Driver assistance system for motor vehicles with a method for detecting a driver's desire to evade a collision

Abstract

Driver assistance system for motor vehicles (34), comprising an electronic data processing system (10) in which a method for detecting a driver's evasive maneuver is implemented, in which the steering angle (φ) and the steering angle velocity (dφ / dt) as well as the vehicle speed (V) are evaluated as primary input variables, characterized in that the method comprises the following steps: - Determining model trajectories (42', 44') that specify the primary input variables as functions of time, based on additional input variables that characterize the traffic environment and / or the state of the own vehicle (34), wherein at least one model trajectory (42') is determined that corresponds to an evasive maneuver of the vehicle, and checking whether at least one alternative model trajectory (44') exists that does not correspond to an evasive maneuver of the vehicle. - if at least one alternative model trajectory (44') exists, selecting one of the determined model trajectories based on the degree of agreement with measured values ​​of the primary input variables, - Decide whether a driver has requested an evasive maneuver based on the selected model trajectory.

Description

State of the art

[0001] The invention relates to a driver assistance system for motor vehicles, comprising an electronic data processing system in which a method for detecting a driver's evasive maneuver is implemented, in which the steering angle and / or the steering angle velocity as well as the vehicle speed are evaluated as primary input variables.

[0002] Driver assistance systems for motor vehicles are under development that support the driver in the event of an impending collision (especially in longitudinal traffic) during an evasive maneuver (Evasive Steering Support, ESS). For this to work, the driver must first initiate the evasive maneuver (e.g., after a collision warning from the system). If the system detects that the driver intends to avoid the obstacle, the vehicle is stabilized on a suitable evasive trajectory via a superimposed steering torque or by asymmetrically applying the brakes. This makes executing the demanding evasive maneuver considerably easier and safer for the driver.

[0003] Since the evasive steering assist is only activated after a swerve has been detected, the algorithm for detecting the swerve must have the highest possible detection rate with the lowest possible false trigger rate. The applicant has disclosed a method in which an environmental sensor, e.g., a radar sensor, detects whether an object exists with which a collision is imminent. If this condition is met, the steering angle (e.g., steering wheel angle) and the steering angle velocity (preferably low-pass filtered) are evaluated. If both values ​​exceed a threshold, the evasive steering assist is activated.

[0004] At high speeds, for example above 60 km / h, where ESS is currently available, the established method works very reliably. However, if an evasive steering assist function is also to be implemented at lower speeds, e.g., in city traffic, there is a risk that the error rate of the conventional method will increase because, as long-term tests have shown, strong steering inputs and high steering angle velocities occur more frequently at low speeds, even when no evasive maneuvers are taking place. Reasons for this at very low speeds, where evasive steering assist would not be useful anyway, include parking and maneuvering. At somewhat higher speeds, high values ​​for steering angle and steering angle velocity occur, particularly in connection with turning maneuvers. However, evasive maneuvers can also occur in this speed range, e.g.,To avoid a collision with a pedestrian crossing the road, it would be quite useful, so it would be desirable to be able to offer evasive steering support even in this speed range.

[0005] From DE 10 2006 043 676 A1 a driver assistance system is known in which primary input variables that are characteristic of the lateral dynamics of the vehicle are evaluated in order to assess the driver's attention and then to decide on this basis whether a collision warning should be issued to the driver or not.

[0006] DE 10 2008 040 241 A1 describes a method for assisting the driver of a motor vehicle during an evasive maneuver, also checking whether the driver initiates an evasive maneuver, and taking into account the driver's steering torque.

[0007] In DE 10 2005 003 274 A1, a method and a device for avoiding and / or reducing the consequences of collisions when a vehicle swerves to avoid obstacles are presented. Here, too, the driver's intention is determined, e.g., based on the steering angle.

[0008] DE 10 2009 003 203 A1 discloses a method for detecting an evasive maneuver intended by the driver of a vehicle, taking into account the steering angle. Disclosure of the invention

[0009] The object of the invention is to provide a method that enables the detection of a driver's intention to evade the road with a high detection rate and a low false trigger rate, even in the lower speed range.

[0010] This problem is solved according to the invention by providing a driver assistance system for motor vehicles with a method comprising the following steps: - Determining model trajectories that specify the primary input variables as functions of time, based on additional input variables that characterize the traffic environment and / or the state of the vehicle itself, whereby at least one model trajectory is determined that corresponds to an evasive maneuver of the vehicle, and it is checked whether at least one alternative model trajectory exists that does not correspond to an evasive maneuver of the vehicle. - if at least one alternative model trajectory exists, select one of the specified model trajectories based on the degree of agreement with measured values ​​of the primary input variables, and - Decide whether a driver has requested an evasive maneuver based on the selected model trajectory.

