffic assignment approach, aiming to provide a unified core network driven by the user data, while preserving key network structures.
        

Using data from the Canadian Freight Analysis Framework (CFAF), which provides estimates of Canadian freight flows, we used origin-destination matrices and applied them to the study network using various traffic assignment models within the professional transport modelling software, PTV Visum. To achieve network abstraction, we used three distinct link removal policies based on the sets of link volumes, resulting in three candidate abstract networks. Each candidate network is then evaluated against the original network in terms of criteria-based performance measures, such as link volumes, betweenness centrality, and overall network connectivity. The selection of the most suitable abstract network is guided by ensuring optimal balance between these criteria. Following the selection process, we apply an algorithm to merge roadway segments between intersections, further streamlining the network representation. This additional step enhances the manageability and interpretability of the network while preserving critical connectivity and traffic flow characteristics.
        

Results demonstrate that the proposed abstraction approach significantly reduces network complexity without compromising structural accuracy, providing a robust foundation for practical applications, such as unified asset management strategies, investment prioritization, and regulatory adjustments. While this study focuses on freight networks in the Prairie region of Canada, the methodology is flexible and can be extended to other transportation contexts, including passenger vehicle networks or future integration with Artificial Intelligence (AI) models, to address a wider range of planning and management needs.