Method for determining an indicative value statistically characterizing a description of events acquired by a vehicle in relation to a location

Abstract

This invention relates to a method for determining an indicator value (Z) for statistically characterizing a location-related event description acquired by at least one vehicle, wherein, based on a road or navigation map, road segments with no inflow and no outflow are identified, with a segment length greater than or equal to a predetermined minimum segment length. Adjacent identified road segments are connected to form a road segment chain (10) having a start point (11) and an end point (12). An event description, including the vehicle's geographic location (P1, P2) and at least one parameter of the vehicle's environmental model, is acquired and transmitted to an analysis system. The analysis system assigns the nearest road network node (O1, O2) of the road segment chain (10) to the geographic location (P1, P2) of the event description accordingly, the nearest road network node being described by a longitudinal distance (L, L1, L2) relative to the start point (11) of the road segment chain (10). For at least one road segment chain (10), a curve of the density value (p) of the event description with respect to the longitudinal distance (L, L1, L2) to the starting point (11) is determined by means of a kernel density estimator with a predetermined bandwidth. The curve of the density value (p) is transformed into a Z curve (Z1, Z2) of the Z value (Z) with zero mean relative to the bandwidth by means of bandwidth-based local Z-normalization.

Description

Technical Field

[0001] The present invention relates to a method for determining an indication value / Z value according to the preamble of claim 1, the indication value being used for statistical characterization of a location-related event description acquired by at least one vehicle. Background Technology

[0002] Document DE 10 2022 105 919 A1 discloses a system and method for early identification of structural hazards in road traffic based on digital traffic route network maps or maps. For this purpose, a computer system assigns accident data, user input data, and sensor data to segments of the map according to their geographic reference (data). According to one aspect, the computer system performs an evaluation including at least frequency determination and / or data comparison. For the purpose of early identification, if the sensor data or user input data has been assigned at a predetermined frequency, and / or if the data comparison of the user input data and / or sensor data—particularly with a comparison with an artificial intelligence (AI) based model—indicates a feature match or feature correlation of a predetermined degree with a critical pattern, then no geographic reference segment associated with the accident data is identified as a potential hazard. This makes early hazard identification possible. According to another aspect, it is proposed to determine a hazard score for the identified hazard.

[0003] US 2023 / 0245560 A1 discloses a system for determining location-related hazard classification based on a database of events and / or accidents acquired by a vehicle. The system includes an interface and a processor. The interface is configured to receive accident data and / or event data. The processor is configured to determine event groups, wherein each event group is assigned a set of accidents from the accident database and / or a set of events from the event data, grouped according to their location proximity. The processor is also configured to determine a set of road traffic estimates associated with a set of events and to determine the order of the event groups based on the number of accidents, the number of events, and the number of road traffic estimates.

[0004] Document DE 10 2019 215 587 A1 discloses a method for calculating relative positive acceleration or negative acceleration (RPA, RNA) through crowdsourcing, comprising the following steps: a) generating a digital map in which different road segments are interconnected by nodes; b) calculating the relative positive acceleration or negative acceleration (RPA, RNA) of vehicles traveling in the road segments with respect to the road segment (A, B, C, ..., M), wherein the relative positive acceleration or negative acceleration (RPA, RNA) of the vehicles is calculated and sent to a backend, or wherein the vehicles send data including current geographic coordinates, current vehicle speed and / or current vehicle acceleration to the backend, thereby the backend calculates the relative positive acceleration or negative acceleration (RPA, RNA) of the vehicles with respect to the route portion; and c) calculating the average or distribution of the calculated relative positive or negative acceleration (RPA, RNA) of multiple vehicles in the route portion at the backend.

[0005] Document DE 102019000630 A1 discloses a method for assigning a common geographic location to multiple events caused by a common root cause and acquired by at least one vehicle, wherein a single geographic location is acquired for each individual event. The acquired events are indexed for lexicographical sorting. The geographic locations associated with multiple acquired events are called hotspots. A method is proposed to determine the maximum hotspot by combining spatially adjacent and sufficiently close hotspots. If the set of events assigned to a hotspot is a proper subset of the set of events assigned to another hotspot, then that hotspot is eliminated. Hotspots merged into the maximum hotspot in this way are assigned a common geographic location.

[0006] Document WO 2021191051 A1 discloses a method for identifying potential hazards in road traffic using connected vehicles. Events indicating potential hazards are acquired, and their geographic locations are transmitted to a central computing unit. When multiple similar events with the same geographic location are acquired, the transmitted events are input as hotspots into a digital map. Contextual information is added to the transmitted events. Hotspots are analyzed to identify potential hazards. Current hotspots are compared with confirmed hotspots. Hotspots are displayed visually on a platform. The geographic location of hotspots for specific traffic-critical events is transmitted to vehicles located near those hotspots.

