Assignment of traffic lights to corresponding lanes

Abstract

Method for assigning traffic lights (a1, ..., a4) and traffic light groups (g1, g2) to the corresponding lanes (S1, S2, S3) of a roadway for a given approach direction from swarm data of a large number of ego vehicles (EF) in a backend server (S) with the steps: - Receiving consolidated data per traffic light crossing of the vehicles of a swarm for the roadway in the corresponding direction of travel, wherein the consolidated data for each intersection crossing contains the traffic light data recorded by each vehicle, namely positions and phase profiles of the traffic light objects, vehicle-related data and environmental data, - Identifying the traffic lights and traffic light groups from the received traffic light objects of a given number m, m>1, of traffic light crossings for the combination of roadway and approach direction and determining the number n of the real traffic light groups, - Adding a dummy traffic light group (g3) to the identified real traffic light groups (g1, g2), - Assigning the Traffic light groups (g1, g2,g3) to the lanes (S1, S2, S3) of the roadway by means of an iteration based on a Bayes filter, in which, for the given number of traffic light crossings for a given lane (S1, S2, S3), an a posteriori probability distribution is calculated from an a priori probability distribution for each traffic light crossing, wherein the a posteriori probability distribution of the (m-1)th step forms the a priori probability distribution of the mth step, where: - the a priori probability distribution of the 0th step for the real traffic light groups (g1, g2) is initialized with the value 2 / 2n+1, where n is the number of real traffic light groups (g1, g2), - the a priori probability distribution of the 0th step for the dummy traffic light group (g3) is initialized with the value 1 / 2n+1, where n is the number of real Traffic light groups (g1, g2) are.

Description

The invention relates to a method for assigning traffic lights to the associated lanes of a roadway from swarm data of a plurality of vehicles and to a use of the method in a vehicle.When driving a motor vehicle autonomously or automatically, it must cope with complex traffic light situations at intersections. For this purpose, automatic recognition of the traffic lights present at an intersection is necessary, wherein this recognition must also comprise an assignment of the traffic lights to the respective lanes for which the signals of the traffic lights are responsible. However, such automatic recognition of traffic lights at intersections is applicable not only to autonomous driving but can also be applied to assisted or semi-autonomous driving of a motor vehicle to assist the driver in driving the motor vehicle.Automatic recognition of traffic lights at an intersection in automatic driving follows the following goals:recommending a possible speed to the traffic light system to the driver,- automatically to stop before the traffic light,warning the driver in the event of an accidental driving over of the traffic light, andgiving the driver a continuing notice when the traffic light is going green again.To implement these examples, attempts are made to check the traffic light with the front camera in the vehicle. The following measurement task must be accomplished in this case:The camera recognizes the traffic light in advance and measures it with regard to position and alignment.The camera recognizes the lanes and markings on the ground and thus detects an assignment of the traffic light to the lanes.The camera constantly monitors the state of the relevant traffic light and passes on this information to the corresponding function.In reality, however, the assignment of traffic light to driving lane is very difficult to determine due to a wide variety of and complex intersection situations, which leads to high incorrect rates in the assignment of traffic light to driving lane.The document DE 10 2014 205 953 A1 relates to a method for analyzing a traffic environment situation of a vehicle having a navigation system and a sensor system, wherein the navigation system ascertains a vehicle position and the sensor system acquires movement data of the vehicle and ascertains a movement behavior of the vehicle in the form of movement trajectories of the vehicle, wherein the vehicle position and the movement trajectories of the vehicle make it possible to draw conclusions about the presence and / or the type of at least one feature of the environment, wherein a sensor of the sensor system acquires the feature and stores at least one additional information about the feature.The document DE 10 2018 210 125 A1 describes a method for assigning traffic lights of a traffic light of a roadway section controlled by this traffic light to the lanes of the roadway section, wherein the roadway section has at least one lane, the lanes of the roadway section have a common direction of travel, and a plurality of vehicles travel on the roadway section, comprising the steps:detecting a traffic light system ahead by each vehicle traveling the roadway section and moving toward the traffic light system,continuously detecting the traffic light states of the traffic lights of the traffic light, the driving lane, the vehicle speed and the vehicle position by each vehicle, wherein the detection starts at a predetermined distance from the traffic light and is continued until passing the traffic light,transmitting the data recorded by each vehicle "traffic light states", "driving lane" and "speed" in the form of a data record to a backend computer,performing the