A computer-implemented method and apparatus for providing a condensed representation of a road map

Abstract

The invention provides a computer-implemented method and apparatus for deriving a condensed representation of a road map which can be used for assessing driving data for an vehicle comprising the steps of: extracting (S1) road map characteristics from a data source containing map data of an original road map (ORM); condensing (S2) the original road map (ORM) based on the extracted road map characteristics (RMC) to provide a condensed road map (CRM); and evaluating (S3) a similarity between the original road map (ORM) and the condensed road map (CRM) to determine whether the condensed road map (CRM) represents the original road map (ORM) sufficiently in terms of an applied similarity metric (SM).

Description

[0001] 202412498

[0002] 1

[0003] Description

[0004] A computer-implemented method and apparatus for providing a condensed representation of a road map

[0005] The invention relates to a computer-implemented method and apparatus for deriving a condensed representation of a road map which can be used for assessing a driving risk and / or a performance of a vehicle including an autonomous vehicle.

[0006] The deployment of autonomous vehicles introduces new types of risk. Different types of risk may be of interest to different types of stakeholders. For instance, city developers and regulators may be interested in the safety impact of an autonomous vehicle while insurance companies focus more on financial impact. Commonly risk can be assessed by frequency and severity of certain events such as traffic accidents. However, for new endeavors, such as the deployment - a combination of operating conditions and location - of autonomous vehicles on public roads, there is not enough historical data that can be used to estimate its risk. Especially for larger deployment areas this constitutes a big challenge since - both in real world as well as in simulation environments - the time required to collect driving data of an entire area takes a very long time and is very resource-intensive. Conventional approaches to assess risks caused by deployment of autonomous vehicles involve simulating all roads in an area, simulating individual routes and / or edge cases real-world driving of autonomous vehicles through all roads in an area, real-world driving of autonomous vehicles through a subset of roads in an area and / or real-world driving of autonomous vehicles on test tracks.

[0007] Against this background, a problem addressed by the present invention is that of providing an efficient method for condensing map data to facilitate collecting driving data of vehicles which can be used for risk assessment.

[0008] This problem is solved according to the invention by a computer-implemented method for providing a condensed representation of a road map having the features of claim 1 and by an server apparatus having the features of claim 17 and a system having the features of claim 18.

[0009] The invention provides according to a first aspect a computer-implemented method for providing a condensed representation of a road map comprising the steps of: extracting road map characteristics from a data source containing map data of an original road map; 202412498

[0010] 2 condensing the original road map based on the extracted road map characteristics and their relative frequency to provide a condensed road map; and evaluating a similarity between the original road map and the condensed road map to determine whether the condensed road map represents the original road map sufficiently in terms of at least one applied similarity metric.

[0011] In a preferred embodiment also a relative frequency or distribution of road map characteristics is determined within the original road map and then used for condensing the original road map.

[0012] The condensation of the original road map can be performed in a possible embodiment iteratively.

[0013] The applied similarity metric can comprise a driving risk related similarity metric.

[0014] The computer-implemented method according to the first aspect of the present invention is adapted to condense a driving area and / or a pre-defined route into a smaller but representative version of it. While the method according to the present invention does focus on representative risk, other similarity metrics such travel times can also be used to define similarity.

[0015] Extracting road map characteristics from a stored road map involves analyzing and processing the map data to identify various features and details relevant to e.g. risk, transportation or navigation.

[0016] Road maps can be stored in various formats, such as Geographic Information System (GIS) files, OpenStreetMap (OSM) data, satellite imagery, or proprietary map data.

[0017] Key characteristics of a road map that can be extracted can include:

[0018] Road Network: Identifying the roads, their classifications (highways, streets, alleys, etc.), and connectivity.

[0019] Intersections and Junctions: Locating points where roads intersect, including complex junctions like roundabouts and interchanges.

[0020] Attributes of Roads: Extracting data on road attributes such as speed limits, lane counts, one- way / two-way status, road width, and surface type.