[0011] This method makes it possible to distinguish, in situations where the driver makes violent steering maneuvers, whether the most plausible cause for these steering maneuvers is that the driver intends to perform an evasive maneuver, or whether it is more likely that the driver is making these steering maneuvers for other reasons.

[0012] For the mathematical description of the model trajectories, the steering angle and steering angular velocity are particularly useful. However, within the scope of the invention, it is not excluded that other input variables are measured and evaluated, for example, the yaw rate and / or the yaw acceleration, in which the information about the steering angle and steering angular velocity is only implicitly contained. Likewise, other input variables may also be measured and evaluated.

[0013] Advantageous further developments and embodiments of the invention are specified in the dependent claims.

[0014] Additional input variables that may be considered include, in particular: - the history of the vehicle speed in the seconds before the maneuver, - the status of the turn signal indicator (indicator) - Information about upcoming curves, intersections or junctions, especially based on data from a digital map, - Movement data of vehicles ahead, - the existence of a route programmed into the navigation system, - Position, speed and class (e.g. pedestrian, car, truck, ...) of the object at risk of collision.

[0015] To determine the model trajectory, in addition to the input variables mentioned above (or a suitable selection thereof), a knowledge base can also be used. This knowledge base contains information or rules about typical human driver behavior, indicating which vehicle trajectories are to be expected in which traffic situations. This knowledge base can be compiled, for example, from empirical data and / or a corresponding collection of rules and principles (expert knowledge).

[0016] In the following, exemplary embodiments of the invention are explained in more detail with reference to the drawing.

[0017] They show: Fig. 1 a block diagram of a driver assistance system with which the invention can be carried out; Fig. 2 a sketch of a traffic situation to illustrate the method according to the invention; Fig. 3 a schematic representation of model trajectories in the sketch according to Fig. 2; and Fig. 4 and Fig. 5 flowcharts for embodiments of the method according to the invention.

[0018] The in Fig. The driver assistance system shown in Figure 1, depicted as a block diagram, features an electronic data processing system 10 that receives data from an environmental sensor 12, represented here by a radar sensor. The radar sensor is installed, for example, at the front of the vehicle and serves to locate vehicles and other objects in front of the vehicle.

[0019] The data processing system 10 receives further input from a vehicle speed sensor 14, which measures the vehicle's speed V, and from a steering wheel sensor 16, which measures the steering angle φ (steering wheel angle) and the steering angle velocity dφ / dt. Furthermore, the data processing system communicates with the vehicle's navigation system 18, which provides information about road layouts, intersections, and the like in the form of an electronic map and, if a route guidance system is activated, also provides information about the currently followed route.

[0020] A flasher switch 20 provides information about the status of the vehicle's own turn signal indicator.

[0021] The data from the environmental sensors 12 are evaluated by a detection module 22, which determines in particular the distances, relative velocities and azimuth angles of objects located in front of the vehicle.

[0022] The data from the vehicle speed sensor 14 and the steering wheel sensor 16, and in the example shown also the data from the navigation system 18 (position of the vehicle) and the turn signal switch 20, are continuously recorded by a recording module 24, each time for the duration of a sliding time window of predetermined length.

[0023] Based on data from the detection module 22 and the recording module 24, a collision warning module 26 continuously assesses the traffic situation and, using location data and the vehicle's own dynamic data, checks whether at least one of the detected objects represents a potential obstacle, i.e., an object that poses a collision risk. Based on this assessment, the collision warning module 24 decides whether to issue a collision warning (visual or audible) to the driver. When a collision warning is issued, the recording module 24 simultaneously receives a signal that "freezes" the recording window; that is, the data stored at that moment is no longer deleted, but newly arriving data continues to be recorded.

[0024] Based on the newly arriving dynamic data, in particular the current values ​​of the steering angle and steering angle velocity, as well as the history recorded in recording module 24, a decision module 28 then evaluates the driver's behavior to determine whether the driver intends to initiate an evasive maneuver (either proactively or in reaction to the collision warning). In doing so, decision module 28 also accesses a knowledge base 30, which stores experiential knowledge about typical driver behavior. The knowledge base 30 also evaluates supplementary information about the traffic environment provided by the detection module 22, such as information about the classes of located objects, especially the potential obstacle (pedestrian, car, truck, etc.), as well as current movement data of the potential obstacle and, if applicable,other relevant objects.

