Method for predicting a specific driving maneuver

Abstract

Method for predicting a specific driving maneuver (H2) from all driving maneuvers currently available for selection (H1, H2, H3) by evaluating predetermined driver behavior characteristics (VV, BV, KV) for each driving maneuver available for selection (H1, H2, H3), whereby for each driver behavior characteristic (VV, BV, KV) for each driving maneuver (H1; H2; H3) a measure (SCOREv, SCOREb, SCOREk) for the probability of occurrence of this driving maneuver is determined by determining the degree of agreement between each observed driver behavior characteristic (VV, BV, KV) and the empirically determined driver behavior characteristic (VVt, BVt, KVt) typical for the respective driving maneuver for the respective driving maneuver (H1; H2;H3) is determined in that a predetermined driver behavior characteristic (BV) is formed on the basis of the detected parameter indicator signal (b), wherein a minimum measure greater than zero (SCOREb=0.2) for this turning maneuver (H2) is assigned to a lack of the indicator signal (b) at a correspondingly defined short distance (D) of the vehicle (F1) to a reference point (R2) for a turning maneuver (H2);

Description

[0001] The invention relates to a method for predicting a specific driving maneuver.

[0002] The increasing traffic volume in German urban areas is placing ever greater strain on drivers. This also increases the risk of accidents, particularly for pedestrians and cyclists, who, unlike vehicle passengers, are inadequately protected in the event of a collision. One approach to preventing collisions is the development of driver assistance systems that provide warning and / or delay intervention.

[0003] The basis for this is knowledge of the driver's intention, which allows predicting traffic conditions in the coming seconds and detecting impending conflict situations at an early stage. Reference is made, for example, to the published patent application DE 10 2006 040 537 A1, which infers, for example, the driver's intention to change lanes based on steering movements in conjunction with the turn signal and a constant speed.

[0004] For example, DE 10 2005 020 429 A1 discloses a driver assistance system for support in intersection areas, which uses environmental and route data by means of sensors and the digital map data of a navigation system.

[0005] Furthermore, DE 10 2009 047 264 A1 discloses a method for assisting a driver of a vehicle during a driving maneuver. The method first scans the vehicle's surroundings and checks whether a junction, intersection, or curve is to be negotiated. The expected direction of travel at the junction, intersection, or curve is determined by evaluating a trajectory.

[0006] DE 10 2010 049 721 A1 describes a method for displaying a criticality of a current and / or future driving situation of a first vehicle. Depending on a current driving situation mapped to an environment model, at least one criticality of the current driving situation of the first vehicle and / or at least one criticality of at least one future driving situation of the first vehicle is determined. Furthermore, depending on the at least one criticality, a visual warning is displayed on a display unit of the first vehicle.

[0007] The object of the invention is to improve a method of the type mentioned above with regard to its reliability.

[0008] This object is achieved according to the invention by the subject matter of patent claim 1. The dependent patent claims are advantageous developments of the invention.

[0009] The invention is based on the following considerations, findings and ideas: This application contains a concept for driver intention detection, particularly in intersections. The focus is on detecting turning maneuvers without a separate turning lane (particularly right-turn maneuvers in countries with right-hand traffic or left-turn maneuvers in countries with left-hand traffic). Studies show that the direction indicator (turn signal) is not suitable as a sole indicator of driver intention due to its inconsistent use in road traffic, as a significant proportion of drivers fail to activate their turn signal despite intending to turn. Therefore, it is necessary to incorporate additional parameters and derived driver behavior characteristics.Specifically, this application uses a driving speed feature, a turn signal feature, and a head turn feature as examples of several driver behavior features to evaluate driver intent, particularly in intersections. The generic approach of the underlying architecture also allows for intention recognition based on other or additional driver behavior features, not only for detecting driver intent in the driver's own vehicle, but also, with appropriate data acquisition and / or data transmission, for detecting driver intent in other vehicles.

[0010] The evaluation of the algorithms for intention recognition based on driver behavior characteristics in intersection areas to detect an intended turning maneuver was carried out in the tests leading to the invention in real traffic for five representative intersections.