[0021] Geometric Properties: Determining the precise geometry of roads, including their lengths, curves, and elevations. 202412498

[0022] 3

[0023] Attributes of the environment: pedestrian infrastructure, road furniture, features surrounding the network such as bus stops, park and subway entrances, pedestrian areas, information about zones the road is embedded in (commercial, residential, industrial, ...) among others.

[0024] When comparing an original road map ORM with a condensed version, the goal is to measure how well the condensed map preserves the essential features and characteristics as well as their relative frequency within the original road map ORM. The relative frequency is the ratio of a characteristic to the overall amount of characteristics in the original road map ORM. Various similarity metrics can be applied to assess this.

[0025] The invention provides according to a further aspect a server apparatus comprising a data extraction module adapted to extract road map characteristics from at least one data source containing map data of the road map to be condensed; a map condensing module adapted to condense the original road map based on the extracted road map characteristics and their relative frequency to provide a condensed road map; and a similarity evaluation module adapted to evaluate a similarity between the original road map and the condensed road map to determine whether the condensed road map represents the original road map sufficiently in terms of an applied similarity metric.

[0026] The invention provides according to a further aspect a system comprising data access layer including at least one data source containing map data of an original road map; a server apparatus having a data extraction module adapted to extract road map characteristics of the original road map stored in the at least one data source of the data access layer, a map condensing module adapted to condense the original road map based on the extracted road map characteristics to provide a condensed road map, and a similarity evaluation module adapted to evaluate a similarity between the original road map and the condensed road map to determine whether the condensed road map represents the original road map sufficiently in terms of an applied similarity metric; and a client apparatus adapted to run a client application based on the condensed road map.

[0027] Advantageous configurations and developments emerge from the further dependent claims and from the description with reference to the figures of the drawings.

[0028] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the road map characteristics used for condensing the original road map comprise road network characteristics and / or route characteristics. 202412498

[0029] 4

[0030] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the road map characteristics used for condensing the original road map comprise road characteristics of roads within a road network including road types, road surfaces, road lane numbers, road widths, road lane widths, number of roundabouts, radius of roundabouts, number of signalized and / or un-signalized intersections, number of regulated and / or unregulated pedestrian crossings, road speed limits, road turn radius and road turn angles.

[0031] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map an artificial road network is generated as the condensed road map by a generative artificial intelligence module based on the road map data of the original road map and based on the extracted road map characteristics.

[0032] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map a sub-network representative for the original road network of the road map is identified based on road network characteristics.

[0033] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the sub-network representative for the original road network of the road map is identified based on administrative divisions of the original road map including area states, urban and / or rural districts and / or municipalities.

[0034] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the sub-network representative for the original road network of the road map is identified based on a specific shape.

[0035] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map a sub-route representative for the original route of the original road map is identified by performing of a standard graph search.

[0036] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map condensing the original road map based on the extracted road map characteristics is performed iteratively until the evaluated similarity between the original road map and the condensed road map is sufficient in terms of the applied similarity metric according to predefined acceptance criteria. 202412498

[0037] 5

[0038] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map condensing the original road map based on the extracted road map characteristics is performed according to a selected condensing algorithm.

[0039] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the applied similarity metric is selected from a group of available predefined similarity metrics, preferably including at least one of a risk related similarity, a geometric similarity, a topological similarity, an attribute-based similarity, a path-based similarity, coverage and completeness, a semantic similarity, a visual similarity, and / or a functional similarity.

[0040] This allows for flexibility and a wide range of possible use cases.

[0041] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the applied similarity metric comprises a risk-related similarity metric.

[0042] The use of a risk-related similarity metric is helpful for deployment of autonomous vehicles.

[0043] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the representativeness of the condensed road map is validated using historical real world data or simulated data.

[0044] The validation results can be used to improve the accuracy, robustness and reliability of the applied condensing algorithm and / or the applied similarity metrics.

[0045] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the map data of the original road map comprises a machine readable map data having a machine readable data format.

[0046] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map in the map data of the original road map comprises not machine readable map data converted automatically into a machine readable data format. 202412498

[0047] 6

[0048] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the condensed road map is stored in a data memory for further processing and / or for being output via an interface.