[0025] If the decision module 28 decides that the driver actually intends to evade, it activates an evasive steering assist 32 (ESS), which then actively intervenes in the steering and / or braking system of the vehicle to actively assist the driver in carrying out the evasive maneuver.

[0026] The functionality of decision module 28 will now be explained using the example of a traffic situation, which is sketched out in Fig. 2 is shown.

[0027] A vehicle 34, equipped with the driver assistance system (the “own vehicle”), is approaching an intersection 36 where two other vehicles 38 and 40 are currently colliding. Vehicle 40 is blocking part of the lane being traveled by the own vehicle 34 and thus constitutes an obstacle posing a collision risk. The collision warning module 26 then issues a collision warning. Fig. Figure 2 shows a possible evasive trajectory 42 for the driver's own vehicle 34. If the driver of vehicle 34 originally intended to continue straight ahead at intersection 36, he would now likely attempt to bypass the obstacle (vehicle 40) on this evasive trajectory 42.

[0028] However, it is also conceivable that the driver of vehicle 34 originally intended to turn right at the intersection and accordingly follow a turning trajectory 44 by first moving into a turning lane 46 and then turning into a cross street 48 at the intersection. In that case, it would be irritating, disruptive, and possibly even dangerous for this driver if the evasive steering assist 32 were nevertheless to attempt to keep the vehicle on the evasive steering trajectory 42.

[0029] Therefore, in decision module 28 an algorithm is implemented that makes it possible to distinguish between the two possibilities shown above with a low error rate.

[0030] For this purpose, the decision module 28 first calculates a model trajectory 42', which is then Fig. 3 is represented as a sequence of circles and the avoidance trajectory 42 in Fig. 2 corresponds to the diameter of the circles in Fig. 3 indicates the instantaneous speed of the vehicle 34, i.e., the distance the vehicle travels within a time interval dt of fixed length. The curvature of the model trajectory and its derivative correspond to the steering angle φ and its derivative dφ / dt. The mathematical representation of the model trajectory 42' is thus equivalent to specifying the steering angle φ and the steering angular velocity dφ / dt as functions of time (or also as functions of position, which can be converted into time-dependent functions if the vehicle speed V is known).

[0031] In the Fig. In example 3, it is assumed for simplicity that the driver travels the evasive trajectory at a constant speed. In general, however, the driving speed is time-dependent. When determining the model trajectory 42', movements that the vehicle 40 may still be performing can also be taken into account, as well as parameters indicating the presumed dimensions of the vehicle 40 and the like. The dynamic data of the vehicle 34 itself, in particular its initial speed upon obstacle detection, also influences the course of the model trajectory. Algorithms for calculating such model trajectories are known and are also used, for example, in calculating target trajectories for the evasive maneuver in the evasive steering assist system 32.

[0032] Furthermore, the decision module 28 calculates at least one alternative model trajectory 44', which corresponds to another route that the driver of vehicle 34 might desire, in the example shown, the turning trajectory 44. In determining this and, if applicable, further alternative model trajectories, the decision module 28 accesses the digital map provided by the navigation system 18, which contains information about the presence of intersection 36 and cross street 48, as well as, if applicable, information about the presence of turning lane 46. Optionally or additionally, supplementary information from the environmental sensors 12 can also be accessed in this context, for example, from a video camera that detects road and lane contours.

[0033] Furthermore, the expected speed profile on the model trajectory 44' is modeled using theoretical models or empirical data from knowledge base 30. For example, the driver will reduce speed shortly before the actual turning maneuver, which in Fig. 3 can be seen from the fact that the diameters of the circles decrease in this area of ​​the trajectory.

[0034] Typically, the driver will also reduce their speed before changing to turning lane 46.

[0035] In Fig. Line 3 marks the time T0 at which the obstacle is detected and the collision warning is issued. However, the model trajectories 42' and 44' also include a time interval prior to this time T0. For this time interval, the measured data for vehicle speed V, steering angle φ, and steering angle velocity dφ / dt are available in recording module 24.

[0036] Comparing the model data with the measured data provides a criterion for distinguishing which of the two model trajectories the driver was more likely to choose. The decrease in driving speed V even before time T0, i.e., before the obstacle was even recognizable, indicates that the driver intended to turn at the intersection anyway.

[0037] In a similar way, characteristic signatures in the course of the steering angle and / or the steering angle speed can also be taken into account.

