Method and device for determining crossing parameters

Abstract

Method (100) for determining a value (117, 118) of a map attribute of a node between a plurality of traffic routes for a digital map (111); wherein the map attribute is a property of the node; wherein the method (100) comprises - determining (102) a simulated traffic situation (114) at the intersection for an assumed value (117) of the attribute; - Determining (103) a measured traffic situation (115) at the intersection based on a plurality of measured trajectory data (112) of vehicles at the intersection; wherein the trajectory data (112) of a vehicle comprise a trajectory of the vehicle when crossing the intersection; - comparing (104) the measured traffic situation (115) with the simulated traffic situation (114); - determining (105) the value (118) of the attribute so that the simulated traffic situation (114) approximates the measured traffic situation (115); and - Providing the value (118) of the attribute as an update of the digital map (111) on the vehicles for use in various driver assistance systems.

Description

[0001] The invention relates to a method and a corresponding device for determining one or more parameters or attributes of a junction, in particular a road intersection.

[0002] Increasing the level of detail in digital maps plays an important role in the further development of driver assistance systems. One example of missing details in digital maps is intersection parameters, such as the type of traffic control at an intersection or the position of stop lines. Manually measuring these parameters or extracting them from city maps is labor-intensive and costly.

[0003] Such parameters can, however, be helpful in various assistance systems. For example, switching off an internal combustion engine using an automatic engine start / stop system is typically only beneficial in terms of consumption if the engine remains switched off for several seconds. However, if the engine is switched off for just one second, for example, consumption can increase compared to if the engine had not been switched off. Based on knowledge of the reason for a stop (e.g. the presence of a stop sign or traffic light) and / or the relative position of the stopped vehicle to a stop line, an estimate of the expected downtime can be made. This can then enable the engine to be switched off or not switched off in a situation-adaptive manner, resulting in optimized consumption for the vehicle.

[0004] Another example is camera-based assistance systems. Driver assistance systems are increasingly based on information from cameras. For example, the driver can be supported in complex intersection scenarios using a right-of-way and intersection assistant. The basis for such systems is the camera-based recognition of the right-of-way rules at an intersection (presence of a traffic light, a yield sign, a give-way sign, a stop sign, the right-of-way rule, etc.). To increase safety, intersection parameters in digital maps could be incorporated into the detection of the right-of-way situation as a priori knowledge. This could prevent potential false triggers from a purely camera-based system.

[0005] Driver assistance systems can, if necessary, access an internet connection and a central server system, the backend. This allows them to access information from a large number of other vehicles (crowdsourcing). One example of this is the display of current traffic flow in the vehicle, in an internet browser, or via a smartphone application. However, the use of internet-based services in vehicles is currently limited to navigation and entertainment.

[0006] DE 10 2005 053 461 A1 describes a method for traffic control. DE 10 2011 083 677 A1 describes a method for predicting a traffic situation for a vehicle.

[0007] This document deals with the acquisition of data for enriching digital maps with additional map attributes (such as intersection parameters). In particular, this document addresses the technical task of determining such map attributes efficiently and precisely from a large number of measured vehicle trajectories.

[0008] The object is achieved by the independent claims. Advantageous embodiments are described, inter alia, in the dependent claims.

[0009] According to one aspect, a method is described for determining a value of an attribute (also referred to as a parameter) of a junction between a plurality of traffic routes. The plurality of traffic routes can be joined by the junction. The junction can comprise a level intersection or a roundabout, and the plurality of traffic routes can comprise a corresponding plurality of roads that are joined by the junction. The attribute can be a property of the junction that is currently not typically obtainable from a digital map, such as the position of a stop line at an access road (i.e. on a traffic route) to the junction and / or the absolute traffic densities (possibly time-dependent) on the traffic routes at the junction.

[0010] The method comprises determining a simulated traffic situation at the intersection for an assumed value of the attribute. In particular, a plurality of simulated traffic situations can be determined for a corresponding plurality of assumed values ​​of the attribute. The simulated traffic situation can be determined using a microscopic traffic simulation. In a microscopic traffic simulation, the behavior of individual vehicles at the intersection can be simulated. A simulation model of the intersection can be used to carry out the traffic simulation. Carrying out a simulation is advantageous because it typically does not require significant amounts of training data. The simulation can therefore be carried out with relatively little effort.