[0049] The interface can comprise a user interface and / or a control and data interface of a control system. The control system may form part of a vehicle and may comprise client devices adapted to run applications on the basis of the condensed road map in near real time.

[0050] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the condensed road map is used for collecting more efficiently driving data for an autonomous vehicle.

[0051] The invention further provides a computer program comprising instructions adapted to perform the method according to the first aspect of the present invention.

[0052] Where appropriate, the above-mentioned configurations and developments can be combined with each other as desired, as far as this is reasonable. Further possible configurations, developments and implementations of the invention also include combinations, which are not explicitly mentioned, of features of the invention which have been described previously or are described in the following with reference to the embodiments. In particular, in this case, a person skilled in the art will also add individual aspects as improvements or supplements to the basic form of the present invention.

[0053] In the following different aspects of the present invention are described in more detail with reference to the enclosed figures.

[0054] Fig.1 shows a flow chart of an exemplary embodiment of a computer-implemented method for providing a condensed representation of a road map according to a first aspect of the present invention;

[0055] Fig.2 shows an example of a road network and its characteristics to illustrate the operation of a computer-implemented method according to the first aspect of the present invention;

[0056] Fig.3 shows an example of a map and its representative counterpart to illustrate the operation of a computer-implemented method according to the first aspect of the present invention; 202412498

[0057] 7

[0058] Fig.4 shows an example of a map and a representative route through the map to illustrate a condensation approach used by a possible embodiment of a computer-implemented method according to the first aspect of the present invention;

[0059] Fig.5 shows an example of a map and a selection that is representative of the original map to illustrate another condensation approach used by a further possible embodiment of a computer-implemented method according to the first aspect of the present invention;

[0060] Fig.6 shows an exemplary component diagram of a possible embodiment of a system according to a further aspect of the present invention;

[0061] Fig. 7 shows an exemplary process flow of a possible implementation of a computer- implemented method according to the first aspect of the present invention.

[0062] The invention provides according to a first aspect a computer-implemented method for providing a condensed representation of a road map. The method can comprise in a possible embodiment several main steps as illustrated in the flow chart of Fig. 1 .

[0063] In a first step S1 of the computer-implemented method illustrated in in the flow chart of Fig.1 road map characteristics RMCs from a data source containing map data of an original road map ORM are extracted. Fig. 2 illustrates schematically an example of an original road map ORM.

[0064] In a possible embodiment road map characteristics RMCs are not only identified but also their distribution or relative frequency among the road network of the original road map ORM is determined (e.g. 80% of the intersections are roundabouts).

[0065] The road map characteristics RMCs can be extracted in step S1 from an arbitrary data source containing map information, for example map data from OpenStreetMap, OpenDRIVE or even satellite imagery.

[0066] The map data of the original road map ORM can comprise machine readable map data having a machine readable data format. The map data of the original road map ORM can also comprise not machine readable map data which can be converted automatically into a machine readable data format for further processing. 202412498

[0067] 8

[0068] In a further step S2 of the computer-implemented method illustrated in in the flow chart of Fig.1 the original road map ORM is condensed based on the extracted road map characteristics to provide a condensed road map CRM. This can also take into account the determined relative frequency of road map characteristics RMCs within the original road map ORM.

[0069] In a possible embodiment the road map characteristics used for condensing the original road map ORM in step S2 can comprise road network characteristics and / or route characteristics.

[0070] As illustrated schematically in Fig. 2 the road map characteristics RMC used for condensing the original road map ORM in step S2 can comprise different kinds of road characteristics of roads within a road network represented by the original road map ORM. These road map characteristics RMCs can include road types, road surfaces, road lane numbers, road widths, road lane widths, number of roundabouts, radius of roundabouts, number of signalized and / or un-signalized intersections, number of regulated and / or unregulated pedestrian crossings, road speed limits, road turn radius, and / or road turn angles. These are only some examples for road map characteristics RMCs. Other characteristics can be used as well. The characteristics can also include context around the road network such as schools, bus stops, bike paths etc.

[0071] In a further step S3 of the computer-implemented method illustrated in in the flow chart of Fig.1 a similarity between the original road map ORM and the condensed road map CRM is evaluated to determine whether the condensed road map CRM represents the original road map ORM sufficiently in terms of an applied similarity metric SM.

