Method for determining the lane-change intention of another motor vehicle located in the immediate vicinity of a motor vehicle, as well as an assistance system

Abstract

Method for determining a lane-change intention (3) of another motor vehicle (4) located in the immediate vicinity (7) of a motor vehicle (2) by means of an assistance system (1) of the motor vehicle (2), in which the environment (7) of the motor vehicle (4) is detected by means of at least one environment detection device (5) of the assistance system (1) and an environment model (8) for the environment (7) is generated on the basis of the detected environment (7) by means of an electronic computing device (6) of the assistance system (1), and in which, depending on the generated environment model (8), at least one situation feature (9, 10, 11, 12, 13) characterizing a lane change of the other motor vehicle (4) is determined by means of the electronic computing device (6), as the at least one situation feature (9, 10, 11, 12, 13) characterizing the lane change of the other motor vehicle (4) is determined.13) a lane-changing motive (9) is determined for the other motor vehicle (4), wherein the lane-changing motive (9) depends on a relative distance (s, rel ) of the further motor vehicle (4) to an object (15) in the vicinity (7), depending on a relative speed (v rel) of the further motor vehicle (4) to the object (15) and depending on an object type, characterized in that the lane change intention (3) is determined by means of a Bayes network model (14) of the electronic computing device (6) depending on the at least one characterizing situation feature (9, 10, 11, 12, 13), wherein as a further situation feature (9, 10, 11, 12, 13) characterizing the lane change of the further motor vehicle (4) a context evaluation (13) is carried out by means of the electronic computing device (6), wherein for this purpose context information for a potential lane change in the environment (7) is taken into account, and wherein traffic sign monitoring is used as context information.

Description

[0001] The invention relates to a method for determining the lane-change intention of another motor vehicle located in the immediate vicinity of a motor vehicle by means of an assistance system of the motor vehicle, in which the environment of the motor vehicle is detected by means of at least one environment detection device of the assistance system and an environment model for the environment is generated by means of an electronic computing device of the assistance system based on the detected environment, and in which, depending on the generated environment model, at least one situation feature characterizing a lane change of the other motor vehicle is determined by means of the electronic computing device, wherein the lane-change intention is determined by means of a Bayesian network model of the electronic computing device as a function of the at least one characterizing situation feature. The invention further relates to an assistance system.

[0002] Especially for the autonomous operation of a motor vehicle, a situation interpretation that is as generic as possible is necessary to autonomously plan appropriate driving maneuvers. This includes the detection of lane changes by other road users, which must be predicted early and reliably. This, in turn, can then be used to perform a corresponding evaluation for subsequent driving functions. The detection of driving maneuvers using object-oriented Bayesian networks in, for example, highway scenarios is already known in the state of the art. Furthermore, the detection of a lane change intention for adaptive cruise control systems is already known.

[0003] DE 10 2014 003 343 A1 relates to a method for determining the need for a lane change by a system vehicle or a vehicle driving ahead of it, whereby a number of feature parameters and a temporal trend development of the feature parameters as well as a probability of a number of hypotheses created from the feature parameters are determined on the basis of an environment representation of a current vehicle environment.The plan is to determine the probability of an actual value of the respective feature parameters using a continuous Gaussian distribution function and the temporal trend development of the respective feature parameter, and to determine the respective probabilities using hypotheses based on a sigmoid function, whereby at least one of the feature parameters is determined as a feature parameter of a relative dynamic between the system vehicle or a vehicle driving ahead of it and another vehicle or object located in the vehicle environment.

[0004] DE 10 2010 020 047 A1 concerns a method for adapting a target distance specified for a vehicle's distance control system to a current traffic situation.

[0005] DE 10 2004 027 983 A1 relates to a method and a device for detecting lane-changing maneuvers for a vehicle. In this method, at least one observation parameter is determined that describes the lane-changing behavior of an observed vehicle, and a lane-changing parameter is determined based on this at least one observation parameter, which characterizes the intention of the vehicle to change lanes from a lane assigned to that vehicle.

[0006] The object of the present invention is to create a method and an assistance system by means of which an improved prediction of a lane-changing intention of another motor vehicle can be achieved.

[0007] This problem is solved by a method and an assistance system according to the independent claims. Advantageous embodiments are specified in the dependent claims.

[0008] One aspect of the invention relates to a method for determining the lane-change intention of another motor vehicle located in the immediate vicinity of a motor vehicle by means of an assistance system of the motor vehicle, in which the environment of the motor vehicle is detected by means of at least one environment detection device of the assistance system and an environment model for the environment is generated on the basis of the detected environment by means of an electronic computing device of the assistance system, and in which, depending on the generated environment model, at least one situation feature characterizing a lane change of the other motor vehicle is determined by means of the electronic computing device, wherein the lane-change intention is determined by means of a Bayesian network model of the electronic computing device depending on the at least one characterizing situation feature.