[0007] Therefore, hotspots known from the prior art can be understood as (locally restricted) clusters / aggregates of events associated with the same or very similar geographical locations. These clusters (or hotspots) are characterized by absolute frequency and maximum radius. Absolute frequency represents the frequency with which events are acquired within the maximum radius surrounding the hotspot (over a predetermined time period).

[0008] The drawback of this known form of describing hotspots is that traffic-related events typically do not occur in circular areas surrounding a geographic location, but rather along traffic routes.

[0009] Furthermore, the analysis of such hotspots becomes difficult: what is usually relevant for assessment is not the absolute frequency of the event, but its relative frequency (the number of times a vehicle traverses the location), because it estimates the probability of the event occurring when a vehicle passes through that location.

[0010] However, due to technical reasons and data protection regulations, the number of vehicles passing through hotspot geographic locations is often unknown or only imprecise, so the relative frequency of events cannot be determined or can only be roughly determined.

[0011] Therefore, there is a need for an improved method for identifying, classifying, and analyzing hotspots, characterized by the salience of an indicator value that represents the event description obtained at such a hotspot.

[0012] The publication E. Schubert, A. Zimek, H.-P. Kriegel: Generalized Outlier Detection with Flexible Kernel Density Estimates. Proceedings of the 14th SIAM International Conference on Data Mining (SDM), Philadelphia, PA, 2014, pp. 542–550, https: / / doi.org / 10.1137 / 1.9781611973440.63 describes a method for analyzing the association between density function estimation based on this type of density function and outlier identification. Summary of the Invention

[0013] The purpose of this invention is to provide an improved method for determining indication values ​​to statistically characterize location-related event descriptions acquired by at least one vehicle.

[0014] According to the present invention, this objective is achieved by a method having the features of claim 1.

[0015] Advantageous designs of the present invention are the subject of the dependent claims.

[0016] To acquire hotspots, parameters of the vehicle environment are continuously acquired in at least one vehicle. For example, the dynamic vehicle environment can be acquired by measuring the distance, speed, and trajectory of surrounding vehicles, pedestrians, and / or other road users (relative to the vehicle's own position) using stereo cameras and / or radar and / or lidar equipment or similar measurement methods. Additionally or alternatively, the vehicle sensor system can acquire static vehicle environment parameters, such as the arrangement and characteristics of lane markings and / or traffic signs and / or traffic control equipment. For example, lane marking spacing, lane marking type, lane width, curvature, and similar parameters can be acquired and measured.

[0017] In parallel with acquiring these vehicle environmental parameters, the vehicle's location is continuously determined, i.e., by means of a geographic location determination system that uses a Global Navigation Satellite System (GNSS) or a similar geographic location determination system.

[0018] Objects and their attributes acquired by various sensors in static and / or dynamic vehicle environments are incorporated into an environment model, which may also include map material from roads or navigation maps, as well as optional additional dynamic, geolocation-related real-time data, such as current traffic volume.

[0019] Within this complex environmental model, triggers are defined to describe noteworthy (particularly traffic-related) events for vehicle driving. For example, such events could be warnings or interventions from assistance systems used to monitor the vehicle's environment. When such a trigger is triggered, an event description is generated and sent from the corresponding vehicle to the backend of the analysis system. The event description includes the corresponding geographic location that triggered the trigger, as well as sensor data and parameters related to the event description acquired by the vehicle's sensor systems.

[0020] The analytics system can be implemented, for example, on a server or in the cloud, and as a backend, it provides a set of predefined services and / or functions via an Application Programming Interface (API).

[0021] The analysis system generates an analytical map based on road or navigation maps or similar map materials, from event descriptions transmitted from one or more vehicles to the backend. In this map material, traffic routes, particularly roads, are assigned geographic locations, which are referred to below as road network nodes. In other words, while geographic locations can generally be associated with any terrain feature, road network nodes refer to those geographic locations that are accessible by road.

[0022] First, identify the longest possible continuous road segment that is not interrupted by intersections. Specifically, select road segments that meet or exceed a predetermined minimum length to create the analysis map. Typically, such road segments belong to highways or expressways with only a few intersections. Road segments shorter than the predetermined minimum length are filtered out and not used for the analysis map.