[0072] A similarity of an original and condensed network or route can be evaluated in step S3 using a similarity measure or similarity metric SM in order to determine a best fitting condensed road map or route. In a possible embodiment of the computer-implemented method the applied similarity metric SM is selected from a group of available predefined similarity metrics, including at least one of a risk related similarity, a geometric similarity, a topological similarity, an attribute- based similarity, a path-based similarity, coverage and completeness, a semantic similarity, a visual similarity, and / or a functional similarity. Preferably, the applied similarity metric SM can comprise in a possible embodiment at least a risk-related metric.

[0073] When comparing an original road map with a condensed version, the goal is to measure how well the condensed map preserves the essential features and characteristics of the original map. Various similarity metrics can be applied to assess this. Different types of similarity metrics SMs that can be used: 202412498

[0074] 9

[0075] Risk related Similarity

[0076] It measures the similarity of traffic accident risks or other traffic related risks for the road network represented in the original road map ORM and represented in the condensed road map CRM.

[0077] Geometric Similarity

[0078] Hausdorff Distance measures the maximum distance from a point in one set to the closest point in the other set. This is useful for comparing the overall shape and geometry of the road networks.

[0079] Frechet Distance Measures the similarity between curves, accounting for the location and ordering of the points along the curves. This is suitable for comparing the paths and road shapes in the maps.

[0080] Topological Similarity

[0081] Graph Isomorphism checks if there is a one-to-one correspondence between the nodes and edges of the two graphs (road networks) and ensures that the connectivity and structure of the road networks are preserved.

[0082] Subgraph Matching identifies common substructures within the graphs. This is useful when the condensed map retains only a subset of the original map’s roads.

[0083] Attribute-Based Similarity

[0084] Road Attributes such as road lengths, widths, speed limits, and number of lanes and ensures that the condensed map accurately reflects the key characteristics of the roads in the original map.

[0085] Statistical Measures can be used to compare distributions of attributes like road types, traffic volumes, and speed limits. Metrics include mean, median, standard deviation, and histograms of the attributes.

[0086] Path-Based Similarity 202412498

[0087] 10

[0088] Route Similarity can be used to compare specific routes or paths between the original and condensed maps. It can use metrics like edit distance (number of changes needed to transform one path into another) or dynamic time warping (DTW) to compare paths of different lengths.

[0089] Shortest Path Preservation assesses whether the shortest paths between key points are preserved in the condensed map. It can use algorithms like Dijkstra’s or A* to find and compare shortest paths.

[0090] Coverage and Completeness

[0091] Intersection Over Union (loU) measures the overlap between the road networks in the original and condensed maps. It can be calculated as the ratio of the intersection area to the union area of the two maps.

[0092] Coverage Ratio can be used to compare the length or area of roads covered by the condensed map relative to the original map. It ensures that a significant portion of the original map’s roads are represented.

[0093] Semantic Similarity

[0094] POI Matching can be used to compare the locations and types of Points of Interest (POIs) in both maps. It ensures that important landmarks and features are retained in the condensed map.

[0095] Land Use and Zoning Similarity can be used to compares the classification of areas (residential, commercial, industrial) between the maps. It ensures that the condensed map reflects the original map’s land use patterns.

[0096] Visual Similarity

[0097] A Structural Similarity Index (SSIM) measures visual similarity between two images. It can be applied if the maps are represented as raster images.

[0098] Pixel-Based Comparison can be used to compare pixel values in rasterized versions of the maps. It uses metrics like Mean Squared Error (MSE) or Peak Signal-to-Noise Ratio (PSNR). 202412498

[0099] 11

[0100] Functional Similarity

[0101] Traffic Flow Comparison can be used to compare simulated or real traffic flows in both maps. It may use similarity metrics like traffic volume, speed, and congestion levels.

[0102] An accessibility analysis can be performed to compare the accessibility of key locations (e.g., hospitals, schools) in both maps. It ensures that the condensed road map CRM maintains similar accessibility patterns.