[0011] The invention is based on the following principles, which are known per se: Using digital (preferably high-precision) maps and GPS data for determining the vehicle's position, which are already available through in-vehicle navigation systems, possible, immediately ahead, drivable route sections (segments of a digital map) can be determined in advance. Based on the vehicle's position and the predetermined route sections, firstly, a vehicle's self-localization on these route sections is possible. Secondly, all currently available driving maneuvers that must be finally decided upon at a reference point in the near future can be determined. The determination of all available driving maneuvers can be achieved, for example, using a so-called hypothesis tree.

[0012] According to the invention, based on the determination of the driving maneuvers available for selection, the driver's intention for a specific one of these driving maneuvers is determined as follows: Predefined driver behavior characteristics are evaluated for each available driving maneuver. For each driver behavior characteristic, a measure of the probability of occurrence of this driving maneuver is determined by determining the degree of agreement between each observed driver behavior characteristic and the empirically determined driver behavior characteristic typical for the respective driving maneuver. Based on the recorded parameter "turn signal," a driver behavior characteristic is a specific turn signal characteristic.

[0013] Preferably, in order to predict a specific driving maneuver, an absolute probability of occurrence is determined for each driving maneuver available for selection by calculating a (preferably weighted) mean value from the measures of all driver behavior characteristics for each driving maneuver available for selection.

[0014] The following driver behavior characteristics can be evaluated in a particularly advantageous manner alone or in any combination with one another or with other driver behavior characteristics: Based on the recorded parameter driving speed, a driver behavior characteristic is a specific driving speed characteristic and based on the recorded parameter driver head rotation, a driver behavior characteristic is a specific head rotation characteristic.

[0015] First, typical driver behavior characteristics are empirically determined for each possible maneuver and stored as target characteristics. During vehicle operation, the current driver behavior characteristics are observed and compared as actual characteristics with the target characteristics. This determines the degree of agreement and thus the aforementioned metric. The metric is preferably a standardized number, for example, a value between 0 and 1.

[0016] While the invention is not limited to predicting a turning maneuver, an advantageous further development is particularly focused on predicting a turning maneuver. In this context, particular importance is attached to the three driver behavior characteristics mentioned above, whereby insights into the evaluation of the driver behavior characteristics can depend on the basic idea of ​​the invention as well as represent independent ideas: Each measure of the probability of occurrence of a specific driving maneuver is determined depending on certain distances (distances) to a reference point, whereby the reference point is approximately assigned to the actual occurrence of the specific driving maneuver (location or time of implementation of the driver's intention in reality).

[0017] In particular, to predict a turning maneuver when a predetermined degree of head rotation (or viewing angle) is exceeded at a certain distance from a reference point assigned to the turning maneuver onset, a comparatively high measure of the probability of occurrence of this turning maneuver is assigned.

[0018] In addition, in order to predict a turning maneuver when the turn signal is not activated, even at a comparatively short distance from a reference point assigned to the turning maneuver occurrence, the measure for the probability of occurrence of this turning maneuver is set greater than zero.

[0019] The drawing shows an embodiment of the invention, which is intended to explain the invention in more detail. It shows Fig. 1 a schematic representation of the general overall concept of the invention, Fig. 2 a schematic diagram to explain the procedure for determining the driving maneuvers available for selection, Fig. 3a and Fig. 3b shows an example of the inventive determination of a measure for the probability of occurrence, in particular of a turning maneuver, based on a turn signal feature and Fig. 4a and Fig. 4b shows an example of the inventive determination of a measure for the probability of occurrence, in particular of a turning maneuver, based on a head rotation feature.

[0020] In Fig. 1 shows an electronic control unit 1 of a motor vehicle that may be part of a driver assistance system. The control unit 1 records the vehicle speed v, the turn signal signal b, and the driver's head rotation angle k as input signals for observing and evaluating, for example, the following driver behavior characteristics: driving speed characteristic VV, turn signal signal characteristic BV, and head rotation characteristic KV. Typical empirically determined driver behavior characteristics VVt, BVt, KVt are stored in the control unit 1 for each driving maneuver, here, for example, driving straight ahead H1, turning right at the next opportunity H2, and turning right at the next but one opportunity H3.