[0009] It is intended that at least one situational feature characterizing the lane change of the other motor vehicle is a lane change motive for the other motor vehicle, whereby the lane change motive is determined depending on a relative distance of the other motor vehicle to an object in the environment, depending on a relative speed of the motor vehicle to the object and depending on a type of object.

[0010] Thus, a generic model can be created on the basis of which the lane-changing intention of the other vehicle can be reliably determined.

[0011] In particular, a geometric approach, or rather different geometric approaches, are chosen for assessing the situation, making them more powerful than the model-based or numerical methods of the prior art. The task is thus to detect lane-changing maneuvers by other vehicles in the vicinity at an early stage. In a first step, at least one situational feature, or more specifically, a multitude of situational features, is determined, particularly based on feature extraction, which serve as quantified indicators of a possible lane change by the other vehicle. A prerequisite for this is the existing situation description in the environment model. For this purpose, information obtained from various sensor data from different environmental detection devices is processed and made available in a unified environment model.This environmental model can contain, for example, track information, object information, boundaries, or other information. In a second step, a Bayesian network model is used, which is modulated based on various events, such as crossing a track marker, a possible track change, or other track-changing maneuvers. The events are related to each other as random variables using conditional probabilities. This makes it possible to make statements or predictions about events that are not measurable or observable.

[0012] In a final step, the previously determined situational characteristics are fed into the Bayesian network model as "evidence." Evidence refers to observable events. Incorporating evidence is a prerequisite for calculating predictions / probabilities for unobservable events.

[0013] This step therefore makes it possible to determine the probability of the "lane change" event occurring for each other vehicle. This event corresponds to the intention behind the lane change maneuver.

[0014] The lane-change motive, also known as the motivation hypothesis, specifically addresses whether another vehicle intends to change lanes due to reasons such as cooperative behavior, tailgating, speeding, stationary objects, or an active turn signal (e.g., a turn signal). For example, a suitable algorithm can analyze the current scene for tailgaters / speeders and stationary objects. The hypothesis is then calculated based on the relative speed and the distance to the vehicle ahead, which could also be another vehicle. From these two values, the time until the potential collision of the two vehicles is determined. Additionally, information such as the type of vehicle ahead or the other vehicle is taken into account.This can provide a better indication of the motivation hypothesis, especially in traffic scenarios with high traffic density. This is because, in such cases, the vehicles are closer together and have lower relative speeds.

[0015] To determine the reason for a lane change, the first step involves checking whether there is a vehicle ahead in the same lane as the vehicle in question. If so, the reason for the lane change can be determined. The second step involves determining the distance s. relThe distance between the two vehicles is determined. This distance is defined, for example, by the distance from the front axle of the rear vehicle to the rear axle of the vehicle in front. To ensure accurate distance measurements even on large curves, the center line of the lane can be used to determine the distance between the two objects. This takes the road geometry into account when determining the distance. A third characteristic for the lane-change motive is the consideration of speed v. rel between the objects. Here, the absolute speeds of the vehicles from the environment model are used to determine a relative speed of the objects to each other.

[0016] The method is, in particular, a computer-implemented method. Thus, a computer program product is also disclosed, which includes program instructions that cause an electronic computing device to execute the method. Likewise, a computer-readable storage medium is disclosed, which includes the computer program product and is, in particular, implemented on the electronic computing device.

[0017] Furthermore, it is planned that a second situational feature characterizing the lane change of the other vehicle will be used to perform a contextual evaluation using the electronic computing device, taking into account contextual information for a potential lane change in the surroundings. For example, the surroundings may be evaluated with regard to this contextual information. The contextual information indicates, in particular, whether a lane change is even possible. Should this be the case in the current traffic situation, for example, this situational feature will be taken into account, thus enabling an improved determination of the lane change intention.

[0018] Furthermore, traffic sign monitoring or, additionally, lane line monitoring of the roadway in which the vehicle is located is used. In particular, the context analysis can thus also be considered a traffic sign hypothesis. The traffic sign hypothesis makes a statement about whether a vehicle is theoretically permitted to change lanes due to traffic signs, such as a no-overtaking sign. The value of the hypothesis is either 0 in the case of a no-overtaking sign or 1 in the case of no no-overtaking sign, with reference to the current lane in which the other vehicle is located. Using the traffic sign hypothesis, an offset can be added to the overall lane change probabilities, which in some cases leads to a sharp increase or decrease in a shorter period and can thus positively influence the detection horizon.In order to make the traffic sign hypothesis possible, statements about the currently valid traffic signs, such as overtaking prohibitions, are necessary.