[0103] By using a combination of these similarity metrics SMs, a comprehensive assessment of the similarity between the original road map ORM and the condensed road map CRM can be achieved, ensuring that the condensed version retains the most critical features and characteristics of the original road map ORM.

[0104] In a possible embodiment a risk related similarity metric can be combined with other similarity metrics to define a similarity metric applied by the method according to the present invention depending on the use case. The definition and combination of similarity metrics can be performed in a possible embodiment by a user using an editor tool for similarity metric definition and adjustment.

[0105] Optionally, the representativeness of the condensed map or route can be validated in step S4 of the computer-implemented method illustrated in in the flow chart of Fig.1 in case appropriate data is available. A validation in terms of risk can for example, be conducted in step S4 by using historical incident data or through simulation.

[0106] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the representativeness of the condensed road map CRM is validated in step S4 using historical real world data. In an alternative embodiment of the computer-implemented method for providing a condensed representation of a road map the representativeness of the condensed road map CRM is validated in step S4 using simulated data.

[0107] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map an artificial road network is generated in step S2 as the condensed road map CRM by a generative artificial intelligence module based on the road map 202412498

[0108] 12 data of the original road map ORM and based on the extracted road map characteristics RMCs. Generative Al refers to algorithms and models that can produce new content or data based on patterns and information learned from existing data.

[0109] In a possible further embodiment of the computer-implemented method for providing a condensed representation of a road map a sub-network representative for the original road network of the road map is identified in step S2 based on road network characteristics.

[0110] In a possible further embodiment of the computer-implemented method for providing a condensed representation of a road map the sub-network representative for the original road network ORM of the road map is identified in step S2 based on administrative divisions of the original road map ORM including area states, urban and / or rural districts and / or municipalities.

[0111] In a still further possible embodiment of the computer-implemented method for providing a condensed representation of a road map the sub-network representative for the original road network of the road map is identified in Step S2 based on a specific shape.

[0112] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map a sub-route representative for the original route of the original road map ORM is identified in step S2 by performing of a standard graph search.

[0113] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map condensing the original road map ORM based on the extracted road map characteristics is performed in step S2 iteratively until the evaluated similarity between the original road map ORM and the condensed road map CRM is sufficient in terms of the applied similarity metric SM according to predefined acceptance criteria.

[0114] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map condensing the original road map ORM based on the extracted road map characteristics RMCs is performed according to a selected condensing algorithm.

[0115] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the condensed road map CRM is stored in a data memory for further processing and / or for being output via an interface. The data memory can be provided in a data access layer 2 as illustrated in Fig.6. 202412498

[0116] 13

[0117] The condensed road map CRM requires less memory than the original road map ORM. Further the condensed road map CRM can be processed by an application APP faster and more efficiently.

[0118] In a possible embodiment of the computer-implemented method for providing a condensed representation of a road map the condensed road map CRM is used for generating driving data for an (autonomous) vehicle.

[0119] The invention further provides a computer program comprising instructions adapted to perform the method according to the first aspect of the present invention.

[0120] The invention provides, according to a further aspect, a server apparatus as illustrated in the schematic diagram of Fig.6 forming part of a system 1. The server apparatus 3 of the system 1 shown in Fig. 6 comprises a map processing module 3A, a data extraction module 3B adapted to extract road map characteristics from at least one data source containing map data of an original road map ORM, a map condensing module 3C adapted to condense the original road map ORM based on the extracted road map characteristics RMCs to provide a condensed road map CRM and a similarity evaluation module 3D adapted to evaluate a similarity between the original road map ORM and the condensed road map CRM to determine whether the condensed road map CRM represents the original road map ORM sufficiently in terms of an applied similarity metric SM.

[0121] The invention provides according to a further aspect a system 1 as illustrated in Fig. 6. The system 1 comprises a data access layer 2 including at least one data source containing map data of an original road map ORM. The system shown in Fig. 6 further comprises the server apparatus 3 having a data extraction module 3B adapted to extract road map characteristics of the original road map ORM stored in the at least one data source of the data access layer 2, a map condensing module 3C adapted to condense the original road map ORM based on the extracted road map characteristics RMCs to provide a condensed road map CRM, and a similarity evaluation module 3D adapted to evaluate a similarity between the original road map ORM and the condensed road map CRM to determine whether the condensed road map CRM represents the original road map ORM sufficiently in terms of an applied similarity metric SM.