[0023] Then, adjacent road segments are linked together (as a linked list) to form road segment chains. In other words, an ordered sequence of road segments forms a topologically one-dimensional (i.e., linearly extending only) road segment chain, with a unique start and a unique end point along its linear extension. Each road segment chain is assigned a chain identifier and a start point in a unique manner (relative to the analysis map). Therefore, any position along a road segment chain can be uniquely determined by the chain identifier and its distance from the start point of the road segment chain.

[0024] In a method for determining indication values ​​to statistically characterize location-related event descriptions acquired by at least one vehicle, no-inflow / no-outflow road segments / basic road segments are identified based on road or navigation maps or similar map materials, where traffic routes are drawn with reference to geographic locations. Along such road segments, it can be assumed that the number of vehicles passing through is approximately equal everywhere. In this way, longer traffic routes are divided into segments that connect with at least one other road segment at intersections, junctions, exits, entrances, and similar nodes.

[0025] According to the invention, among the road segments determined in this way, road segments whose length is greater than or equal to a predetermined minimum road segment length are identified. The selection of road segments can be successfully limited to, or at least substantially limited to, specific types of traffic routes for which it is a good approximation to assume substantially the same vehicle throughput (along their respective traffic routes). In particular, the segment length can be selected such that the selected road segments are associated with highways or expressways.

[0026] In subsequent steps, among the identified road segments, those segments that are connected to each other along traffic routes via intersections, junctions, exits, entrances or similar nodes are identified, and the geographical location of one end of the first identified road segment corresponds to the geographical location of one end of the second identified road segment.

[0027] A series of road segments connected to each other in a linked list manner is called a road segment chain. A road segment chain extends linearly from its starting point to its ending point, so there are no branches where traffic flow can diverge. Therefore, for such road segment chains, assuming the same traffic flow everywhere is a particularly good approximation.

[0028] At least one event description is acquired from at least one vehicle and transmitted to the analysis system. This event description includes at least one parameter of the vehicle's environmental model, including the vehicle's static and / or dynamic environment and its operational and / or control parameters. Furthermore, the event description includes the geographical location from which these parameters were acquired.

[0029] In particular, event descriptions are acquired and transmitted when the vehicle environment model describes states or state changes that are particularly important for assessing traffic flow and / or vehicle state.

[0030] The analysis system evaluates the input and / or stored event descriptions. The analysis system then assigns the nearest road network node to each of the event descriptions' geographic locations—which are typically not precisely aligned with a road network node due to tolerances and inaccuracies when captured by the positioning system—usually by (vertically) projecting onto the nearest road network node.

[0031] The nearest road network node is described by a unique identifier of the road segment chain and a longitudinal distance, which (in a predetermined length unit chosen equally for all road segment chains) indicates the distance from the starting point of the road segment chain.

[0032] Furthermore, the analysis system utilizes a kernel density estimator to determine the curve / variation / distribution of event description density with respect to longitudinal distance from the starting point for at least one road segment chain. To this end, a density (density value) is determined over an increasing, preferably equidistant, longitudinal distance starting from the starting point of the road segment chain. This density (density value) estimates the frequency of event descriptions within the increments of longitudinal distance. By using a kernel density estimator, the dispersion of the determined values ​​can be reduced by considering the frequency of event descriptions of adjacent longitudinal distance increments when determining the density within the increments (in a weighted moving average manner).

[0033] The density is determined at specific locations along the longitudinal distance from the start of the road segment chain by submitting the longitudinal distances (L, L1, L2) of the event descriptions assigned to this road segment chain via projection to a kernel density estimator with a predetermined bandwidth. Here, a series of Dirac pulses arranged according to the assigned longitudinal distances are convolved with a non-negative, preferably symmetric, kernel weight function.

[0034] Methods for selecting kernel weight functions, and in particular for selecting the bandwidth of such kernel weight functions, are known from the prior art, for example from the publication Wand, MP; Jones, MC (1995). Kernel Smoothing. London: Chapman & Hall / CRC. ISBN 978-0-412-55270-0.

[0035] Then, the density of each event description determined along the segment chain is converted into Z-values ​​through local Z-normalization.

[0036] Here and below, local Z-normalization is understood as: for density values ​​determined for one or a group of road network nodes. The transformation performed, which takes into account the density value A set of density values ​​defined around each road network node and along its respective road segment chain within a predetermined area called bandwidth. For this set of density values, determine the (local) average value. and (local) standard deviation Local Z-normalization, according to regulations, sets the density value... Convert to Z value

[0037] In other words, local Z-normalization converts density values ​​into Z-values, which are based on a region called bandwidth and represent the deviation of density from the mean using standard deviation as a benchmark.