[0019] According to an advantageous embodiment, the electronic computing device predicts the collision time between the other vehicle and the object based on the relative speed and distance, and this collision time is taken into account when determining the lane change intention. Specifically, the so-called Time-to-Collision (TTC), which corresponds to the collision time, is determined based on the relative speed and distance to the vehicle ahead. Based on this, the lane change intention can be reliably determined. For example, if the Time-to-Collision is very low, the probability of a lane change is increased. If the Time-to-Collision is high, the probability of a lane change is lower.

[0020] Furthermore, it has proven advantageous to consider the vehicle type of the other vehicle when determining the lane-change motive using the electronic computing device. Specifically, the vehicle type of the other vehicle is taken into account. Additionally, the object type of the vehicle, which can also correspond to the vehicle type, can be considered. The vehicle type attribute can take a value between 0 and 1. A value of 1 means that it is very likely that this vehicle type is the motivation for a lane change. This applies, for example, to trucks on a highway. If a passenger car is driving in front of the vehicle under consideration, the attribute is set to a value of, for example, 0.5. Thus, the corresponding hypotheses for the characterizing situation feature can be refined, allowing for a more accurate determination of the lane-change intention.

[0021] Furthermore, it is advantageous if an error assessment is performed using the electronic computing device, depending on the environmental model, and this assessment is taken into account when determining the lane-change intention. In particular, this allows for error handling that, for example, addresses the quality of lane markings based on fitting errors derived from corresponding fusion data, thereby reducing the credibility of, for example, a path hypothesis. This error handling can be applied holistically to various situational characteristics. Thus, potential errors in determining the lane-change intention can also be considered.

[0022] Furthermore, it has proven advantageous to consider, when assessing errors, the curvature of a roadway on which the other vehicle is located, and / or noise in the environmental perception, and / or at least one current environmental parameter, and / or at least one parameter characterizing the environmental perception. In particular, curve geometries can thus be taken into account when determining the characteristic situation features. Furthermore, the impact of noise on object and lane information on the determination of the situation features can be considered. Additionally, other vehicles without lane information can also be taken into account. In particular, all features function reliably even when only the lane information of the vehicle is available.This is particularly important when, for example, adjacent lanes cannot be modeled in the environment model due to bad weather, especially the current environmental parameter, and / or due to obstruction by, for example, a truck, which specifically describes the parameter characterizing the environment perception. With multiple lanes, a single reference lane is then sufficient to determine the lane-change probabilities of all vehicles.

[0023] According to a further advantageous embodiment, a path feature of the other vehicle and / or a position feature of the other vehicle and / or a clearance feature for the other vehicle is generated as a further situation feature characterizing the lane change of the other vehicle and is taken into account when determining the lane change intention. The path feature or trajectory feature can, in particular, be a so-called trajectory hypothesis. The trajectory hypothesis is modeled, in particular, by using a sigmoid function in the lane coordinate system. Specifically, a heading of the other vehicle is determined to define an intersection point with a lane line to the right or left. From this simple straight line, the parameters T required for feature extraction can already be derived. LCR (Crossing a lane marking) and Φ Spur(Course angle relative to the lane markings) is extracted. This is a very simple form of lane change prediction, which is why, after finding an intersection point with the lane marking, another intersection point on the center line of the respective adjacent lane is sought. Depending on the course angle and the time until crossing a lane marking, a future trajectory can then be predicted based on geometric approximation.

[0024] The positional feature can also be referred to as the lateral hypothesis. The lateral hypothesis is derived from two parameters: lateral distance and lateral velocity. Both values ​​provide information about the vehicle's position within the lane relative to the lane markings on the left and right. Based on this, an additional hypothesis or estimate for an upcoming lane change can be determined. The lateral distance is calculated by determining the shortest path between the vehicle's position and an adjacent polyline segment of the lane marking. This yields the lateral distance between the vehicle's position and a point on the polyline segment, and thus the vehicle's position relative to the outer lane markings. The lateral velocity is calculated from the change in distance to the lane marking over a fixed time interval.

[0025] The free space feature can also be referred to as the free space hypothesis. For each object in the vicinity, an occupancy grid is created. This grid is defined using four segments that run orthogonally to the respective vehicle heading and at defined intervals from the vehicle's pose. The edges of the grid cells are determined based on these segment lines and the marking lines of the corresponding neighboring lanes. If no left lane exists for the other vehicle, the grid is created only for the right lane, and vice versa. If, for example, no lane is detected, the free space feature is still determined. The algorithm iterates over all other vehicles in the vicinity, except for the vehicle itself, and determines the nearest point on the centerline of the lane in which the vehicle is located.This point serves to split the central line into a forward and a backward heading line. These two heading lines are then used to determine the intersection point with occupancy time grids on the respective lane. This approach has the advantage that the lane geometry is incorporated into the distance and time calculations. In particular, this allows multiple objects to influence the probability of movement in the vehicle's vicinity. To account for this, the individual probabilities are assumed to be stochastically independent and can be combined as shown in the formula above. This ensures that all vehicles or objects in the vicinity are considered in the free-space analysis.