following evaluations by the backend computer for each vehicle-specific data record:determining the detected traffic light system on a map,detecting the movement of the vehicle through the roadway section,detecting the traffic light conditions when the traffic light system passes through the vehicle, andcreating an assignment table for the detected traffic light system from the evaluated data records, wherein the assignment table assigns the traffic lights of the traffic light system to the lanes of the roadway section.A prerequisite for the creation of an assignment of the traffic lights of a traffic light installation to the lanes is the exact relative localization of the ego vehicles in global coordinates. As a global coordinate system, for example, the semi-global ENU coordinate system can be used, wherein ENU stands for east-north-up, which is valid for a range of several hundred meters around a reference point. This can be achieved on the one hand by a highly accurate localization based on odometry data, GPS data and optionally further visual data. If this is not the case, then on the other hand an ad hoc localization can be carried out based on subsequent matching of odometry data, GPS data and traffic light data of the respective ego vehicle.From the exact localization of the ego vehicles, the positions of the traffic light objects detected by the ego vehicles when entering the traffic lights can then be determined.Now, with known methods such as the method described above, some traffic lights may not be reliably detected, for example turn lights. In addition, traffic signs such as the static right turn arrow are not considered. Since this information is not present, an incorrect traffic light can be assigned to lane. This is particularly problematic when a single traffic light group is left. In this case, the mapping algorithm will necessarily map each lane of this traffic light group and this condition will also remain undetected.The object of the invention is therefore to specify a method for improved assignment of traffic lights to lanes in order to avoid incorrect traffic light to lane assignments.This object is achieved by a method having the features of claim 1 and by the use of the method having the features of claim 10. Preferred embodiments of the invention are the subject matter of the dependent claims.The method according to the invention for assigning traffic lights and traffic light groups to the associated lanes of a roadway for a predefined direction of approach from swarm data of a plurality of ego vehicles in a backend server comprises the steps:receiving consolidated data per traffic light crossing of the vehicles of a swarm for the roadway in a corresponding direction of travel, wherein the consolidated data for each crossing contains the traffic light data recorded by each vehicle, namely positions and phase profiles of the traffic light objects, vehicle-related data and environment data,identifying the traffic lights and traffic light groups from the received traffic light objects of a predefined number m, m>1 of traffic light passes for the combination of roadway and direction of approach and determining the number n of real traffic light groups,adding a dummy traffic light group to the identified real traffic light groups,assigning the traffic light groups to the lanes of the roadway by means of an iteration based on a Bayesian filter, in that a posterior probability distribution is calculated from each traffic light transit for the predefined number of traffic light passes for a predefined lane, wherein the posterior probability distribution of the (m-1)th step forms the a priori probability distribution of the mth step, whereinthe a priori probability distribution of the 0th step for the real traffic light groups is initialized with the value 2 / 2n+1, where n is the number of real traffic light groups,the a priori probability distribution of the 0th step for the dummy traffic light group is initialized with the value 1 / 2n+1, where n is the number of real traffic light groups.The Bayes iteration using a dummy traffic light group on the one hand enables an assignment of traffic light groups to lanes and on the other hand enables an erroneous assignment to be detected.The vehicle-related data preferably comprise at least the movement state and movement direction of the ego vehicle, the orientation vector of the vehicle trajectory, the traffic light line and the stop line on the ego trajectory and also the direction of turn and ego events, wherein ego events comprise the driving over event which corresponds to the point in time at which the ego vehicle drives over the stop line and the standstill event which corresponds to the point in time at which the ego vehicle reaches a standstill.Further preferably, the posterior probability of the Bayesian filter is proportional to a likelihood function multiplied by the a priori probability, wherein the likelihood function is a function of the state of movement of the ego vehicle at the time of the ego event, a function of the position of the ego vehicle relative to the stop line at the time of the ego event and a function of the existence or the state of movement of the preceding vehicle at the time of the ego event.By means of the measured variables of the state of motion of the ego vehicle and the position of the ego vehicle at the time of an ego event, and of the state of motion of a vehicle driving ahead, if present, the likelihood function of this transition can be determined in a simple manner.Preferably, the traffic lights are derived from the transit-specific traffic light objects, whereina