[0011] The method further comprises determining a measured traffic situation at the intersection. The measured traffic situation can be determined on the basis of a large number of measured trajectory data from vehicles at the intersection. The vehicles can in particular be road vehicles (such as cars or motorcycles). A vehicle can record trajectory data when passing over the intersection and make it available to determine the value of the attribute. The trajectory data can, for example, comprise a trajectory of the vehicle when passing over the intersection. A trajectory can, for example, comprise the position of the vehicle at the intersection as a function of time (possibly including the time of day and / or day of the week). From a trajectory of a vehicle, stopping positions and / or stopping periods of the vehicle at the intersection (e.g. on a traffic route to the intersection) can thus be determined.

[0012] Alternatively or additionally, the trajectory data of a vehicle can directly include specific elements of a trajectory. For example, the trajectory data of a vehicle for a node can include one or more stopping positions and / or stopping periods of the vehicle at the node (e.g., on a traffic route to the node). These selected elements of a vehicle's trajectory can, if necessary, be determined directly by the vehicle and provided as trajectory data for determining a value of the node's attribute. This can reduce the amount of trajectory data transmitted by the vehicle to a central server.

[0013] The method further comprises comparing the measured traffic situation with the simulated traffic situation and determining the value (e.g., the estimated value) of the attribute such that the simulated traffic situation approximates the measured traffic situation. Thus, the described method enables the value of the attribute to be determined precisely and effectively.

[0014] As already explained above, the method may comprise determining a plurality of simulated traffic situations at the intersection for a corresponding plurality of assumed values ​​of the attribute. The plurality of simulated traffic situations may be compared with the measured traffic situation to determine the estimated value of the attribute. In particular, a plurality of comparison measures between the measured traffic situation and the corresponding plurality of simulated traffic situations may be determined. The estimated value of the attribute may then be determined or selected from the plurality of assumed values ​​of the attribute. In particular, the value corresponding to a relatively improved comparison measure may be selected from the plurality of comparison measures.

[0015] The comparison measure can include an indication of a difference between the measured traffic situation and the simulated traffic situation. In particular, a relatively large comparison measure can be an indication of a relatively large difference. The assumed value for the attribute can be selected as an estimate, which reduces the comparison measure (and thus the difference). The use of a comparison measure enables the use of an optimization method (e.g., a gradient method). This can reduce the computational effort required to determine the estimated value of the attribute. In particular, this can reduce the number of simulated traffic situations to be determined.

[0016] The method may comprise determining one or more model parameters of a simulation model for the intersection. The one or more model parameters may be determined based on the plurality of measured trajectory data and / or based on a digital map. The one or more model parameters may be constant and used to determine the plurality of simulated traffic situations. The one or more model parameters may comprise one or more of: a geometry of the intersection (which may be determined, for example, from the digital map), a relative traffic density on the plurality of traffic routes of the intersection (which may be determined, for example, from the plurality of measured trajectory data), and / or a maneuver probability for possible maneuvers of a vehicle at the intersection (which may be determined, for example, from the plurality of measured trajectory data).

[0017] The simulation model can also include the assumed value of the attribute as an additional model parameter. As already explained, the assumed value can be changed to determine the multitude of simulated traffic situations. The assumed value of the attribute can thus be used as a variable model parameter to change the simulation model. A changed simulated traffic situation can then be determined based on a changed simulation model. By using a changeable simulation model, it can be ensured that the various simulated traffic situations can be determined precisely and efficiently.

[0018] The simulated traffic situation and the measured traffic situation can include the traffic behavior of the vehicles at the intersection with respect to the intersection attribute to be determined. In other words, the simulated and measured traffic situation can depend on the value of the attribute to be determined. In particular, there can be a (possibly continuous) relationship between the attribute and the traffic situation. This can ensure that the comparison of the traffic situations allows a conclusion to be drawn about the value of the attribute, and thus that the comparison of the traffic situations enables the determination of the estimated value of the attribute.