[0026] In a further advantageous embodiment, the electronic computing device outputs the lane change probability as probability values ​​for a right lane change, a left lane change, and following another vehicle. Specifically, the respective probability values ​​range from 0 to 1. The sum of all probabilities equals 1. Based on these probability values, the lane change intention can then be reliably determined using the electronic computing device. The highest value is given preference. In other words, the highest of the three probability values ​​is determined as the lane change intention.

[0027] A further aspect of the invention relates to an assistance system for a motor vehicle for determining the lane-change intention of another motor vehicle located in the immediate vicinity of the motor vehicle, comprising at least one environment detection device and an electronic computing device, wherein the assistance system is configured to carry out a method according to the preceding aspect. In particular, the method is carried out by means of the assistance system.

[0028] A further aspect of the invention relates to a motor vehicle with an assistance system according to the previous aspect. In particular, the motor vehicle is designed to be at least partially autonomous.

[0029] The invention also includes further developments of the assistance system and the motor vehicle according to the invention, which have features already described in connection with the further developments of the method according to the invention. For this reason, the corresponding further developments of the assistance system and the motor vehicle according to the invention are not described again here.

[0030] The invention also includes combinations of the features of the described embodiments.

[0031] The following describes exemplary embodiments of the invention. This is illustrated by: Fig. 1 a schematic block diagram of an embodiment of a motor vehicle with an embodiment of an assistance system; Fig. 2 a schematic top view of a traffic situation to determine a path feature; Fig. 3. Another schematic top view of a traffic situation to determine a positional feature; Fig. 4. Another schematic top view of a traffic situation to determine a free space characteristic; Fig. 5. A further schematic top view of a traffic situation for determining a lane-changing motive; and Fig. 6 a schematic block diagram of an embodiment of a Bayesian network model.

[0032] The embodiments described below are preferred embodiments of the invention. In these embodiments, the described components each represent individual features of the invention that can be considered independently of one another. Each of these features further develops the invention independently and can therefore be considered part of the invention individually or in a combination other than that shown. Furthermore, the described embodiments can also be supplemented by other features of the invention already described.

[0033] In the figures, functionally identical elements are each provided with the same reference symbols.

[0034] Fig. Figure 1 shows a schematic block diagram of an embodiment of an assistance system 1 of a motor vehicle 2. The motor vehicle 2 can be designed as an at least partially autonomously operated motor vehicle 2. The assistance system 1 is used to determine a lane change intention 3 for another motor vehicle 4 ( Fig. 2) trained. The assistance system 1 has at least one environmental sensing device 5 and one electronic computing device 6 for this purpose. The environmental sensing device 5 can be, for example, an ultrasonic sensor device and / or a radar sensor device and / or a lidar sensor device and / or a camera.

[0035] In the procedure for determining the lane-changing intention 3 of the vehicle located in an immediate vicinity 7 ( Fig. 2) The environment 7 of the motor vehicle 2 is detected by means of at least the environment detection device 5 of the assistance system 1, and an environment model 8 for the environment 7 is generated by means of the electronic computing device 6 based on the detected environment 7. Depending on the generated environment model 8, a situation feature 9, 10, 11, 12, 13 characterizing a lane change by the other motor vehicle 4 is determined by means of the electronic computing device 6, wherein the lane change intention 3 is determined by means of a Bayesian network model 14 of the electronic computing device 6 depending on the at least one characterizing situation feature 9, 10, 11, 12, 13.

[0036] It is provided that a lane-changing motive 9 for the further motor vehicle 4 is determined as at least one situational feature 9, 10, 11, 12, 13 characterizing the lane change of the further motor vehicle 4, wherein the lane-changing motive 9 depends on a relative distance s rel ( Fig. 5) of the further motor vehicle 4 to an object 15 ( Fig. 5) in the vicinity of 7, depending on a relative velocity v rel ( Fig. 5) of the further motor vehicle 4 to the object 15 and is determined depending on an object type.

[0037] The Fig. Figure 1 further shows that, as a further situation feature 9, 10, 11, 12, 13 characterizing the lane change of the further motor vehicle 4, a path feature 10 of the further motor vehicle 4 and / or a position feature 11 of the further motor vehicle 4 and / or a clearance feature 12 for the further motor vehicle 4 are generated and are taken into account when determining the lane change intention 3.