[0038] Further clues are provided by the state of the turn signal switch 20 (if a turn is desired, the driver will activate the turn signal before or at the latest shortly after time T0) as well as a route stored in the navigation system 18.

[0039] In Fig. Figure 4 shows a possible procedure in the form of a flowchart. If the collision warning module 26 has detected an imminent collision risk and, if applicable, issued a collision warning, in step S1 a number of model trajectories M1, M2, ..., Mn are determined that are possible trajectories for the vehicle in the current traffic situation. This includes at least one evasive trajectory on which the obstacle can be bypassed. Naturally, further (alternative) model trajectories can only be calculated if the infrastructure (layout of lanes and tracks, turning options, etc.) allows for such alternative trajectories.

[0040] Input variables in step S1 include, in particular, the location data of the object posing the collision risk, possibly location data of other objects that are to be considered as potential further obstacles, the position of the vehicle itself known from the navigation system, and the map data from the navigation system. Each of the model trajectories determined in this way contains information that specifies the associated temporal profile of the steering angle and / or the steering angle velocity (and preferably the vehicle speed V).

[0041] In a further step S2, the model data are then compared with the measurement data supplied by recording module 24, which specifies the history of the steering angle, steering angle velocity, and vehicle speed. Probabilistic methods are known for trajectory matching that allow each model trajectory M1, M2, Mn to be assigned a probability value P1, P2, Pn, which indicates the probability, given the history of the measurement data, that the model trajectory in question corresponds to the trajectory actually desired by the driver. Other known methods for trajectory matching use fuzzy logic. In this case, corresponding membership measures are obtained instead of the probability values ​​P1, P2, Pn.

[0042] In the Fig. In the example shown, step S2 is followed by a further step S3, in which the preliminary results obtained through trajectory matching are fused with additional features, in particular the turn signal, the route stored in the navigation system, and the like. This fusion may further modify the probability values ​​or membership measures. The effects of the additional features on the probability values ​​can be modeled, for example, using a Bayesian network.

[0043] The model trajectory with the highest probability or membership measure is then selected as the presumably desired trajectory. If the selected trajectory is an avoidance trajectory, step S4 checks whether an actual collision risk still exists. If so, avoidance assistance is initiated in step S5 based on the selected avoidance trajectory.

[0044] Fig. Figure 5 illustrates an alternative procedure in which the fusion with the other features such as turn signal, navigation route and the like is carried out according to step S3 in Fig. 4 is already integrated into step S1. In this case, the probability values ​​P1, P2, Pn obtained in step S2 lead directly to the selection of the most probable trajectory and thus, if applicable, to the detection of an evasive maneuver request.

Claims

[1] Driver assistance system for motor vehicles (34), comprising an electronic data processing system (10) in which a method for detecting a driver's evasive maneuver request is implemented, in which the steering angle (φ) and steering angle velocity (dφ / dt) as well as the vehicle speed (V) are evaluated as primary input variables, characterized by that the procedure comprises the following steps: - Determining model trajectories (42', 44') that specify the primary input variables as functions of time, based on additional input variables that characterize the traffic environment and / or the state of the own vehicle (34), wherein at least one model trajectory (42') is determined that corresponds to an evasive maneuver of the vehicle, and checking whether at least one alternative model trajectory (44') exists that does not correspond to an evasive maneuver of the vehicle. - if at least one alternative model trajectory (44') exists, selecting one of the determined model trajectories based on the degree of agreement with measured values ​​of the primary input variables, - Decide whether a driver has requested an evasive maneuver based on the selected model trajectory. [2] Driver assistance system according to claim 1, wherein the method provides that the additional input variables include information about the traffic infrastructure based on a digital map. [3] Driver assistance system according to claim 1 or 2, wherein the method provides that the additional input variables include a state of a flasher switch (20). [4] Driver assistance system according to one of the preceding claims, wherein the method provides that the additional input variables include information about a route selected for route guidance in a navigation system (18). [5] Driver assistance system according to one of the preceding claims, wherein the method provides that at least one of the primary input variables is continuously recorded during a sliding time window and the additional input variables comprise the history of the recorded variables. [6] Driver assistance system according to one of the preceding claims, wherein the method provides that the selection of one of the specified model trajectories is carried out by probabilistic trajectory matching. [7] Driver assistance system according to one of claims 1 to 5, wherein the method provides that the selection of one of the determined model trajectories is carried out by trajectory matching using fuzzy logic. [8] Driver assistance system according to claim 6 or 7, wherein the method provides that the results of the trajectory matching are subsequently fused with at least one of the additional input variables.

Images