[0038] A higher local Z-value indicates more significant variability of the considered points relative to the region, regardless of absolute density. A higher Z-threshold can reduce the result set to more significant locations.

[0039] Therefore, the Z-value can be used to characterize and identify network nodes as salient, especially those network nodes whose probability of acquiring / reporting event descriptions is much greater than that of network nodes in a predetermined area along the road segment chain. One advantage of this method compared to existing techniques is its ability to reliably characterize this saliency independently of acquiring the absolute (total) number of vehicles passing through the network node.

[0040] On the one hand, this advantage is achieved by forming a road segment chain according to the present invention, thereby selecting road network nodes with relatively uniform vehicle density. On the other hand, this advantage is achieved by converting density values ​​into an index through local Z-normalization, which is largely insensitive to fluctuations in the dispersion of density values ​​along the road segment chain.

[0041] In one embodiment of the method, for the road segments intended to form a segment chain, the selected predetermined minimum length should ensure that at least one segment chain consists entirely or primarily of road segments belonging to highways or expressways. This achieves particularly good uniformity of traffic flow along the formed segment chain. In other words, density values ​​obtained along such a segment chain are particularly comparable due to relatively low inflows and outflows, and in particular, an increase in the density value described by an event reliably indicates abnormal traffic conditions.

[0042] In one implementation of this method, the vehicle environment model integrates static and dynamic vehicle environments. Here, "static vehicle environment" refers to acquiring a vehicle environment that remains substantially unchanged over time, particularly one unaffected by the activities of other road users, such as the arrangement and characteristics of lane markings and / or traffic signs and / or traffic control equipment. For example, the static vehicle environment can be acquired using environmental cameras, distance sensors, radar, and lidar sensors.

[0043] The "dynamic vehicle environment" here should be understood as the vehicle environment that changes over time, especially the vehicle environment that changes due to the movement of other road users. For example, the dynamic vehicle environment can be obtained by using stereo cameras and / or radar and / or lidar equipment or similar measurement methods to measure the distance, speed, and trajectory of surrounding vehicles, people, and / or other road users relative to the vehicle itself.

[0044] In one embodiment of this method, a peak region is identified as a region of road network nodes that are consecutive between the peak start and peak end points along the road segment chain and have a Z-value exceeding a predetermined threshold. Therefore, regions with significant traffic conditions, such as accident centers / accident black spots, can be identified with particularly good reliability.

[0045] Preferably, the predetermined threshold is selected as 1. Based on the statistical principle that density values ​​approximately follow a normal distribution, this method allows for the identification of peak regions where the event description density is higher than approximately 85% of the road network nodes in the road segment chain. Therefore, peak regions with significant traffic conditions can be identified with particular reliability.

[0046] In one embodiment of the method, peak regions are assigned identifiers for corresponding road segment chains (e.g., uniquely assigned as numbers or strings), longitudinal distances from the peak start point to the start point of the corresponding road segment chain, longitudinal distances from the peak end point to the end point of the corresponding road segment chain, the maximum value of all Z values ​​within the peak region, and a description of the geometry of the peak region. The description of the geometry of the peak region can, for example, be specified as a set of geographic locations assigned to all or part (e.g., equidistantly selected) road network nodes within the peak region. Based on these characteristics, peak regions can be evaluated particularly easily and reliably.

[0047] In one embodiment of the method, the vehicle environment model includes at least one object and / or parameter acquired by vehicle sensors. Based on the vehicle environment model, triggers for events worthy of observation, such as events important for vehicle control and guidance, are identified and / or flagged. Based on such triggers, event descriptions are generated and transmitted to the analysis system.

[0048] Using such triggers, you can generate descriptions of specific events that are of interest and relevant to subsequent analysis.

[0049] The trigger preferably flags warnings and / or interventions issued by a driver assistance system. As a result, driving situations where the driver's behavior may require correction can be identified in the event description. For such driving situations, determining statistical indicators based on road network nodes (i.e., based on the specific location of road segment chains) enables particularly convincing assessments of potentially hazardous locations and the identification of corresponding peak areas, such as potential accident black spots. Therefore, indicators that contribute to improving road safety can be determined.

[0050] In one implementation of this method, event categories are defined and recorded in an event category list. Event categories may be (by way of example only) braking operations initiated by the driver assistance system, vehicle parameters indicating a decrease in vehicle stability, or potential collisions with objects in the vehicle environment acquired by the driver assistance system.