[0019] For example, the attribute to be determined may include the position of a stop line of a first traffic route at the intersection. The simulated traffic situation and the measured traffic situation may then include a distribution of stopping positions of the vehicles on the first traffic route.

[0020] The method can be used to determine values ​​of a plurality of attributes of the intersection. The method can then include determining the simulated traffic situation at the intersection for assumed values ​​for the plurality of attributes. Furthermore, the values ​​of the plurality of attributes can be determined so that the simulated traffic situation approximates the measured traffic situation. In this way, the estimated values ​​of a plurality of attributes can be determined efficiently and precisely.

[0021] According to another aspect, a software (SW) program is described. The SW program can be configured to be executed on a processor and thereby to carry out the method described in this document.

[0022] According to a further aspect, a storage medium is described. The storage medium can comprise a software program configured to be executed on a processor and thereby to carry out the method described in this document.

[0023] It should be noted that the methods, devices, and systems described in this document can be used both alone and in combination with other methods, devices, and systems described in this document. Furthermore, any aspects of the methods, devices, and systems described in this document can be combined in a variety of ways. In particular, the features of the claims can be combined in a variety of ways.

[0024] The invention will be described in more detail below using exemplary embodiments. Fig. 1 is a flowchart of an exemplary method for determining an attribute of a node; Fig. 2 a comparison of an exemplary simulated traffic situation and an exemplary measured traffic situation; and Fig. 3 shows a course of an exemplary comparison measure.

[0025] As stated at the beginning, this document deals with the efficient and precise determination of one or more map attributes for a digital map. An example map attribute is, for example, the position of a stop line (or the stop line position) at a road intersection and / or the absolute traffic density at the entrance to an intersection (e.g., depending on the time of day and / or the day of the week). The following describes a method for determining such map attributes. The method is described using the map attributes "stop line position" and "absolute traffic density" as examples. However, it should be noted that the method (and the corresponding device) described in this document can be used analogously for attributes of intersections in general.

[0026] Vehicles can be connected to a central server or a central computer system (e.g., a backend system) via the internet and provide data. In particular, so-called trajectories and / or trajectory data can be recorded when a vehicle crosses an intersection. A trajectory can indicate where the vehicle was at a specific time. A trajectory can thus be used to determine the vehicle's stopping position and the time it spent at that stopping position. Furthermore, the time of day and / or day of the week when the intersection was crossed can be recorded in the trajectory.

[0027] The backend system can thus access a multitude of trajectories or a multitude of trajectory data for crossing a specific intersection. This gives the backend system the ability to determine intersection parameters, such as the absolute traffic density at intersection entrances and / or a stop line position.

[0028] In other words, the data recorded in the vehicle (trajectory data) can be sent via a communications network to a central processing unit, the backend. In the backend, intersection parameters can be extracted from this data and made available to the vehicles, for example, via map updates. Alternatively or additionally, it is possible to perform part of the process locally in the vehicle in order to reduce transmission costs. In particular, specific elements of a trajectory (e.g., one or more stopping positions of the vehicle) can be determined in the vehicle and provided as compact trajectory data.

[0029] One option for extracting continuous parameters from a large amount of data (trajectory data) is machine learning methods such as linear regression, nonlinear regression, or support vector regression. However, the variables used to determine intersection parameters, such as stopping positions, depend on a large number of influencing factors. For example, the intersection geometry and topology, as well as the traffic density at intersection entrances and the probability of turning maneuvers, influence a vehicle's stopping position. Due to the large number of influencing factors, a large amount of training data (i.e., a large number of intersections with known or labeled parameters) is required for a machine learning method, which would be very complex.

[0030] For this reason, it is proposed to simulate the traffic behavior at the intersection for which further attributes or parameters are to be determined. Simulating traffic behavior at an intersection typically requires only a small amount of training data to create a correct simulation. This allows for a simulation of traffic behavior at an intersection to be carried out with relatively little effort and in a precise manner.

[0031] A generic method for determining one or more parameters of an intersection based on a large number of recorded crossings (e.g., GPS trajectories) is described below. This method is based on a simulation of traffic behavior at intersections. Fig. 1 shows a flowchart of an exemplary method 100 for determining an attribute of a junction between a plurality of roads (e.g., an intersection or a roundabout). The method 100 includes determining 101 a priori information 113 about the junction. The a priori information 113 about the junction can be determined, for example, based on digital maps 111. An exemplary a priori information 113 that can be determined based on a digital map 111 is, for example, the geometry of the junction. The a priori information 113 can also be viewed as model parameters of a simulation model.