[0038] Furthermore, especially in Fig. Figure 1 shows that, as a further situational feature 9, 10, 11, 12, 13 characterizing the lane change of the other vehicle 4, a context evaluation 13 is performed using the electronic computing device 6, whereby context information for a potential lane change in the environment 7 is taken into account. In particular, traffic sign monitoring and / or lane line monitoring of the roadway on which the other vehicle 4 is located can be used as context information. In particular, lane-change-relevant information can be taken into account during the context evaluation. For example, no-overtaking signs and solid lines on the roadway can be considered. Thus, context knowledge from the environment model 8 can be taken into account.

[0039] The characterizing situation features 9, 10, 11, 12, 13 can be generated in particular on the basis of a feature extraction 16 from the environment model 8.

[0040] Furthermore, it can be provided, in particular, that an error assessment is carried out using the electronic computing device 6, depending on the environment model 8, and that this error assessment is taken into account when determining the lane-change intention 3. For example, the error assessment can consider the curvature of a roadway on which the other vehicle 4 is located, and / or noise in the environment detection, and / or at least one current environmental parameter, and / or at least one parameter characterizing the environment detection. Thus, error handling can be provided. In particular, curve geometries can be taken into account when determining the situation features 9, 10, 11, 12, and 13. Furthermore, the effects of noisy object and lane information can be considered.In particular, this ensures that the assistance system 1 can operate reliably even when only lane information from vehicle 1 is available. This is especially important when adjacent lanes cannot be modeled in the environment model 8 due to bad weather or obstruction by, for example, trucks.

[0041] With multiple lanes, a single reference lane is sufficient to determine the lane-changing probabilities of all vehicles.

[0042] In particular, it may also be provided that the electronic computing device 6 outputs the lane change intention 3 as probability values ​​for a right lane change, for a left lane change, and for a following journey.

[0043] To utilize the characterizing situation features 9, 10, 11, 12, 13, in particular the lane change motif 9, the path feature 10, the position feature 11 and the free space feature 12, a sigmoid function 20 can be used.

[0044] Fig. Figure 2 shows a schematic top view of a traffic situation for determining the path feature 10. In particular, a geometric interpolation method is used to determine the features for the trajectory. The path feature 10 can also be referred to as the trajectory hypothesis (TR). This involves, in particular, determining the intersection point between so-called heading segments 17 of the other vehicle 4 and a center line 18 of an adjacent lane 19. The resulting segment is bisected, and a difference angle α is bisected, both of which are parameterizable. These steps are repeated until the difference angle α is below a threshold value. A time to cross the lane marking T can then be calculated. LCRFurthermore, an angle can be determined at the intersection between the trajectory or path and the outer track marker, which can be referred to as the Φ-track. P(TLCR)=TLCR→fSIG(x);P(ΦTrack)=ΦTrack→fSIG(x)

[0045] The size for feeding into the Bayes network model 14 is then determined by: P(TR)=PTLCR*P(ΦTrack).

[0046] Fig. Figure 3 shows another schematic top view of a different traffic situation. In the present embodiment, the position feature 11 is determined in particular. The position feature 11 is, in particular, a so-called lateral hypothesis (LE). To determine the lateral evidence, the position of the other motor vehicle 4 in the lane, relative to the outer lane marking, is determined. A distance o LAT between a pose 22 and a track marking 23, in particular to a point P1.

[0047] From this, a lateral velocity v can be calculated. LAT can be derived. The value ranges of o are normalized. LAT and v LAT about the sigmoid function 20 in the range of values ​​between 0 and 1 as probability: P(oLAT)=oLAT→fSIG(x) P(vLAT)=vLAT→fSIG(x)

[0048] The following quantities are now fed into the Bayesian network model: P(LE)=P(oLAT)*P(vLAT)

[0049] Fig. Figure 4 shows another schematic top view of a different traffic situation. Fig. In particular, the free space feature 12 can be determined. The free space feature 12 can also be referred to as the free space hypothesis (FR). For this purpose, an occupancy grid 24 is set for the next motor vehicle 4 if an adjacent lane is present. The initial value for the individual cells Z1 to Z4 is 100 percent, which means that free space is available. As the vehicle, represented here by object 15, approaches cells Z1, Z2, Z3, Z4, the distance to entering and exiting cells Z1, Z2, Z3, Z4 is calculated. The distance S shown TE This describes the distance until entry into a respective cell Z1, Z2, Z3, Z4. The distance S TD This describes the distance until exiting cell Z1, Z2, Z3, Z4. The time until entering and exiting cells Z1, Z2, Z3, Z4 is determined using the equation of motion for uniform motion. T TEThis describes the time until entry into cells Z1, Z2, Z3, Z4 and T. TD describes the time until exit from cells Z1, Z2, Z3, Z4. The probabilities are then determined using the sigmoid function 20. P(STE)=STE→fSIG(x);P(TTE)=TTE→fSIG(x) P(STD)=STD→fSIG(x);P(TTD)=TTB→fSIG(x)