total point cloud is created from the traffic light object point clouds after pre-filtering of individual traffic light object points in the longitudinal direction along the roadway course has taken place and the individual traffic light object points are projected in the transverse direction to the roadway course, wherein the height of the traffic light position points is maintained, andclustering of the traffic light object points of the total point cloud is carried out in order to obtain the traffic lights.Preferably, the traffic light groups are determined based on the phase profile of the traffic lights.Further preferably, for grouping the traffic lights into traffic light groups, the correlation between the phase profiles of individual pairs of traffic light objects is calculated during the same pass and, based on the pair-wise correlations, the individual traffic light object phase profiles are divided into traffic light groups by means of a hierarchical clustering method.A traffic light group is further preferably assigned to a lane if the a priori distribution calculated by the Bayesian filter converges to a high probability via the number of passes. In this context, high probability means a probability close to 1, i.e. P≈1.A low probability conversely means that the lane under consideration cannot be assigned to a traffic light group.Further preferably, the assignments of the traffic light groups to the lanes of a roadway as a function of the direction of approach and intersection identification are stored in a database, which can be called up by autonomous or semi-autonomous vehicles. In this way, the retrieving vehicle can always access current data of a predetermined intersection.Further preferably, the traffic lights and the traffic light groups, in particular an intersection, are checked if the assignments of the traffic light groups to the lanes are not unambiguous. Here, it does not clearly mean that the assignment probability does not converge to one or zero, but rather fluctuates between these values. In this way, errors in the assignment can be detected.When using the method explained above for assigning traffic lights and traffic light groups to the associated lanes of a roadway for a predefined direction of approach from swarm data of a plurality of ego vehicles in a backend server in an autonomous or semi-autonomous vehicle, the vehicle retrieves the assignment from an external database when approaching a traffic light-controlled intersection.A preferred embodiment of the invention is explained below with reference to the drawings. This shows FIG. 1 shows the acquisition of swarm data and the processing thereof for the assignment of traffic lights to driving lanes, FIG. 2 shows clustering of the traffic light objects for determining traffic light positions, FIG. 3 is a graphical representation of the Bayes filter, FIG. 4 shows a real example for assigning traffic lights to lanes, FIG. 5 shows the traffic light groups derived from FIG. 4, FIG. 6 shows the probability that lane 1 is assigned to the first traffic light group, FIG. 7 shows the probability that lane 2 is assigned to the first traffic light group, FIG. 8 shows the probability that lane 3 is assigned to the second traffic light group, FIG. 9 shows a further example of a part of a fictitious intersection, FIG. 10 shows the lane to the traffic light assignment of lane 1 to the intersection of FIG. 9, FIG. 11 shows the assignment result for only one detected traffic light and without the dummy traffic light group of the example of FIG. 9, andFIG. 12 shows a non-convergent traffic light to lane assignment in the case of an erroneous traffic light detection for the case of FIG. 9.FIG. 1 shows the schematic division of the data collection and processing for assigning traffic lights to lanes from swarm data on which the method is based.On the vehicle side, the detection of the traffic lights A in the vehicle area F is started every time the intersection K passes. In this case, the necessary signals, namely movement of the ego vehicle EF, traffic lights A and objects (not shown) in the environment of the ego vehicle EF, are recorded by means of a camera (not shown), and a first processing DE is carried out in the ego vehicle EF. Subsequently, the data record resulting from the first processing DE is transmitted to the backend server S by means of a transmission link Ü, for example a mobile radio connection.The individual steps in the vehicle area F are:detecting the traffic light signals,tracking and ego compensation of the detected traffic lights per pass,filtering invalid traffic light objects,detecting relevant "ego events", namely the driving event, which corresponds to the point in time at which the ego vehicle drives over the stop line, and the standstill event, which corresponds to the point in time at which the ego vehicle reaches a standstill, andtransmitting the consolidated data to the server, namely at leasttraffic light data, namely positions and phase profiles of a traffic light object,vehicle-related data such as the state of movement and direction of movement of the vehicle, orientation vector of the vehicle trajectory, traffic light line and stop line on the ego trajectory and also the direction of turn, ego events, andenvironment data such as pre-driving vehicles including their speed.In order that lanes can be assigned to traffic lights in the area of the backend server S, the position data of the traffic light data transmitted to the backend server S are expressed in coordinates (x, y, z) of a global coordinate system, for example in the ENU system (east-north-up) already