[0051] At least one such event category is assigned to each event description. To determine the density of event descriptions, only those (filtered) event descriptions whose event categories meet specific filtering criteria are used. For example, only event descriptions of a specific event category or a subset of event categories are used. Therefore, particularly specific evaluation of density values ​​and / or particularly specific determination of peak regions are possible.

[0052] In one implementation, the direction of the road segment chain points from the start point to the end point, and the event description includes the direction of travel associated with the direction of the road segment chain. To determine the density of event descriptions, only event descriptions with the same direction of travel are considered, i.e., where vehicles (along the road segment chain) move in the same direction. Therefore, anomalies associated with the direction of travel can be identified particularly specifically (e.g., distinguishing downhill and uphill travel on steep sections, or different hazards depending on the direction of travel, such as hazards caused by obstructed vision, entering or exiting ramps, or the radius of a curve).

[0053] In one embodiment of the method, the Z-values ​​determined at road network nodes along at least one road segment chain are visualized in a map display. For example, visualization can be achieved using color representation, as different colors are assigned to different regions of the Z-values. Additionally or alternatively, the Z-values ​​can be visualized using markers drawn perpendicular to the respective road segment chain, wherein the length of the marker drawn on the road network node is selected based on the Z-value determined for that road network node. Preferably, the length of such markers is selected to be proportional to the corresponding Z-value.

[0054] Therefore, the gradual differences in salient points (especially dangerous points) can be presented particularly clearly along the road segment chain without the need for arbitrary threshold settings and peak area segmentation. Attached Figure Description

[0055] Embodiments of the present invention will now be explained in more detail with reference to the accompanying drawings.

[0056] in: Figure 1 An analytical map with road segment chains is shown schematically; Figure 2 The determination of the Z-normalized Z-value for the event description density determined by means of kernel density estimation is illustrated schematically. Figure 3 The diagram illustrates the relationship between the Z-value and the direction of travel along the road segment chain. Detailed Implementation

[0057] Corresponding parts in all the accompanying figures are labeled with the same reference numerals.

[0058] Figure 1 An analytical map with segment chain 10 is schematically shown. Segment chain 10 is composed of non-intersection road segments (not shown in detail) that have a length equal to or exceeding a predetermined minimum length. Each of these road segments (in...) Figure 1 (Not specifically specified) includes at least one, typically multiple, road network nodes O1 and O2. These road network nodes can, in principle, be distributed arbitrarily densely (limited only by the resolution of the underlying map material). To improve overview, Figure 1 Only the first and second road network nodes O1 and O2 are shown as examples.

[0059] For example, road segment chain 10 can be obtained by analyzing road maps or navigation maps. Figure 1 It shows its position and extent relative to a geographic coordinate system, which is shown in a roughly simplified manner and schematically has a longitudinal coordinate (geographic longitude), represented in this example as the x-direction x, and a latitude coordinate (geographic latitude), represented in this example as the y-coordinate y.

[0060] The road segment chain 10 has a starting point 11 at the first end and an ending point 12 at the opposite second end, and is identified by an identifier, which in this case is chosen as the string "#2" for example only. At least the starting point 11 and the ending point 12 are associated with a road network node (not specifically described in this example) and a geographical location (also not specifically described in this example), respectively.

[0061] Each point along segment chain 10 can be uniquely identified by the longitudinal distance L to the starting point 11 (in a distance unit that can be arbitrarily chosen but is the same for all segment chains 10) and the identifier of segment chain 10.

[0062] Therefore, for each point on a road segment chain 10, a road network node O1 or O2 on the road segment in the underlying road map or navigation map can be assigned based on its distance from the starting point 11. For example, the sequence of synthetic road segments of the road segment chain 10 can be traced from the starting point 11 until the corresponding longitudinal distance L from the starting point 11 is reached. Thus, by referring to the geographical location of the starting point 11, coordinates in the geographic coordinate system can be assigned to each point on the road segment chain 10.

[0063] For clarity, Figure 1 The diagram illustrates two geographic locations, P1 and P2, which are determined by one or more vehicles through a positioning system and transmitted along with an event description. For example, at each of these geographic locations P1 and P2, a braking process may have been triggered by a driver assistance system. Due to inaccuracies in the geographic location determination system, the geographic locations P1 and P2 reported by different vehicles are typically off-target from the route of road segment chain 10 (i.e., the set of road network nodes O1 and O2 of road segment chain 10 in the geographic coordinate system).