[0050] The probabilities are derived by summarizing the characteristics: P(TE)=P(STE)*P(TTE) P(TD)=P(STD)*P(TTE)

[0051] The parameters for feeding into the Bayes network model 14 are the movement probabilities: P(FR(I))=P(TE)+P(TD)

[0052] If several objects have an influence on cells Z1, Z2, Z3, Z4, the movement probabilities are determined by the formula: PGES(FRCell(i))=POBJ1(FRCell(i))*POBJ2(FRCell(i)) In particular, multiple objects (15) can thus influence the probability of movement in the vehicle's vicinity. To account for this, the individual probabilities are assumed to be stochastically independent and can be combined as shown in the formula above. This ensures that all vehicles or objects in the vicinity are considered in the free-space analysis.

[0053] The discretization of the vehicle's surroundings is not achieved using grids, but rather lines. The features are calculated taking the road geometry into account. The discretization lines follow the orientation of object 15. In particular, only four discretization lines are required.

[0054] Fig. Figure 5 shows another schematic top view of a different traffic situation. In particular, the lane-change motive 9 is shown. The lane-change motive 9 can also be referred to as the motivation hypothesis. It estimates the motivation for the lane change of the other vehicle 4. This is based on relative dimensions to the object 15 of the lane-change candidate. To formulate this hypothesis, the following dimensions between the lane-change candidate and its vehicle in front, i.e., the other vehicle 14 and the object 15, are determined. The relative distance s is then determined. rel , the relative velocity v rel as well as the predicted time until the two vehicles meet, in particular a so-called collision time point (TTC). Furthermore, the vehicle types of vehicles 4 and objects 15 can also be taken into account.

[0055] Fig.Figure 6 shows a schematic block diagram of an embodiment of the Bayesian network model 14. In particular, a multitude of different nodes K1 to K17 are shown. The Bayesian network model 14 allows conclusions to be drawn taking into account uncertainties and missing a priori information. This is, in particular, a discrete Bayesian network model 14, which means that an event can only assume discrete values. In our case, for example, these are only true and false. The events are random variables. The parameterization is based on expert knowledge, especially on findings from test drives and trials. The calculated situational features 9, 10, 11, 12, 13 are fed into the Bayesian network model 14 as observable events, specifically as so-called a priori knowledge. The conclusion regarding a possible lane change is drawn by the Bayesian network model 14.

[0056] All specific situational features 9, 10, 11, 12, 13 must be normalized to a value between 0 and 1 for processing in the Bayesian network model 14. For this purpose, the sigmoid function 20 is used, which maps a function value between 0 and 1 for each input value with a fixed range of values. This function value is then fed into the Bayesian network model 14 as observable variables (parent nodes).

[0057] The range of values ​​for the input variables is determined from measured values. The extreme values ​​of this range always represent the safe state, i.e., whether a lane change has occurred or not.

[0058] From this point onward, the situational features 9, 1, 11, 12, 13 are interpreted as events that occur or do not occur. The function values ​​of the sigmoid function 20 indicate the measure, i.e., a probability for both events.

[0059] The general form of the sigmoid function 20 for determining the probability of event X is described as follows: P(X=Yes|E)=η⋅1a+exp(b⋅E)

[0060] The non-occurrence of the event is calculated as follows. P(X=No|E)=1−P(X=Yes|E)

[0061] E: Input values ​​from the situation characteristics, normalization factor a, b gradient of the curve from 0 to 1

[0062] Value ranges and parameters for normalizing situational characteristics 9, 10, 11, 12, 13

[0063] Value ranges and parameters for normalizing the hypothesis Lateral Evidence (LE): P(LE=Yes|v_lat):a_v_lat=0.04, b_v_lat=9, _v_lat=0.04 v_lat{−1.5…0.0} P(LE=Yes|o_lat):a_o_lat=60, b_o_lat=3.5, _o_lat=60 o_lat{0.0…2.0}

[0064] Calculation of the probability for the event Lateral Evidence. Serves as input for node LE in Bayesian network model 14. P(LE=Yes|v_lat, o_lat)=P(LE=Yes|v_lat) P(LE=Yes|o_lat)

[0065] For further consideration in the Bayesian network model 14, we use the abbreviated notation for probability: P(LE=1)=P(LE=JA|v_lat)P(LE=JA|o_lat)