mentioned.The backend server area S receives the described data from the vehicle area F from the crossing of a plurality of vehicles and processes them in the data processing and assignment DV, i.e. the external backend server, as follows:matching and clustering the collected traffic light data in order to generate a traffic light image per intersection and direction of approach,grouping the resulting traffic lights into groups having a common phase characteristic,assigning the resulting traffic light groups to lanes, andentry of the results into a database T and / or a digital map K.The function of assigning traffic lights to traffic lanes from swarm data must work with traffic lights A at different levels. In order to avoid confusion, the following terms are used: Traffic light object: A traffic light object supplied (tracked) by the detection system, for example by a corresponding camera. Erroneous detection may result in multiple traffic light objects being generated for the same physical traffic light signal during the same pass. Traffic light: These are representative of a physical traffic light A at a given intersection K and direction of approach and is generated by clustering a plurality of traffic light objects. Traffic light group: group of traffic lights which have the same phase behavior and thus correspond to a single logical traffic light.In order to be able to carry out the data processing of interest here in the backend server S, the backend server S therefore collects data of different vehicles with respect to their intersection crossings and firstly arranges these in a corresponding table after the intersection identification and the direction of approach.The backend server area S thus has two main tasks:Matching and clustering the traffic light positions of the traffic light objects provided here from a plurality of passes in order to derive traffic lights and traffic light groups therefrom. This step can normally be performed when there are data of at least 20 passes, regardless of the lane being driven, and should be repeated with new data at regular intervals to cover the case where structural or programming changes have been made at the intersection.The assignment of lanes to traffic light groups. This online procedure may begin as soon as the traffic light positions and traffic light objects are fixed. With each new data set from individual passes, the assignment can be corrected accordingly.During processing in the backend server area S, a distinction must be made between data which originate from different passes. For this purpose, the relevant variables, in particular positions of the traffic light objects and the traffic light lines, are provided with an additional index c, c=1... C, with C number of the transmitted transit datasets. Thus, the expression p nc would indicate the n c- th traffic light object of the cth pass.FIG. 2 shows the clustering of the traffic light objects transmitted from the ego vehicles to the backend computer S for determining traffic light positions. A total point cloud is created from the corrected, drive-over-specific traffic light object point clouds, which total point cloud contains all the traffic light positions previously seen by the vehicles. In this case, as shown in FIG. 2, the traffic light positions of a predefined direction are selected on the basis of the ego trajectories transmitted by the vehicles, the traffic light objects of different lanes being shown differently in the illustration. Thus, in the example of FIG. 2, the sign "+" means traffic light objects of a first lane and the sign "x" means traffic light objects of a second lane.The traffic light objects transmitted to the server by the plurality of vehicles with respect to crossing of the intersection and their positions have a relatively high level of variation in the longitudinal direction, that is to say along the course of the road, so that filtering of the individual traffic light object points takes place in the longitudinal direction and traffic light object points which do not pass a predefined condition in the longitudinal direction are discarded. One possible condition would be that traffic light objects that are more longitudinally than two standard deviations from the median across all points are discarded.The individual traffic light object points are projected onto the transverse direction Q, wherein the height H is directly adopted, as is illustrated in FIG. 2. The traffic light object points are clustered using the known DBSCAN algorithm to derive individual "physical" traffic lights. Said algorithm forms clusters of so-called "density-connected" points, i.e. points which are not further than a predetermined distance from a core point in the same cluster. In the present case, the predefined distance is 2 m. The discrete traffic lights A 1, A 2, A 3 and A 4 are thus obtained by means of the DBSCAN algorithm.The next step is now to group the identified traffic lights into groups with the same phase profile in order to be able to ascertain traffic light groups. FIG. 2 shows the traffic light object points of two lanes. The phase profiles show that the physical traffic lights A 2 and A 4 belong to a traffic light group AG 1 and the traffic lights A 1 and A 3 belong to a second traffic light group AG 2. This assignment of traffic lights to traffic light groups is basically carried out in two steps:The basis for the classification of the traffic lights into traffic light groups is the correlation between the phase profiles of individual pairs of traffic light objects that were detected during the same pass. For each