[0122] The system 1 illustrated in Fig.6 further comprises a client apparatus 4 adapted to run a client application APP based on the condensed road map CRM. 202412498

[0123] 14

[0124] The computer-implemented method and apparatus according to present invention is applicable to real-world as well as fictional map or route data.

[0125] Fig. 2 provides an example of a road network and its characteristics. Based on the map data and the extracted road network characteristics, two types of condensation approaches can be applied in step S2. Fig 2 illustrates schematically an original road map ORM before condensation. Thus, the original road map ORM is a representation of the original road network and its characteristics.

[0126] A first type of condensation approaches applied in step S2 takes the entire map I route data as well as the road network characteristics as input and generates a completely new, artificial road network that is a smaller but representative version of the original network of the original road map ORM, e.g. by using generative Al.

[0127] Such an approach can for example identify archetypes within the road traffic network and generate a new condensed representation through concatenating the learned archetypes such that the sequence of archetypes exhibits the same risk profile as the original map or route. This type of approach is applicable for real world as well as simulated environments.

[0128] Fig.3 gives an example of an original road map ORM (Fig.3 left side) and its representative condensed counterpart CRM (Fig.3 right side).

[0129] The second type of condensation approaches applied in step S2 of the method also takes the original road network or route characteristics as input and then identifies a sub-route or identifies a sub-network that is representative of the entire network or original route.

[0130] Examples of sub-networks representative of the original network can be administrative divisions of the original road map ORM (e.g. area states, urban / rural districts or municipalities or any combination thereof) or can be defined through an arbitrary shape (e.g. the location of the center of a circle and its radius, the envelope around one or more series of roads and intersections, i.e. sets of routes).

[0131] Selecting a representative route R for a larger route or area can be an arbitrary or a systematic search, e.g. standard graph search techniques like Depth-First Search (DFS) and Breadth-First Search (BFS). Areas can for example also be identified through leveraging spatial search algorithms, such as arranging the data in a spatial tree and traversing it using a priority queue. 202412498

[0132] 15

[0133] Defining a grid over the entire map and brute force traversing it. Such areas and shapes can be connected or disjoint. This type of approach is applicable in both real-world as well as simulated environments.

[0134] Fig.4 shows an example of an original road map ORM (Fig.4 left side) and a representative route R (Fig.4 right side) through the map according to a first condensation approach.

[0135] Fig.5 shows an example of an original road map ORM (Fig.5 left side) and a selected area A that is representative of the original road map ORM according to an alternative second condensation approach (Fig.5 right side).

[0136] To evaluate in step S3 of the computer-implemented method whether the generated output is representative of the original road map ORM, similarity metrics SM can be used to analyze the similarity and ultimately the representativeness of the generated output.

[0137] For example, the similarity can be evaluated in step S3 of the computer-implemented method through computing the absolute difference per extracted road network characteristic RMCs, and optionally combining these into a single number (e.g. computing the Manhattan distance between the networks under study). Such combinations require making assumptions on the importance of the road network characteristics and their relationship, which can for example be obtained from expert studies available in the literature, through transforming the data into a new coordinate system where the dimensions are independent (e.g. through a principal component analysis), or through learning a parameterization of the relationship between road network characteristics and risk.

[0138] To validate in step S4 of the computer-implemented method the derived approach (e.g. whether the road map characteristics are selected appropriately) real world or simulated data can be used.

[0139] As the method according to the present invention does focus on representative risk, one option is to compare historical risk events (e.g. collisions) between the original road network or route and the condensed version. If the representative map is synthetic, traffic simulations can be used.

[0140] Fig.6 shows an exemplary component diagram of a possible embodiment of a system 1 according to the present invention. In the illustrated embodiment, there is a data access layer 2 202412498

[0141] 16 to access the data sources such as database 2A or a file storage 2B, a server apparatus 3 where the computations are done, and a client application APP run on a client apparatus 4 to start the execution of the map processing.