[0064] For each geographic location P1 and P2, assign the nearest road network node O1 or O2 along the road segment chain 10. Typically, the nearest point can be uniquely determined by vertical projection onto the road segment chain 10. When multiple nearest road network nodes O1 and O2 exist at equal distances, one of these nodes can be randomly selected.

[0065] As an example, Figure 1 A first geographic location P1 is shown, described by a first pair of geographic coordinates x1 (along the x-direction x) and y1 (along the y-direction y). A first road network node O1 is assigned to this first geographic location P1 on the road segment chain 10, located at a first longitudinal distance L1 from the starting point 11. As an example, assume the first longitudinal distance L1 has 5.2 length units. The first geographic location P1 can then be described by the identifier "#2" of the road segment chain 10 and the longitudinal distance L1 from the starting point 11 "5.2 length units".

[0066] In the analysis, the event description reported to the first geographic location P1 is assigned the first road network node O1 and the lateral distance d1 from the first geographic location P1 to the nearest first road network node O1, which is "1.0 length unit" in this example. For example, all information about the identifier of the associated road segment chain 10 and the location and lateral distance of the nearest road network node O1 can be summarized in a single string "#2, 5.2, +1", which is assigned to the first geographic location P1. Typically, this mapping will not be unique; that is, event descriptions may exist at other (different) geographic locations, and these geographic locations may be assigned the same string.

[0067] Similarly, the second geographic location P1 is described by a second pair of geographic coordinates x2 (along the x-direction x) and y2 (along the y-direction y). The second network node O2 is assigned to the second geographic location P2 as the nearest point in the road segment chain 10. Furthermore, based on the second longitudinal distance L2 (in this example, "4.1 length units") from the nearest network node O2 to the starting point 11 and the lateral distance d2 (in this example, "1.0 length units") between the network node O2 and the second geographic location P2, the specified string "#2, 4.1, +1" is assigned to the second geographic location P2.

[0068] Using this allocation rule, which is explained here only as an example, a point can be assigned to each geographic location P1, P2, in a typical curvilinear coordinate system given by the set of road segment chains 10. Therefore, determining the position relative to the nearest road segment chain 10 can be simplified to an indication of the longitudinal distance L relative to its starting point 11.

[0069] By performing all subsequent evaluations on each segment chain 10, the complexity of the captured data can be reduced from two dimensions (coordinates along the x-direction and y-direction) to only one dimension (the longitudinal distance from the starting point 11 of the respective segment chain 10).

[0070] The density distribution is calculated from the event descriptions obtained along the road segment chain 10 by applying a method known in the prior art as kernel density estimation, based on the longitudinal distance L of the road network points O1 and O2 assigned to their respective geographical locations P1 and P2.

[0071] Therefore, for a road segment chain 10, the density p of the event description is determined at regular intervals—for example, every meter along a longitudinal distance L starting from the starting point 11. Figure 2 The upper part is shown schematically.

[0072] The longitudinal distances (L, L1, L2) assigned to the event description by projecting onto the nearest segment chain (10) are submitted to a kernel density estimator with a predetermined bandwidth for kernel density estimation to determine the density p. In this case, the Dirac pulse sequence arranged over the assigned longitudinal distances (L, L1, L2) is convolved with a non-negative, preferably symmetric, kernel weight function.

[0073] Here, density p can be determined based on one or more categories of event descriptions. For example, one category of event description (also known as event category) can be defined by braking intervention of a driver assistance system (such as an anti-lock braking system), while another category of events can be defined by a sudden decrease in vehicle speed.

[0074] Therefore, the sum of the density values ​​p determined in this way on road segment chain 10 is equal to the number of data points used (i.e., the total number of event descriptions for the corresponding event categories obtained on road segment chain 10). Thus, different road segments can be compared, density values ​​p determined based on different data bases can be compared, and density values ​​p based on such different data bases can be weighted.

[0075] Typically, at some road network nodes O1 and O2, the density p of event descriptions is significantly greater than the statistical average. These particularly significant locations are referred to below as peak regions R1, R2, and R3. Peak regions R1, R2, and R3 may be caused by, for example, chaotic traffic rules, special weather exposures (e.g., lanes prone to black ice or fog, or lanes with increased risk of slippage during rain), or they may also be caused by structural features of traffic management (road narrowing, pavement replacement, road damage, etc.).

[0076] To identify peak regions R1, R2, and R3, road network nodes O1 and O2, or regions comprising multiple such road network nodes O1 and O2, are classified. This classification is also known as scoring.