such pair of traffic light objects, both the total time at which both phase characteristics temporally overlap and the matching time (matching time) in which the two phase characteristics assume the same phase value are determined. The quotient of the two values corresponds to the correlation between the two phase curves.Based on such pairwise correlations, the individual phase profiles of the traffic light objects can be assigned to traffic light groups by means of the hierarchical clustering method. This hierarchical clustering algorithm forms groups in which each group element, i.e. traffic light, has a minimum correlation with all other elements of the group. This assignment is carried out iteratively by expanding the existing clusters or groups in successive steps by new elements which have a small correlation with all elements already in clusters.FIG. 3 shows a schematic illustration of the iteration algorithm with which it is now possible to perform a lane-to-traffic light group assignment on the basis of the present information and the traffic lights or traffic light groups derived therefrom. This is done via an online algorithm, into which the individual drive-over events are fed in sequence, so that the algorithm converges over the time profile to a fixed probability, which indicates that a given lane is assigned to a given traffic light group.Assume that the ego vehicle has traveled in ego at a given crossing on lane s. The calculated probability that a certain traffic light group g* is to be assigned to this lane isIn other words, a random variable G sego is defined which describes the assignment of the ego lane s ego to a traffic light group G, and the above expression describes the probability that the traffic light group G associated with the ego lane s ego is the traffic light group g*, where g* ∈ g 1,... g n.The determination of the assignment of the traffic light groups G to the ego lane s ego is based on a Bayesian filter which is based on a model with which it is possible to determine the probability that a certain set of variables D can assume given values d if G sego assumes a given value g. The model therefore takes the following form:This function is also referred to in this context as likelihood function L(gld).For each update step, which in the present case corresponds to a given ego event, the variables are measured in D, wherein it is assumed here that these assume the values d*. From the previously determined probability P(G sego= g*), also referred to in this context as a priori probability, an updated or a posterior probability can be calculated via the set of bayes:In other words, the a posterior probability is proportional to the likelihood function multiplied by the a priori probability. The denominator in the above-mentioned Bayes formula corresponds to a normalization factor.It is further assumed that the variable D depends not only on g, but also on an additional set of further variables M, which can be measured for each update step corresponding to an ego event, wherein the measured value of a variable M is given by m*.In the present case, the following variables are used, which are all measured in the ego vehicles and transmitted to the server:- Dependent variable D:V ego is the state of motion of the ego vehicle at the time of the event, wherein the variable can assume one of the following values:v holds ego= andv ego= travels.- Independent variables M:X ego is the position of the ego vehicle relative to the stop line at the time of the event with the following values:x is ego= at the stop line, andx ego= before the stop line.v obj is the existence or the state of movement of the preceding vehicle at the time of the event with the following values:v obj= no preceding vehiclev obj= holds the preceding vehicle, andv obj= the preceding vehicle is traveling.Thus, the following likelihood function is obtained:This situation is schematically illustrated in FIG. 3 as an iteration method. In the first step, the a priori probability is determined, namely:Based on this a priori probability, in the next step, the likelihood function is determined including the independent variable M, namely the position of the ego vehicle relative to the stop line at the time of the event and the existence or the movement state of the preceding vehicle at the time of the event: from which the a posterior probability results. In other words, in each update step, the current a priori assignment probability is corrected via the likelihood function, including the measured variables, in order to generate an updated a posterior assignment probability. This again serves as a priori probability in the subsequent step, wherein the iteration is indicated by the dashed line in FIG. 3.It can now happen that one or more traffic lights have not been correctly captured by the camera system used or that they have depended on the formation of the traffic light groups, matching or clustering. If a complete traffic light group fails, as it were, the assignment algorithm which assigns the traffic light groups to the lanes will most likely make an incorrect assignment. This is particularly problematic if only a single traffic light group remains when the traffic light groups are formed. The assignment algorithm will then necessarily assign each lane to this traffic light group, wherein this erroneous assignment remains undetected, which can lead to a risk potential.A similar related problem is presented by turn lights. These can be detected only unreliable and, in particular, the traffic light phase information cannot be interpreted.To make both problems manageable, an