[0142] The required input data (e.g. map data, satellite imagery, ...) can be stored in a data storage, such as a database 2A or a file storage 2B of the data access layer 2 as illustrated in Fig.6. On the server side there can be provides in a possible embodiment different modules 3A to 3D handling different steps of the process performed by the method according to the present invention.

[0143] A map processing module 3A of the server 3 is adapted to process map data from an arbitrary data source. The map data can either directly be machine readable (e.g. map data from OpenStreetMap) or need to be converted into a machine-readable format (e.g. satellite image data).

[0144] A data extraction module 3B of the server 3 is adapted to extract and compute a defined set of road network characteristics from the processed map data. Furthermore, the data extraction module 3B may address data quality issues (e.g. incomplete data, inaccurate data, inconsistent data).

[0145] In a map condensation module 3C of the server 3 the extracted road characteristics are used to either identify a sub-route, a sub-network (e.g. generate a set of routes through the original road network or select a part of the original network) that is representative of the entire map or a completely new road layout that is smaller than, but also representative of, the entire original road map.

[0146] Lastly, a similarity evaluation module 3D of the server 3 ensures a previously defined quality threshold of the results through quantifying the similarity, and therefore representativeness e.g. in terms of risk, of the original map and the generated output. Any additional validation techniques (e.g. comparison to real-world accident data) may also be conducted in this module.

[0147] An exemplary process flow of a possible implementation of a computer-implemented method according to the first aspect of the present invention is illustrated in Fig.7.

[0148] The process begins in step SA with reading and processing the provided map data. A set of road network characteristics is specified in Step SB from a list of possible road network 202412498

[0149] 17 characteristics. The specified characteristics are then extracted and counted in step Sc from the map data.

[0150] In a next step SD, a condensation algorithm is selected. Once a condensation algorithm has been selected, an iterative approach is used. If it is found in step SE that no condensation algorithm has been chosen a generative representative road within the network can be generated in step SF. Otherwise a smaller but representative road network is generated in step SG. From a defined (set of) starting point(s) (e.g. intersection, road sections, road network characteristic, small area within the map) the route or area is enlarged within each iteration.

[0151] To assess the quality of the iteration’s result, a similarity check is conducted in step SH. TO achieve this, a similarity metric SM is selected in step Si from a set of similarity metrics.

[0152] After conducting the similarity analysis in step SG, the results can be compared in step SK to predefined acceptance criteria. If the acceptance criteria are met, the iterative process is terminated in step SL. If the acceptance criteria are not met, the selected condensation algorithm options can be changed iteratively in step SD to improve the generated output until the acceptance criteria are met (for one of the selected starting points) in step SK. If the similarity check in step SH is negative the process is terminated in step SM.

[0153] The computer-implemented method according to the present invention does significantly reduce effort and higher scalability compared to driving through I simulating all roads in a network.

[0154] A list of characteristics can be extended or can be reduced arbitrarily. This allows the application of the computer-implemented method in different domains (e.g. by specifically adding use-case specific characteristics such as for risk or traffic-flow etc.)

[0155] The computer-implemented method according to the present invention can use tunable similarity measures depending on the chosen condensation algorithm and characteristics. The computer-implemented method further enables the creation of higher fidelity models for a condensed network compared to larger network within the same budget. Condensation techniques applied by the computer-implemented method can evolve over time and produce better representative networks / adapt to requirements.

[0156] Although the present invention has been described in the above by way of embodiments, it is not limited thereto, but rather can be modified in a wide range of ways. In particular, the 202412498

[0157] 18 invention can be changed or modified in various ways without deviating from the core of the invention.

Claims

20241249819Claims1. A computer-implemented method for providing a condensed representation of a road map comprising the steps of:- extracting (S1) road map characteristics (RMC) from a data source containing map data of an original road map (ORM);- condensing (S2) the original road map (ORM) based on the extracted road map characteristics (RMC) and their relative frequency to provide a condensed road map (CRM); and- evaluating (S3) a similarity between the original road map (ORM) and the condensed road map (CRM) to determine whether the condensed road map (CRM) represents the original road map (ORM) sufficiently in terms of an applied similarity metric (SM).