[0077] This invention is based on the understanding that it is based on a constant threshold. The score is determined when the density p of event descriptions acquired there (e.g., the number of braking operations acquired within a one-meter radius around road network nodes O1 and O2) exceeds a predetermined constant threshold. (For example, in the case of 100 braking operations), classifying road network nodes O1 and O2 into peak regions R1, R2, and R3 is not a satisfactory approach.

[0078] In particular, such scoring presupposes knowledge of the number of vehicles passing through each road segment. However, this requirement is difficult to meet in practice due to the explained technical and / or organizational legal reasons. In methods known in the prior art, this results in traffic-dense road network nodes O1, O2 being marked as significant, even though there is no higher risk of intervention, for example, by driver assistance systems, there.

[0079] To avoid this problem, the present invention is based on the assumption that the vehicle density along segment chain 10, particularly between spatially adjacent road network points O1 and O2, fluctuates only slightly. This assumption is particularly reasonable for segment chain 10 constructed as a highway or expressway, because for such traffic routes, changes in the number of passing vehicles can only occur at inbound and outbound ramps. In particular, this makes the score values ​​between adjacent inbound or outbound ramps highly comparable.

[0080] It is recommended to assign a score, called the Z-value or Z-score, to each data point. To do this, within a symmetrical influence region called the bandwidth surrounding road network nodes O1 and O2, the standard deviation of the density p, also determined within this symmetrical influence region, is used. Using this as a benchmark, determine the deviation of density p from its average value.

[0081] To illustrate this method, assuming a road segment, and assuming the number of passing vehicles is approximately independent of location and constant, density values ​​for event descriptions are determined at the first to Nth road network points O1 and O2, respectively. These event descriptions are reported within a predetermined bandwidth around each of the road network points O1 and O2. An empirical average (sample average) is then calculated from this. and empirical standard deviation .exist and The calculation only includes event descriptions of road network points O1 and O2 located within a predetermined bandwidth around the points O1 and O2 to be scored. In other words: and Determined to be around the corresponding density value The moving average or moving standard deviation within the bandwidth range .

[0082] In this case, it is possible to determine the moving average by weighted average. and / or sliding standard deviation The density p measured at a greater distance from road network nodes O1 and O2 has a smaller weight than the density p measured at a closer distance. For example, adjacent densities p can be weighted in a way that decreases linearly with their distance from their respective road network nodes O1 and O2.

[0083] average value Curve of standard deviation The curve along link 10 is in Figure 2 The upper part is plotted as a supplement to the density p curve.

[0084] Based on average and standard deviation For each density value Assign a normalized density value or Z-value Z:

[0085] In other words, the corresponding density value of the event description A statistical Z-standardization process is applied, limited to its range of influence (its bandwidth). Within this bandwidth, the corresponding determined Z-values ​​have a Z-mean of zero and can have a standard deviation of 1. As a result, even though the Z-values ​​were determined for road segments with different conditions (especially different traffic densities), the Z-values ​​were still highly comparable.

[0086] exist Figure 2 In the middle, a curve showing the Z value Z along a road segment chain 10 is displayed, which is shown according to the distance length L.

[0087] Based on the Z-value Z, the peak regions R1, R2, and R3 can be easily identified as statistically significant regions of road segment chain 10. For example, peak regions R1, R2, and R3 can be identified as continuous regions along road segment chain 10 where the Z-value Z exceeds a predetermined threshold. .

[0088] As an example, in Figure 2 The threshold is shown in the central area. In other words, exceeding that threshold... This indicates that the density value p is relative to the (sliding) mean. The deviation exceeded the (slipping) standard deviation. However, other thresholds can also be used. .

[0089] Based on the threshold exceeded in this way, the curve of the Z-value Z can be transformed into a binary score S:

[0090] exist Figure 2 In the lower region, as an example (for The curve of the score value S along road segment chain 10 is shown. The first to third peak regions R1, R2, and R3 are determined based on the score value S.

[0091] According to the present invention, the peak regions R1, R2, and R3 are allocated as follows: - The identifier for the corresponding road segment chain 10 (in this example, the string "#2"). - The peak starting point Lmin, which is the point closest to the starting point 11 among the peak regions R1, R2, and R3, and - The peak endpoint Lmax is the point in the peak region R1, R2, R3 that is farthest from the starting point 11.

[0092] To improve clarity, Figure 2 Only the peak start point Lmin and peak end point Lmax are marked for the second peak region R2.