additional "dummy" traffic light group is introduced into the method. Thus, the total number of traffic light groups used in the Bayes algorithm is n+1 and for each traffic light group g*, where traffic light group g n+1 is the dummy traffic light group.Thus, the assignment algorithm always has the choice between at least two traffic light groups and it is to be expected that in cases where correct assignment is not possible, the algorithm will fluctuate between the dummy traffic light group and the other real traffic light groups without converting to one. This opens up the possibility of looking at the intersection in a targeted manner.In addition to the n recognized traffic light groups g 1 to g n a further traffic light group g n+1 is added. Thus, if conflicting observations occur that cannot be mapped by n traffic light groups, they can be mapped to the n+1thgroup. The aim is not primarily to get a correct traffic light to track allocation, but rather to detect an erroneous allocation and, if appropriate, to intervene manually.The aim of the algorithm is now to calculate the probability for each of the n real traffic light groups g*, g* ∈ g 1,... g n, that the currently traveled ego lane s ego is assigned to this traffic light group or not. For the dummy traffic light group g*=g n+1 the assignment probability must therefore converge towards zero after a few passes in the specific case.Before the ego lane s ego has been traveled over at all, all assignment probabilities including those of the dummy traffic light are initialized, so that the initial a priori distribution of the Bayes iteration is defined as follows:In other words, the initial a priori probability is evenly distributed for all real traffic light groups g 1,... g n and the above-mentioned function of the number n of traffic light groups and for the dummy traffic light group g n+1 the a priori probability is half the initial a priori probability of the real traffic light groups.After each crossing of the intersection on an ego lane ego the ego likelihood function is thus calculated and the probability is updated via the Bayer update in order to obtain the above-indicated posterior distribution. This a posterior distribution is used as a priori distribution in the next step, i.e. during the next pass of an ego vehicle on the same intersection and lane. Over time, the a priori distributions should convert to the correct probabilities, i.e. high and in principle converging towards one probability for the traffic light group under consideration, low and converging towards zero probabilities for all others.FIG. 4 shows a real intersection situation in Wolfsburg, namely the intersection Lessing street and Goethe street with the entry direction South. The three lanes S 1, S 2 and S 3 can be seen from right to left, wherein the right lane S 1 forms a straight lane and turn lane to the right, the middle lane S 2 forms a pure straight lane and the left lane S 3 forms a pure turn lane to the left. The traffic behavior of this direction of approach is controlled by two traffic light groups g 1 and g 2, the first traffic light group g 1 comprising the traffic lights a 1 and a 3 and the second traffic light group g 2 consisting of the traffic lights a 2 and a 4. In this case, the traffic light group g 1 controls the lanes S 1 and S 2, and the traffic light group g 2 is responsible for the left turn lane S 3.FIG. 5 shows the traffic light image resulting for the intersection with the direction of approach South shown in FIG. 4, this traffic light image being generated in the server on the basis of a plurality of passes in a manner analogous to the method described on the basis of FIG. 2. The traffic light image with the traffic lights a 1, a 2, a 3 and a 4 is shown in a coordinate system with the transverse direction Q and the height H in metric units, wherein the traffic lights a 1 and a 3 belong to a first traffic light group g 1 and the traffic lights a 2 and a 4 belong to a second traffic light group g 2, as follows from the analysis of the phase characteristics measured at the respective drive-over event.FIG. 6 shows the result of the above-explained Bayesian iteration for the right lane S 1 of FIG. 4 : The probability P(g1, S 1) that the traffic light group g 1 controls the right lane S 1 is shown as compared to the number N of traffic light passes of the ego vehicles in this lane. The starting point is the initialization of the method with the probability 0.4 for the 0th driveover, since there are two real traffic light groups g1and g2and consequently the initial driveover probability is 2 / 5=0.4. It can be seen that the method converges with the probability P(g1, S1) of one from approximately eight passes that the lane S1 is controlled by the traffic light group g1.FIG. 7 shows the result for the lane S 2 of FIG. 4 Here, too, it can be seen that, when the method is initialized with the probability of 0.4 with approximately six passes, the result is fixed, namely that the traffic light group g 1 is responsible for the lane S 2 with a probability P(g1, S 2) of approximately one.FIG. 8 also shows the course of the probability P(g2, S3) as a function of the passes N that the traffic light group g2 is responsible for the left-turn lane S3 starting with the initialization of the probability P(g2, S3) at 0.4th latest in the fourth pass, the Bayesian method converges to the probability value P(g2, S3) of one. Consequently, it is found out within a few passes that the traffic light group g 2 checks the left turn lane S 3.FIG. 9 shows a schematic illustration of part of an intersection K, wherein FIG. 9 shows two lanes