2. The computer-implemented method for providing a condensed representation of a road map according to claim 1 , wherein the road map characteristics (RMC) used for condensing the original road map (ORM) comprise road network characteristics and / or route characteristics.

3. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 2, wherein the road map characteristics (RMC) used for condensing the original road map (ORM) comprise road characteristics of roads within a road network including road types, road surfaces, road lane numbers, road widths, road lane widths, number of roundabouts, radius of roundabouts, number of signalized and / or unsignalized intersections, number of regulated and / or unregulated pedestrian crossings, road speed limits, road turn radius, and / or road turn angles.

4. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 3, wherein an artificial road network is generated as the condensed road map (CRM) by a generative artificial intelligence module based on the extracted road map characteristics (RMC) and their distributions.

5. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 4, wherein a sub-network representative for the original road network of the original road map (ORM) is identified based on20241249820- road network characteristics and / or- administrative divisions of the original road map (ORM) including area states, urban and / or rural districts, and / or municipalities.

6. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 5, wherein the sub-network representative for the original road network of the original road map (ORM) is identified based on a specific shape.

7. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 6, wherein a sub-route representative for the original route of the original road map (ORM) is identified by performing of a standard graph search.

8. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 7, wherein condensing the original road map (ORM) based on the extracted road map characteristics (RMC) is performed iteratively until the evaluated similarity between the original road map (ORM) and the condensed road map (CRM) is sufficient in terms of the applied similarity metric according to predefined acceptance criteria.

9. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 8, wherein condensing the original road map (ORM) based on the extracted road map characteristics (RMC) is performed according to a selected condensation algorithm.

10. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 9, wherein the applied similarity metric (SM) is selected from a group of available predefined similarity metrics.

11. The computer-implemented method for providing a condensed representation of a driving network according to any of the preceding claims 1 to 10, wherein the applied similarity metric (SM) comprises a risk-related metric.

12. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 11 , wherein the representativeness of the condensed road map (CRM) is validated (S4) using historical real world data or simulated data.2024124982113. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 12, wherein the map data of the original road map (ORM) comprises- machine readable map data having a machine readable data format or- not machine readable map data converted automatically into a machine readable data format.

14. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 13, wherein the condensed road map (CRM) is stored in a data memory for further processing and / or for being output via an interface.

15. The computer-implemented method for providing a condensed representation of a road map according to any of the preceding claims 1 to 14, wherein the condensed road map (CRM) is used for assessing the driving risk for an autonomous vehicle.

16. A computer program comprising instructions adapted to perform the method according to any of the preceding claims 1 to 15.

17. A server apparatus (3) comprising- a map processing module (3A) adapted to process map data from a data source;- a data extraction module (3B) adapted to extract road map characteristics (RMC) from at least one data source containing map data of an original road map (ORM);- a map condensing module (3C) adapted to condense the original road map (ORM) based on the extracted road map characteristics (RMC) and their relative frequency to provide a condensed road map (CRM); and- a similarity evaluation module (3D) adapted to evaluate a similarity between the original road map (ORM) and the condensed road map (CRM) to determine whether the condensed road map (CRM) represents the original road map (ORM) sufficiently in terms of an applied similarity metric (SM).

18. A system (1) comprising20241249822- a data access layer (2) including at least one data source containing map data of an original road map (ORM);- a server apparatus (3) having a map processing module (3A) adapted to process map data from a data source;- a data extraction module (3B) adapted to extract road map characteristics (RMC) of the original road map (ORM) stored in the at least one data source of the data access layer (2);- a map condensing module (3C) adapted to condense the original road map (ORM) based on the extracted road map characteristics (RMC) and their relative frequency to provide a condensed road map (CRM); and- a similarity evaluation module (3D) adapted to evaluate a similarity between the original road map (ORM) and the condensed road map (CRM) to determine whether the condensed road map (CRM) represents the original road map (ORM) sufficiently in terms of an applied similarity metric (SM); and a client apparatus (4) adapted to run a client application (APP) based on the condensed road map (CRM).

Images