[0021] The control unit 1 comprises a first evaluation unit 3, by means of which the degree of agreement of each observed driver behavior characteristic VV, BV, KV with the empirically determined driver behavior characteristic VVt, BVt, KVt typical for the respective driving maneuver H1, H2 and H3 can be determined in the form of a measure SCOREv, SCOREb, SCOREk for the probability of occurrence of this driving maneuver.

[0022] The control unit 1 comprises a second evaluation unit 5, by means of which the absolute probability of occurrence P1, P2, P3 for each driving maneuver H1, H2, H3 available for selection can be determined by averaging 4 the metrics SCOREv, SCOREb, SCOREk of all driver behavior characteristics VV, BV, KV for each driving maneuver H1, H2, H3 available for selection.

[0023] The absolute probabilities of occurrence, here P1 for maneuver H1, P2 for maneuver H2, and P3 for maneuver H3, are preferably determined by calculating a preferably weighted mean gM1, gM2, gM3 from the metrics SCOREv, SCOREb, SCOREk of all driver behavior characteristics VV, BV, KV for each available maneuver H1, H2, H3. For this purpose, each metric can be assigned a fixed weighting W1, W2, W3.

[0024] Different driver assistance responses (ADAS1 or ADAS2) can be assigned to different absolute probability values, e.g., P1=10%, P2=40%, P3=30%, or P1=5%, P2=90%, P3=5%. For example, in the first case, where clear intent recognition is not possible, no driver assistance response or only a visual warning display can be activated as a first-level ADAS1 driver assistance response. For the second case, where clear intent recognition is possible, a haptic warning or a delaying intervention can be activated as a second-level ADAS2 driver assistance response.

[0025] The Fig. The exemplary embodiment illustrated in Figure 1 shows an inventive concept that can be applied to any number of driving maneuvers and driver behavior characteristics. The three driver behavior characteristics mentioned here are used only in advantageous embodiments of the invention. The use of the driver behavior characteristics depends, for example, on the availability of sensors for detecting a specific driver behavior in the vehicle.

[0026] In Fig. 2 shows an exemplary current section of a digital map with the route sections S1 to S6 immediately ahead of a vehicle F1. The vehicle F1, which here is to have a device implementing the method according to the invention (in particular control unit 1 with a corresponding software program module) on board, can perform a self-localization 2 ( Fig. 1) by positioning data Pos F1 and the route section data S1 to S6 of the digital map. In the illustrated embodiment, these signals are Fig. 1 also input signals of the control unit 1. After the self-localization 2, all resulting driving maneuvers available for selection - here H1, H2, H3 - are determined.

[0027] In Fig. 2 shows reference points R1 to R3, at which a final decision of the driver for a specific driving maneuver H1, H2 or H3 will be made. The predictive determination of the driver's intention at these reference points R1 to R3 using the metrics SCOREv, SCOREb, SCOREk from the driver behavior characteristics VV, BV, KV is carried out at defined distances D from these reference points R1 to R3 (see also examples according to Fig. 3a to 4b).

[0028] A reference point can also be generally defined as a “fork point” to lane alternatives from which corresponding driving maneuvers result.

[0029] For details of a particularly advantageous evaluation of the driving speed v in connection with the formation of a measure SCOREv, for example for the probability of occurrence of a turning maneuver (e.g. H2), reference is made to the applicant's unpublished patent application 10 2013 200 724.8.

[0030] The following comments on the Fig. 3a and Fig. 3b represent a method according to the invention. The explanations regarding the Fig. 4a and Fig. 4b are embodiments of advantageous developments of the invention.