S 1 and S 2 from right to left in the direction of travel of the intersection K, namely a turn lane S 1 to the right and a lane S 2 leading straight ahead. The right lane S 1 is controlled by a traffic light a 1. Furthermore, instead of the traffic light a 1, a turn arrow ap may be responsible for the right turn lane, which allows a right turn. In the case of the turn signal a 1, this is identical to the traffic light group g 1 necessary for the Bayesian iteration, since in the example it is only a traffic light. The straight-ahead left lane S 2 is checked in the example by a traffic light a 2, which is likewise identical to its traffic light group g 2.FIG. 10 now shows the result of the association of lane with the traffic light group for the right turn lane S 1, wherein the traffic light group g 1 is realized by the traffic light a 1 and the traffic light group g 2 is realized by the traffic light a 2. Furthermore, in the iteration method, a virtual traffic light group g 3 is used as a dummy traffic light group. The probability P is shown with respect to the number of passes N.The upper curve shows the probability P(g1, S1) that the lane S1 of FIG. 9 is controlled by the traffic light group g1. It is obvious that even after seven passes this probability P(g1, S1) converges to the value one, the initial probability of the 0th pass being 0.4 according to the above definition, since two real traffic light groups exist.The middle curve of FIG. 10 shows the probability P(g2, S1), determined as a function of the passes N, that the traffic light group g2 controls the right lane S1, which is obviously not the case in the example. Even after six passes, the probability converges to zero. This curve also begins with the initial probability of 0.4.The same applies to the probability P(g3, S1) that the virtual traffic light group g3 would have to do anything with the lane S1. Here again, the method for the dummy traffic light group g 3 converges quickly to zero starting from the initial value 0.2=1 / 5 according to the above definition.It can also be seen from FIG. 10 that the following applies: since the sum of all possible probabilities of the assignments of traffic light groups to the lane under consideration must always result in one.FIG. 11 shows the traffic light lane allocation at the intersection K of FIG. 9, assuming that either the first traffic light group g 1 controlling the right turn on the lane S 1 has not been recognized, for example, because the traffic light a 1 is out of operation or defective, or there is a case that, instead of the traffic light a 1, a right turn arrow ap controls the traffic on the right lane S 1 that has not been recognized. In both cases, therefore, only the traffic light group g 2 which is identical to the traffic light a 2 is recognized.In other words, for this scenario, there is therefore only one real traffic light group g 2, realized by the traffic light a 2 of FIG. 9.FIG. 11 now shows the case of a Bayes iteration without a dummy traffic light group. In the example of the defective traffic light a 1 or the unrecognized turning arrow ap, the number n of recognized real traffic light groups is n=1 since there is only one real traffic light group g 2. Consequently, this traffic light group g2 is assigned to the right turn lane S1, which is represented by the assignment probability P(g2, S1), which assumes the value 1 for all passes. In this case, the initial assignment probability for the 0th driveover also has the value 1, since the sum of all probabilities for a lane must always result in 1. It is therefore not recognizable that at this point somewhat wrong has been run.FIG. 12 now shows a Bayesian iteration including a fictitious dummy traffic light group in the case of the example of a defective traffic light a 1 or an unrecognized turning arrow ap in the situation of FIG. 9 ; if a dummy traffic light group is added to the method, denoted here by g 3, both the assignment probability P(g 1, S 1) and the assignment probability P(g 3, S 1) can be determined for the lane S 1, wherein the following must apply:It can be seen that the two assignments do not converge towards one or zero, but rather oscillate between the said values one and zero as a function of the number of passes N. In other words, since the method does not converge with the dummy traffic light group g 3 to a clear statement as to which traffic light group is to be assigned to which lane, the state of the intersection is, so to speak, indeterminate. In this case, the method can thus generate a warning and the intersection K can be checked.List of reference charactersF vehicle area EF ego vehicle K intersection A traffic lights DE data acquisition and preprocessing Ü transmission (wireless) S server DV data processing and assignment T table K map H height in m Q transverse direction in m A1 traffic light 1 A2 traffic light 2 A3 traffic light 3 A4 traffic light 4 AG1 traffic light group 1 AG2 traffic light group 2 S1 lane 1 S2 lane 2 S3 lane 3 g2 traffic light group g1 traffic light group a1 traffic light a2 traffic light a3 traffic light a4 traffic light ap turn arrow N number of the drive-overs considered P probability P(g1, S 1) Probability of the association of the traffic light group g 1 with the lane S 1 P(g 1, S 2) Probability of the association of the traffic light group g 1 with the lane S 2 P(g 2, S 3) Probability of the association of the traffic light group g 2 with the lane S 3 P(g 2, S 1) Probability of the association of the traffic light group g 2 with the lane S 1 P(g 3, S 1) Probability of the association of the dummy traffic light group g 3 with the lane S 1