[0093] Furthermore, explicit lengths and / or maximum values ​​and / or geometric descriptions can be assigned to peak regions R1, R2, and R3. The length represents the distance between the peak starting point Lmin and the peak ending point Lmax. The maximum value represents the maximum density p and / or the maximum Z value Z in peak regions R1, R2, and R3. The geometric description describes the arrangement and location of peak regions R1, R2, and R3 in a geographic coordinate system, for example, based on the longitude and latitude of the centers of peak regions R1, R2, and R3.

[0094] Figure 3 The route of a road segment chain 10, which consists of multiple road segments connected by intersections, junctions, exits, entrances, and similar nodes, is schematically shown in a simplified geographic coordinate system. On both sides of the road segment chain 10, to the right of the travel directions I1 and I2 respectively, Z curves Z1 and Z2 with Z values ​​Z are given by a marker M perpendicular to the road segment chain 10.

[0095] The length of the vertical marker M assigned to the first Z-curve Z1 indicates the Z-value Z as a function of position, which is obtained corresponding to the first direction of travel I1. Correspondingly, the length of the vertical marker M associated with the second Z-curve Z2 indicates the Z-value Z obtained by a vehicle moving in the opposite second direction of travel I2.

[0096] In addition to Figure 2 In addition to the method described above for determining peak regions R1, R2, and R3 based on exceeding a threshold θ, Figure 3 The representation shown can also clearly display road sections where the incidence of events is increasing.

Claims

1. A method for determining an indication value (Z) for statistically characterizing a location-related event description acquired by at least one vehicle. Its features are, - Based on road or navigation maps, identify road segments with no inflow and no outflow that have a length greater than or equal to a predetermined minimum segment length. - Link the identified adjacent road segments into a road segment chain (10) with a start point (11) and an end point (12). - The vehicle acquires an event description including its geographic location (P1, P2) and at least one parameter of the vehicle's environmental model, and transmits this event description to the analysis system. - The analysis system assigns the nearest road network node (O1, O2) of the road segment chain (10) to the geographical location (P1, P2) described for each event, and the nearest road network node is described by the longitudinal distance (L, L1, L2) relative to the starting point (11) of the road segment chain (10). - For at least one segment chain (10), the density value (p) of the event description is determined as a curve of the longitudinal distance (L, L1, L2) to the starting point (11) by means of a kernel density estimator with a predetermined bandwidth. - The density value (p) curve is transformed into a Z curve (Z1, Z2) of the Z value (Z) over the bandwidth by using local Z-normalization, which has zero mean over the bandwidth.

2. The method according to claim 1, Its features are, The predetermined minimum segment length is selected such that the at least one segment chain (10) is formed entirely or primarily by segments associated with highways or expressways.

3. The method according to any one of the preceding claims, Its features are, The vehicle environment model integrates static and dynamic vehicle environments.

4. The method according to any one of the preceding claims, Its features are, The Z-values ​​(Z) that exceed a predetermined threshold will be connected along the road segment chain (10) between the peak start point (Lmin) and the peak end point (Lmax). The regions of road network nodes (O1, O2) are identified as peak regions (R1, R2, R3), and the predetermined threshold is preferably a value of 1.

5. The method according to claim 4, Its features are, Assign identifiers for corresponding road segment chains (10), longitudinal distances (L, L1, L2) from the peak start point (Lmin) to the starting point (11) of the corresponding road segment chain (10), longitudinal distances (L, L1, L2) from the peak end point (Lmax) to the ending point (12) of the corresponding road segment chain (10), the maximum value of all Z values ​​(Z) in the peak region (R1, R2, R3), and a description of the geometry of the peak region (R1, R2, R3).

6. The method according to any one of the preceding claims, Its features are, The vehicle environment model includes at least one object and / or parameter acquired by vehicle sensors, generates an event description based on triggers, and transmits the event description to an analysis system, wherein the triggers mark events that are identified as observable by the vehicle environment model.

7. The method according to claim 6, Its features are, The trigger flags interventions and / or warnings from the driver assistance system.

8. The method according to any one of the preceding claims, Its features are, Assign an event category selected from the list of event categories to at least one event description, considering only the event descriptions of one or more event categories in the determination of density (p).

9. The method according to any one of the preceding claims, Its features are, The direction of the road segment chain (10) is from the starting point (11) to the ending point (12). The event description includes the travel direction (I1, I2) related to the direction of the road segment chain. Only the event descriptions of the same travel direction (I1, I2) are considered in the determination of density (p).

10. The method according to any one of the preceding claims, Its features are, The Z value (Z) determined at the road network node (O1, O2) along at least one road segment chain (10) is visualized by a sign (M) extending perpendicular to the road segment chain (10), the length of which is chosen to be proportional to the corresponding Z value (Z).

Images