[0032] Alternatively or additionally, a priori information (or model parameters) 113 can be determined from the plurality of trajectory data 112. The trajectory data 112 can comprise a trajectory. As explained above, a trajectory can indicate the time of day and / or day of the week a vehicle was at an intersection and how the vehicle behaved at this intersection (e.g., whether it turned right / left or whether it continued straight ahead). Alternatively or additionally, the trajectory data 112 can comprise selected elements of a trajectory (e.g., stopping positions and / or stopping periods).

[0033] If a large number of trajectory data 112 from a large number of vehicles are available, a-priori information 113 such as a relative traffic density at the individual arms / traffic routes of the intersection and / or the maneuver probabilities for a vehicle (e.g. probability of turning right, probability of turning left and / or probability of driving straight ahead) can be determined therefrom.

[0034] The method 100 further comprises carrying out 102 a traffic simulation. The traffic simulation can, in particular, be a microscopic traffic simulation in which the behavior of a large number of vehicles at the intersection is simulated, thereby determining the traffic situation at the intersection. The a priori information or model parameters 113 make it possible to describe the intersection for the simulation (i.e., the simulation model) in terms of geometry and relative traffic densities. Furthermore, the a priori information makes it possible to simulate the behavior of individual vehicles at the intersection (in particular, with regard to the driving maneuvers performed by the vehicle at the intersection).

[0035] Furthermore, assumptions 117 can be made regarding one or more unknown attributes of the intersection. In other words, assumed values ​​117 can be determined for the one or more attributes. In particular, assumptions can be made regarding stopping positions and / or absolute traffic densities of the intersection. These assumptions 117 regarding the unknown attributes can be taken into account in the simulation. In particular, it can be determined which traffic situation would be observed at the intersection if the assumptions 117 were true. This traffic situation can be referred to as the simulated traffic situation 114.

[0036] Carrying out 102 a traffic simulation thus comprises determining a simulated traffic situation 114, taking into account the a priori information (or model parameters) 113 regarding the intersection and taking into account assumptions 117 regarding one or more unknown attributes of the intersection. The assumptions 117 can be considered variable model parameters of a simulation model.

[0037] The method 100 further comprises determining 103 a measured traffic situation 115 based on the plurality of trajectory data items 112. As already explained above, the trajectory data 112 can include information regarding one or more stopping positions of the vehicle at an arm (also referred to as a traffic route) of a junction. From the plurality of trajectory data items 112, for example, a distribution of the stopping positions of the vehicles at the arm of the junction can be determined. The distribution of the stopping positions can reflect the measured traffic situation 115 at the junction. In other words, a traffic situation 114, 115 can be described, for example (among other things), by a distribution of stopping positions of vehicles at the arms of a junction.This component of the traffic situation can be both simulated (to determine the simulated traffic situation 114) and determined from the multitude of trajectory data 112 (to determine the measured traffic situation 115).

[0038] The method 100 further comprises comparing 104 the simulated traffic situation 114 with the measured traffic situation 115. In particular, a comparison measure 116 can be calculated on the basis of the simulated traffic situation 114 and the measured traffic situation.

[0039] The method 100 further includes determining 105 an estimated value 118 for the one or more unknown attributes of the node. The estimated value 118 can be determined as a function of the comparison measure 116, e.g., the value or magnitude of the comparison measure 116. In particular, the estimated value 118 for the one or more unknown attributes of the node can be determined such that the comparison measure 116 is reduced (possibly minimized).

[0040] The reduction or minimization of the comparison measure 116 can be carried out iteratively. For this purpose, the steps of carrying out 102 a traffic simulation and comparing 104 the simulated traffic situation 114 with the measured traffic situation 115 can be repeated for different assumptions (or assumed values) 117 regarding the one or more unknown attributes. The assumptions (i.e., the assumed values) 117 can be selected, e.g., by means of an optimization method, such as a gradient method, such that the comparison measure 116 tends to be reduced from one iteration to the next. Upon reaching a (local) minimum of the comparison measure 116, the iterative process can be terminated. The estimated values ​​118 for the one or more unknown attributes can then be equated with the assumptions 117 for the one or more unknown attributes that led to the comparison measure 116 when the iterative process was terminated.