Claims

Method for assigning traffic lights (a1,..., a4) and traffic light groups (g1, g2) to the associated lanes (S1, S2, S3) of a roadway for a predefined direction of approach from swarm data of a plurality of ego vehicles (EF) in a backend server (S), having the steps: - receiving consolidated data per traffic light crossing of the vehicles of a swarm for the roadway in the corresponding direction of travel, wherein the consolidated data for each crossing contains the traffic light data detected by each vehicle, namely positions and phase profiles of the traffic light objects, vehicle-related data and environment data, - identifying the traffic lights and traffic light groups from the received traffic light objects of a predefined number m, m >1, Traffic light passes for the combination of roadway and direction of approach and determining the number n of real traffic light groups, - adding a dummy traffic light group (g3) to the identified real traffic light groups (g1, g2), - assigning the traffic light groups (g1, g2, g3) to the lanes (S1, S2, S3) of the roadway by means of an iteration based on a Bayesian filter, in that an a posterior probability distribution is calculated from each traffic light pass for the predefined number of traffic light passes for a predefined lane (S1, S2, S3), wherein the a posterior probability distribution of the (m-1)th step forms the a priori probability distribution of the mth step, wherein - the a priori probability distribution of the 0th step for the real traffic light groups (g1, g2) is initialized with the value 2 / 2n+1, wherein n is the number of the real traffic light groups (g1, g2), - the a priori probability distribution of the 0th step for the dummy traffic light group (g3) is initialized with the value 1 / 2n+1, wherein n is the number of the real traffic light groups (g1, g2).Method according to Claim 1, characterized in that the vehicle-related data comprise at least the state of movement and direction of movement of the ego vehicle (EF), the orientation vector of the vehicle trajectory, the traffic light line and the stop line on the ego trajectory and also the direction of turn and ego events, wherein ego events comprise the driving over event which corresponds to the point in time at which the ego vehicle drives over the stop line and the standstill event which corresponds to the point in time at which the ego vehicle reaches a standstill.Method according to Claim 1 or 2, characterized in that the posterior probability of the Bayesian filter is proportional to a likelihood function multiplied by the a priori probability, the likelihood function being a function of the state of movement of the ego vehicle (EF) at the time of the ego event, a function of the position of the ego vehicle relative to the stop line at the time of the ego event and a function of the existence or the state of movement of the vehicle driving ahead at the time of the ego event.Method according to one of the preceding claims, characterized in that the traffic lights (a1, a2, a3, a4) are derived from the drive-specific traffic light objects, wherein - a total point cloud is created from the traffic light object point clouds after pre-filtering of individual traffic light object points in the longitudinal direction along the roadway course has taken place and the individual traffic light object points are projected in the transverse direction to the roadway course, wherein the height of the traffic light position points is maintained, and - clustering of the traffic light object points of the total point cloud is carried out in order to obtain the traffic lights (a1, a2, a3, a4).Method according to Claim 4, characterized in that the real traffic light groups (g1, g2) are determined on the basis of the phase profile of the traffic lights (a1, a2, a3, a4).Method according to Claim 5, characterized in that, in order to group the traffic lights (a1, a2, a3, a4) in real traffic light groups (g1, g2), the correlation between the phase profiles of individual pairs of traffic light objects is calculated during the same journey and, based on the pairwise correlations, the individual traffic light object phase profiles are divided into traffic light groups (g1, g2) by means of a hierarchical clustering method.Method according to one of the preceding claims, characterized in that a traffic light group (g1, g2, g3) is assigned to a lane (S1, S2, S3) if the a priori distribution calculated by the Bayesian filter converges to a high probability P ≈ 1 via the number of passes.Method according to one of the preceding claims, characterized in that the assignments of the traffic light groups (g1, g2, g3) to the lanes (S1, S2, S3) of a roadway as a function of the direction of approach and intersection identification are stored in a database which can be called up by autonomous or semi-autonomous vehicles.Method according to one of the preceding claims, characterized in that the traffic lights (a1, a2, a3, a4) and the traffic light groups (g1, g2) are checked if the assignments of the traffic light groups (g1, g2, g3) to the lanes (S1, S2, S3) are not unambiguous.Use of the method for assigning traffic lights (a1, a2, a3, a4) and traffic light groups (g1, g2, g3) to the associated lanes (S1, S2, S3) of a roadway for a predefined direction of approach from swarm data of a plurality of ego vehicles (EF) in a backend server (S) according to one of the preceding claims, in an autonomous or semi-autonomous vehicle, wherein the vehicle retrieves the assignment from an external database when approaching a traffic light-controlled intersection.

Images