[0031] In connection with the Fig. 3a and Fig. 3b details of a possible evaluation of the indicator signal b to determine a measure SCOREb depending on a typical indicator signal characteristic BVt for turning maneuvers are explained. The driving maneuver H2 (right-turn maneuver at the next opportunity, see also turning from section S1 into section S3 according to Fig. 2) and the driving maneuver H3 (right-turn driving maneuver at the next but one possibility, see also turning from section S4 into section S6 according to Fig. 2) are examples of the applicability of this indicator signal feature BVt.

[0032] According to Fig. 3a, the SCOREb measure is determined depending on the location where the indicator b is activated. For example, a typical indicator signal characteristic BVt was determined by mixed distribution of study results and statistics, which represents a density function of the driver's activation of the indicator for different distances D from a reference point - here R2 or R3. The SCOREb measure - here, for example, for the driving maneuver H2 or for the driving maneuver H3 - is then calculated using the density function according to Fig. 3a determines when an indicator b is activated.

[0033] If a turn signal is not activated or a turn signal b is not present, the measure SCOREb is preferably calculated here, as an example for the driving maneuver H2, by a distribution function according to Fig. 3b. This takes into account that some drivers do not activate their turn signals even when intending to perform a turning maneuver. Therefore, even in the absence of a turn signal b, a low minimum value greater than zero is preferably assigned for this turning maneuver (here, for example, H2) at a correspondingly defined short distance D [m] of vehicle F1 to the reference point – here R2, the beginning of road section S6 (=right-turn lane). Fig. Figure 3b, for example, shows a minimum SCOREb of 0.2 at a distance of approximately D = 7 m from the reference point R2. It should be noted again that all metrics are preferably normalized between 0 and 1.

[0034] With regard to the two turning maneuvers H2 and H3 that are available for selection and follow one another in quick succession according to Fig. 2 can be calculated from the density function according to the indicator signal characteristic BVt according to Fig. 3a, for a turn signal at a comparatively short distance D (e.g., 20 m) from the reference point R2, a significantly higher SCOREb value is set for the driving maneuver H3 than for the driving maneuver H2 when the reference point R3 is at a distance D of approximately 40-60 m. This assessment is based on the finding that drivers who usually activate their turn signals do so early.

[0035] In principle, due to the realization that some drivers do not activate their turn signals even when turning, the use of the turn signal feature BV alone for detecting the driver's intention has not always proven to be reliable enough.

[0036] Head rotation feature BV as an example of a particularly advantageous driver behavior feature: Details of a possible evaluation of the head rotation k are presented in connection with the Fig. 4a and Fig. 4b is explained as follows:

[0037] Fig. Figure 4a shows a head rotation feature KVt for determining a measure SCOREk for the probability of a straight-ahead driving maneuver—here, maneuver H1. It has been found that during a straight-ahead driving maneuver, a comparatively small, almost always constant head rotation angle k—of approximately 7°—is present, even at a comparatively short distance D (e.g., 15 m) from the reference point (here, R1 for maneuver H1).

[0038] Fig. Figure 4b shows a head turn feature KVt for determining a measure SCOREk for the probability of a turning maneuver – here maneuver H2 or H3. According to Fig.4b, in connection with the head rotation feature KVt, the head rotation angle k is again strongly dependent on the location, i.e., dependent on the distance D of the vehicle F1 from a relevant reference point - here R2 or R3. At a distance D of 5 m, 10 m, 15 m, and 45 m from the reference point (here, for example, R2 or R3) for a turning maneuver (here, for example, H2 or H3), strongly distinguishable head rotations k of approximately 70°, 50°, 30°, and 10°, respectively, result.

[0039] Due to these well-distinguishable head rotation characteristics KVt for a straight-ahead driving maneuver on the one hand and for a turning driving maneuver on the other hand, the determination of the measure SCOREk for a specific driving maneuver at least from these two driving maneuvers available for selection has proven to be a comparatively very reliable driver behavior characteristic.

[0040] The direction of gaze is considered an analogous parameter to head rotation.