[0041] Fig. 2 shows an example of a simulated traffic situation 114 and a measured traffic situation 115. In the example shown, the traffic situation is described by the distribution of the stopping positions (relative to the center 201 of the intersection) of vehicles at an intersection. Fig. 2 further shows the actual stop line position 219, as well as the assumed stop line position 217, which was used to determine the simulated traffic situation 114. Fig. 2 shows that the simulated traffic situation 114 deviates from the measured traffic situation 115, which is typically reflected in a relatively high value for the comparison measure 116. The Fig. The method 100 described in Figure 1 results in the simulated traffic situation 114 being adapted or approximated to the measured traffic situation 115 (by changing the assumptions 117, 217 regarding the attributes, i.e., by changing the assumed values ​​117, 217). This then leads to the assumed stop line position 217 also approaching the actual stop line position 219, so that a precise estimated value 118, 217 for the actual stop line position 219 can be determined.

[0042] Fig. Figure 3 shows a curve of an exemplary comparison measure 116 depending on the assumptions 117 regarding the attributes of the node to be determined. In the example shown, the comparison measure 116 has an (absolute) minimum. The assumptions 117 regarding the attributes to be determined at the minimum of the curve of the comparison measure 116 result in the estimated values ​​118 of the attributes to be determined.

[0043] In other words, the method 100 comprises the acquisition of measurement data or trajectory data 112 (e.g., trajectories) in a plurality of vehicles. The measurement data 112 can include, in particular, the vehicle position, the vehicle orientation, and the longitudinal speed (possibly as a function of time). The vehicle (or an external server) can be configured to extract characteristic variables of the intersection, e.g., the distance from stopping positions to the intersection center 201, the duration of stopping positions, and / or the direction of entry into the intersection. The extracted characteristic variables of the intersection can then be transmitted to a backend server. By extracting the characteristic variables from the measurement data 112 in the vehicle, the required transmission capacity between the vehicle and the backend server can be reduced if necessary.

[0044] Measured features 115 of a traffic situation at the intersection can then be determined from the transmitted characteristic variables (step 103). The measured features 115 can, in particular, describe the traffic behavior of the vehicles in relation to the parameters or attributes to be estimated. The measured features 115 can, for example, include the distribution density of the stopping positions at intersection entrances. The distribution density can be approximated using a histogram or a kernel density estimator. Such an approximation typically does not require any model knowledge, such as modeling using a Gaussian bell curve.

[0045] The method 100 further comprises simulating 102 the traffic at the respective intersection (e.g., at an intersection). A discrete value range can be defined for each of the parameters or attributes to be estimated, with the combination of all possible parameter values ​​resulting in a discrete parameter space. A simulation can be performed iteratively for each individual parameter combination 117. As a result, simulated features 114 of the traffic situation at the intersection are available for each individual parameter combination. The simulated features 114 can, in particular, include the distribution density of the stopping positions of the vehicles.

[0046] The method 100 further comprises comparing 104 the measured features 115 and the simulated features 114 of the traffic situation. In particular, the approximated distribution density from the real data can be compared with the distribution densities from the individual simulations. Statistical methods, in particular Match Distance, Bhattcharyya, Bhattcharyya2, Histogram Intersection and corresponding comparison measures 116 can be used for the comparison. For each individual simulation with a specific parameter combination 117, a comparison measure 116 is thus obtained. For the example of a one-dimensional parameter space (e.g., stop line position or absolute traffic density), exemplary comparison measures 116 result according to Fig. 3. The parameter combination 117 can result from the simulation as an estimated value 118 for the parameters, for which the minimum comparison measure 116 is determined.

