10:00 AM - 10:15 AM
[HTT15-05] A Macroscopic Method for Road Network Dataset Comparison Using Geographical Traffic Distribution
★Invited Papers
Keywords:road network dataset, network simplification, transportation network analysis, Wasserstein distance
Road network data plays an important role in transportation network analysis as the foundation where all travelers move and traffic flows. Road network data is a type of geographical information system (GIS) data which organizes real-world roads into a network structure with their connections and shapes. Nowadays, an increased accessibility to various types and formats of road network data, from open databases such as OpenStreetMap (OSM) to which generated from Global Navigation Satellite System (GNSS) trajectory data (Zhong et al. 2023) and satellite imageries (Hinz and Baumgartner 2003), has dramatically improved the efficiency of research on road networks.
The comparison of different road network datasets is essential for understanding the quality of data and evaluating their accuracy and usefulness. Previous studies mainly focused on the characteristics of network, such as positional accuracy, completeness and topological correctness (Boeing 2024, Hakley 2010). However, these methods only focus on a specific feature of network and may not be appropriate to directly apply to transportation network analysis, in which various local and global features are essential. For example, simplified road networks may not be accurate in completeness (Fig. 1), but tends to computationally efficient and are expected to keep the major traffic states (Fig. 2). Therefore, existing methods could only evaluate the accuracy but not traffic and computational issues, thus the usefulness of data could not be evaluated properly.
In this study, we developed a comprehensive method of comparing road network datasets that can capture the effects of local and global features for transportation analysis. To evaluate road network data on a macroscopic scale, we consider the entire network represented by the data as one geographical distribution weighted by traffic parameters. The distribution is on a two-dimensional grid space, refers to the shape of road network using positional information of links and nodes. Possible traffic states on road network data is obtained by static traffic assignments with user equilibrium. Necessary traffic parameters (e.g. capacity and max speed) and OD demands can be hypothetically assumed from road network data. Differences between two road networks is measured by differences of possible route choices on them, in proportion to traffic volume and geographical distance between routes. This is defined as traffic-weighted distance, calculated from the Wasserstein distance between two corresponding distributions.
We did comparisons and evaluations of different simplified OSM networks within a study area in Tokyo. Network extraction methods are applied for simplification, where nodes and links which considered as insignificant are removed to simplify the network but keep the major traffic distribution on it. Results in Fig. 3 show that for such simplified networks, they have high link length reduce rate but small traffic-weighted distance to the original network, refers to small changes on traffic distributions. Therefore, the proposed method could evaluate changes of possible traffic states over different networks, suggests a potential use in evaluating and selecting road network datasets.
References
G. Boeing: Modeling and Analyzing Urban Networks and Amenities with OSMnx. Working paper. 2024.
M. Haklay: How good is volunteered geographical information? a comparative study of openstreetmap and ordnance survey datasets,Environment and Planning B: Planning and Design, Vol.37, No.4, pp.682–703, 2010.
S. Hinz and A. Baumgartner: Automatic extraction of urban road networks from multi-view aerial imagery, ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 58, Issues 1–2, pp. 83-98, 2003.
H. Zhong, T. Seo, W. Nakanishi, S. Yasuda, Y. Asakura and T. Iryo: Generation of aggregated road network by vehicle trajectory data, in 10th International Symposium on Transportation Data and Modelling (ISTDM2023), Ispra, 19-22 June 2023.
The comparison of different road network datasets is essential for understanding the quality of data and evaluating their accuracy and usefulness. Previous studies mainly focused on the characteristics of network, such as positional accuracy, completeness and topological correctness (Boeing 2024, Hakley 2010). However, these methods only focus on a specific feature of network and may not be appropriate to directly apply to transportation network analysis, in which various local and global features are essential. For example, simplified road networks may not be accurate in completeness (Fig. 1), but tends to computationally efficient and are expected to keep the major traffic states (Fig. 2). Therefore, existing methods could only evaluate the accuracy but not traffic and computational issues, thus the usefulness of data could not be evaluated properly.
In this study, we developed a comprehensive method of comparing road network datasets that can capture the effects of local and global features for transportation analysis. To evaluate road network data on a macroscopic scale, we consider the entire network represented by the data as one geographical distribution weighted by traffic parameters. The distribution is on a two-dimensional grid space, refers to the shape of road network using positional information of links and nodes. Possible traffic states on road network data is obtained by static traffic assignments with user equilibrium. Necessary traffic parameters (e.g. capacity and max speed) and OD demands can be hypothetically assumed from road network data. Differences between two road networks is measured by differences of possible route choices on them, in proportion to traffic volume and geographical distance between routes. This is defined as traffic-weighted distance, calculated from the Wasserstein distance between two corresponding distributions.
We did comparisons and evaluations of different simplified OSM networks within a study area in Tokyo. Network extraction methods are applied for simplification, where nodes and links which considered as insignificant are removed to simplify the network but keep the major traffic distribution on it. Results in Fig. 3 show that for such simplified networks, they have high link length reduce rate but small traffic-weighted distance to the original network, refers to small changes on traffic distributions. Therefore, the proposed method could evaluate changes of possible traffic states over different networks, suggests a potential use in evaluating and selecting road network datasets.
References
G. Boeing: Modeling and Analyzing Urban Networks and Amenities with OSMnx. Working paper. 2024.
M. Haklay: How good is volunteered geographical information? a comparative study of openstreetmap and ordnance survey datasets,Environment and Planning B: Planning and Design, Vol.37, No.4, pp.682–703, 2010.
S. Hinz and A. Baumgartner: Automatic extraction of urban road networks from multi-view aerial imagery, ISPRS Journal of Photogrammetry and Remote Sensing, Vol. 58, Issues 1–2, pp. 83-98, 2003.
H. Zhong, T. Seo, W. Nakanishi, S. Yasuda, Y. Asakura and T. Iryo: Generation of aggregated road network by vehicle trajectory data, in 10th International Symposium on Transportation Data and Modelling (ISTDM2023), Ispra, 19-22 June 2023.