[0066] Value ranges and parameters for normalizing the Trajectory Hypothesis (TR): P(TR=JA|T_lcr):a_T_lcr=150,b_T_lcr=2.5,_T_lcr=151 T_lcr{−1.0..4.0} P(TR=YES|spur):a_spur=0.01,b_spur=200,_spur=0.14 _spur{−0.05...0.0}

[0067] Calculation of the probability for the event trajectory. Serves as the input node for node TR in Bayes network model 14. P(TR=Yes|T_lcr,_spur)=P(TR=Yes|T_lcr)P(TR=Yes|_spur)

[0068] For further consideration in the Bayesian network model 14, we use the abbreviated notation for probability: P(TR=1)=P(TR=Yes|T_lcr)P(TR=Yes|_spur)

[0069] Value ranges and parameters for normalizing the hypothesis Free Space (FR): P(Entry=Yes|T_TE):a_T_TE=0.2,b_T_TE=−3,_T_TE=0.24 T_TE{−1.0...1.0} P(Entry=Yes|S_TE):a_S_TE=0.4,b_S_TE=−0.4,_S_TE=0.41 S_TE{−10.0...1.0} P(Exit=Yes|T_TD):a_T_TD=0.2,b_T_TD=−3,_T_TD=0.24 T_TD{−1.0...1.0} P(Exit=Yes|S_TD):a_S_TD=0.4, b_S_TD=0.4,_S_TD=0.41 S_TD{−1.0...10.0}

[0070] Calculation of the probability of the event that a vehicle enters the cell: P(Entry=Yes|T_TE,S_TE)=P(Entry=Yes|T_TE)P(Entry=Yes|S_TE)

[0071] Calculation of the probability of the event that a vehicle exits the cell: P(Exit=Yes|T_TD,S_TD)=P(Exit=Yes|T_TD)P(Exit=Yes|S_TD) Calculation of the free-space probability:

[0072] The probabilities for the events "entry" and "egress" of the other vehicle 4 with respect to a cell Z1, Z2, Z3, Z4 can be summed due to their stochastic dependence. Thus, the free-space probability exerted by another vehicle 4 on a cell Z1, Z2, Z3, Z4 is as follows: P_ze(Entry=Yes Exit=Yes)=P(Entry=Yes)+P(Exit=Yes) P_ze: Free space probability of a vehicle for a cell

[0073] If another vehicle is located in front of a cell Z1, Z2, Z3, Z4 and yet another vehicle is located behind a cell Z1, Z2, Z3, Z4, both vehicles can have the same influence on the free-space probability of a cell Z1, Z2, Z3, Z4. That is, the free-space probability of the vehicle in front of the cell can be exactly the same as the free-space probability of the vehicle behind the cell Z1, Z2, Z3, Z4. Therefore, both events are stochastically independent, and the free-space probability, taking both vehicles into account, can be calculated by multiplying the individual probabilities {P_{ze}}. P_z(P_ze1 P_ze2)=P_ze1 P_ze2 P_z: Total free-space probability for a cell with two vehicles

[0074] General formula for calculating the free space probability of a cell Z1, Z2, Z3, Z4: P_z(P_z_i P_z_i+1 P_n)=_i=1^nP_ze_i ,n>0 n: Number of vehicles which have an influence on a cell

[0075] The Bayesian network model 14 for detecting a lane change is performed by determining the intermediate nodes "Cross lane marking left" K11 and "Cross lane marking right" K2. Both partial results feed into the result node "Lane change" K6. The hypothesis "Lane change" provides us with a probability for the events "Lane change to the left", "Lane change to the right", and "Following vehicle" per vehicle.

[0076] The input nodes for situation features 9, 10, 11, 12, and 13 are "Free Space" K7 and K10, "Trajectory" K1 and K9, and "Lateral Evidence" K0 and K8. The calculated situation features 9, 10, 11, 12, and 13 from feature extraction 16 are modeled as parent nodes in the Bayesian network model 14. The input variables of these nodes are the calculated probabilities from the normalization of situation features 9, 10, 11, 12, and 13 for the Bayesian network model 14.

[0077] A special feature here is the "Free Space" hypothesis. This hypothesis has nodes K3, K4, K5 and K12, K13, K14, which represent the individual cells Z1, Z2, Z3, Z4 of the grid with their corresponding free space probabilities. The free space probabilities of the individual cells are summarized in nodes K7 and K10. In this node, probabilities are assigned to the possible combinations of all individual events—that is, whether a cell Z1, Z2, Z3, Z4 is occupied or not—and these probabilities are fixed (expert knowledge). From these probabilities, the overall probability of the "Free Space" hypothesis is calculated.