[0047] To perform the simulation, an intersection model or a junction model can be created. This model can include fixed model parameters 113, which can be extracted directly from the measurement data 112 and / or from other sources 111. Examples include relative traffic densities at the intersection entrances and maneuver probabilities depending on the time of day. The geometry and topology of the junction can be extracted from available map material 111. These fixed model parameters 113 are not changed between iterations and are therefore constant parameters. In addition, the model contains the parameters to be estimated, which are changed according to the discrete parameter space between iterations and therefore correspond to variable parameters.

[0048] An advantage of the method 100 described in this document is the ability to apply the method 100 generically to different parameters or attributes of a node. Furthermore, it is possible to estimate multiple parameters simultaneously in a precise and efficient manner. Furthermore, implementing the method 100 does not require the establishment of heuristics or rules for individual parameters, which could lead to errors in determining the values ​​of the attributes. Furthermore, only relatively little training data is required to implement the method 100.

[0049] The present invention is not limited to the embodiments shown. In particular, it should be noted that the description and figures are intended only to illustrate the principle of the proposed methods, devices, and systems.

Claims

[1] A method (100) for determining a value (117, 118) of a map attribute of a node between a plurality of traffic routes for a digital map (111); wherein the map attribute is a property of the node; wherein the method (100) comprises - determining (102) a simulated traffic situation (114) at the intersection for an assumed value (117) of the attribute; - Determining (103) a measured traffic situation (115) at the intersection based on a plurality of measured trajectory data (112) of vehicles at the intersection; wherein the trajectory data (112) of a vehicle comprise a trajectory of the vehicle when crossing the intersection; - comparing (104) the measured traffic situation (115) with the simulated traffic situation (114); - determining (105) the value (118) of the attribute so that the simulated traffic situation (114) approximates the measured traffic situation (115); and - Providing the value (118) of the attribute as an update of the digital map (111) on the vehicles for use in various driver assistance systems. [2] Method (100) according to claim 1, wherein the method (100) comprises - determining (102) a plurality of simulated traffic situations (114) at the intersection for a corresponding plurality of assumed values ​​(117) of the attribute; - determining (104) a plurality of comparison measures (116) between the measured traffic situation (115) and the corresponding plurality of simulated traffic situations (114); and - determining (105) the value (118) of the attribute from the plurality of assumed values ​​(117) of the attribute that corresponds to a relatively improved comparison measure (116) from the plurality of comparison measures (116). [3] Method (100) according to claim 2, wherein the comparison measure (116) comprises an indication of a difference between the measured traffic situation (115) and the simulated traffic situations (114). [4] Method (100) according to any preceding claim, wherein - the method (100) comprises determining (101) one or more model parameters of a simulation model for the intersection point on the basis of the plurality of measured trajectory data (112) and / or on the basis of the digital map (111); - the simulation model includes the assumed value (117) of the attribute as a further model parameter; and - the simulated traffic situation (114) is determined on the basis of the simulation model. [5] The method (100) of claim 4, wherein the one or more model parameters comprise one or more of: a geometry of the intersection, a relative traffic density on the plurality of traffic lanes of the intersection, and / or a maneuver probability for possible maneuvers of a vehicle at the intersection. [6] Method (100) according to any preceding claim, wherein the simulated traffic situation (114) is determined by means of a microscopic traffic simulation. [7] Method (100) according to any preceding claim, wherein the simulated traffic situation (114) and the measured traffic situation (115) comprise a traffic behavior of the vehicles at the intersection with respect to the intersection attribute to be determined. [8] Method (100) according to any preceding claim, wherein - the attribute comprises a position of a stop line of a first traffic route at the intersection; and - the simulated traffic situation (114) and the measured traffic situation (115) comprise a distribution of stopping positions of the vehicles on the first traffic route. [9] Method (100) according to any preceding claim, wherein - the method (100) serves to determine values ​​(117, 118) of a plurality of attributes of the node; - the method (100) comprises determining (102) the simulated traffic situation (114) at the intersection for assumed values ​​(117) for the plurality of attributes, and - the method (100) comprises determining (105) the values ​​(118) of the plurality of attributes so that the simulated traffic situation (114) approximates the measured traffic situation (115). [10] Method (100) according to any preceding claim, wherein - the intersection includes a level crossing or a roundabout; and - the plurality of traffic routes comprises a corresponding plurality of roads which are brought together by the junction.

Images