[0041] The invention can be used not only to determine the prediction of a specific driving maneuver of the own vehicle / driver, but also of another road user in the relevant environment of the own vehicle. To do this, the parameters for evaluating the driver behavior characteristics of another road user would have to be recorded via Car2X communication and / or environment detection sensors in the own vehicle.

Claims

[1] Method for predicting a specific driving maneuver (H2) from all currently available driving maneuvers (H1, H2, H3) by evaluating predetermined driver behavior characteristics (VV, BV, KV) for each available driving maneuver (H1, H2, H3), whereby for each driver behavior characteristic (VV, BV, KV) for each driving maneuver (H1; H2; H3) a measure (SCOREv, SCOREb, SCOREk) for the probability of occurrence of this driving maneuver is determined by determining the degree of agreement of each observed driver behavior characteristic (VV, BV, KV) with the empirically determined driver behavior characteristic (VVt, BVt, KVt) typical for the respective driving maneuver for the respective driving maneuver (H1; H2; H3) characterized bythat a predetermined driver behavior characteristic (BV) is formed on the basis of the detected parameter indicator signal (b), wherein a minimum measure greater than zero (SCOREb=0.2) for this turning maneuver (H2) is assigned to a lack of the indicator signal (b) at a correspondingly defined short distance (D) of the vehicle (F1) to a reference point (R2) for a turning maneuver (H2). [2] Method according to claim 1, characterized by that in order to predict a specific driving maneuver (H2), an absolute probability of occurrence (P1, P2, P3) is determined for each driving maneuver available for selection (H1, H2, H3) by forming a mean value (gM1, gM2, gM3) from the measures (SCOREv, SCOREb, SCOREk) of all driver behavior characteristics (VV, BV, KV) for each driving maneuver available for selection (H1, H2, H3). [3] Method according to one of the preceding claims, characterized bythat a predetermined driver behavior characteristic (VV) is formed on the basis of the recorded parameter driving speed (v). [4] Method according to one of the preceding claims, characterized by that a predetermined driver behavior characteristic (KV) is formed on the basis of the recorded parameter head rotation (k) of the driver. [5] Method according to one of the preceding claims, characterized by that when the head is turned (k) by a comparatively high defined angle (70°; 50°; 30°; 10°) at a correspondingly defined distance (5m; 10m; 15m; 45m) of the vehicle (F1) to a turning maneuver reference point (R2; R3), a comparatively high measure (SCOREk) is determined for this turning maneuver (H2). [6] Method according to one of the preceding claims, characterized bythat different driver assistance reactions (FAS1; FAS2) are assigned to different absolute probability values ​​(10%, 40%, 30%; 5%, 90%, 5%). [7] Use of the method according to one of the preceding claims for determining the prediction of a specific driving maneuver of the own vehicle and / or of another road user in the relevant environment of the own vehicle, wherein the parameters for evaluating the driver behavior characteristics of another road user are recorded by Car2X communication and / or by environment detection sensors in the own vehicle. [8] Device for carrying out the method according to one of the claims 1-6, with an electronic control unit (1) which records the parameters (v, b, k) for observing the driver behaviour characteristics (VV, BV, KV) as input signals, in which the typical empirically determined driver behaviour characteristics (VVt, BVt, KVt) for the respective driving manoeuvre (H1; H2; H3) are stored and which comprises a first evaluation unit (3) by means of which the degree of agreement of each observed driver behaviour characteristic (VV, BV, KV) with the empirically determined driver behaviour characteristic (VVt, BVt, KVt) typical for the respective driving manoeuvre (H1; H2; H3) for the respective driving manoeuvre (H1; H2; H3) can be determined in the form of a metric (SCOREv, SCOREb, SCOREk). [9] Device according to the preceding claim, characterized bythat the control unit comprises a second evaluation unit (5) by means of which the absolute probability of occurrence (P1, P2, P3) for each driving maneuver (H1, H2, H3) available for selection can be determined by averaging (4) the measured values ​​(SCOREv, SCOREb, SCOREk) of all driver behavior characteristics (VV, BV, KV) for each driving maneuver (H1, H2, H3) available for selection.

Images