[0078] In child nodes K2 and K11, the hypothesis "lane marking crossed" is calculated for the respective lane markings to the right and left of vehicle 1. All possible combinations of events from the nodes of situation features 9, 10, 11, 12, 13 (free space, trajectory, and lateral evidence) are parameterized with probabilities, allowing a probability for the hypothesis "lane marking crossed" to be calculated.

[0079] The probabilities from the hypotheses "lane marking crossed" are entered as an event into the node "lane change". Within the node "lane change", the event "lane change to the right", "lane change to the left" or "following vehicle" can then be calculated.

[0080] Overall, the invention demonstrates lane change detection or intention detection. Reference symbol list 1 Assistance system 2 motor vehicles 3 Lane change intention 4 additional motor vehicles 5 Environmental sensing device 6 electronic computing equipment 7 Environment 8 Environment model 9 Lane change motif 10 Path feature 11 Position characteristic 12 Free space feature 13 Context analysis 14 Bayesian network model 15 objects 16 Feature extraction 17 Heading 18 Central Line 19 adjacent lanes 20 Sigmoid function 21 adapted lane 22 Vehicle pose Line 23 24 grid α Difference angle O LAT Route v LAT lateral velocity P1 point S TD Exit distance S TE Entry distance T TE Entry time T TD Exit time s relrelative distance v rel relative speed TTC Collision Time K1 to K17 nodes

Claims

[1] Method for determining a lane-change intention (3) of another motor vehicle (4) located in the immediate vicinity (7) of a motor vehicle (2) by means of an assistance system (1) of the motor vehicle (2), in which the environment (7) of the motor vehicle (4) is detected by means of at least one environment detection device (5) of the assistance system (1) and an environment model (8) for the environment (7) is generated on the basis of the detected environment (7) by means of an electronic computing device (6) of the assistance system (1), and in which, depending on the generated environment model (8), at least one situation feature (9, 10, 11, 12, 13) characterizing a lane change of the other motor vehicle (4) is determined by means of the electronic computing device (6), as the at least one situation feature (9, 10, 11, 12, 13) characterizing the lane change of the other motor vehicle (4) is determined.13) a lane-changing motive (9) is determined for the other motor vehicle (4), wherein the lane-changing motive (9) depends on a relative distance (s, rel ) of the further motor vehicle (4) to an object (15) in the vicinity (7), depending on a relative speed (v rel ) of the further motor vehicle (4) to the object (15) and is determined depending on an object type, characterized by, that by means of a Bayesian network model (14) of the electronic computing device (6) the lane change intention (3) is determined as a function of at least one characterizing situation feature (9, 10, 11, 12, 13), wherein as a further situation feature (9, 10, 11, 12, 13) characterizing the lane change of the other motor vehicle (4) a context evaluation (13) is carried out by means of the electronic computing device (6), wherein context information for a potential lane change in the environment (7) is taken into account, and wherein traffic sign monitoring is used as context information. [2] Method according to claim 1, characterized by that depending on the relative speed (v rel ) and the relative distance (s rel) a collision time (TTC) between the other motor vehicle (4) and the object (15) is predicted by means of the electronic computing device (6) and the collision time (TTC) is taken into account in the lane change motive (9). [3] Method according to claim 1 or 2, characterized by , that a vehicle type of the other motor vehicle (4) is taken into account in the lane change motive (9) by means of the electronic computing device (6). [4] Method according to any one of the preceding claims, characterized by , that as additional contextual information a line monitoring of a roadway on which the other motor vehicle (4) is located is used. [5] Method according to any one of the preceding claims, characterized by , that an error assessment is carried out using the electronic computing device (6) depending on the environment model (8) and the error assessment is taken into account when determining the lane change intention (3). [6] Method according to claim 5, characterized by , that in the error assessment a curve curvature of a roadway on which the other motor vehicle (4) is located, and / or noise in the environmental detection and / or at least one current environmental parameter and / or at least one parameter characterizing the environmental detection are taken into account. [7] Method according to any one of the preceding claims, characterized by , that as a further situation feature (9, 10, 11, 12, 13) characterizing the lane change of the further motor vehicle (4) a path feature (10) of the further motor vehicle (4) and / or a position feature (11) of the further motor vehicle (4) and / or a free space feature (12) for the further motor vehicle (4) are generated and are taken into account when determining the lane change intention (3). [8] Method according to any one of the preceding claims, characterized by, that by means of the electronic computing device (6) the lane change intention (3) is output as probability values ​​for a right lane change and for a left lane change and for a following journey. [9] Assistance system (1) for a motor vehicle (2) for determining a lane change intention (3) of another motor vehicle (4) located in the immediate vicinity (7) of the motor vehicle (2), comprising at least one environment detection device (5) and an electronic computing device (6), wherein the assistance system (1) is configured to carry out a method according to one of claims 1 to 8.

Images