METHOD AND SYSTEM FOR ANALYZING THE CONTROL OF A VEHICLE

Abstract

A method and a system analyze the control of a vehicle having an autonomous driving unit. A change in the driving mode from autonomous driving to manual driving is detected, and at least one driving parameter before and/or after detecting the change is monitored. Based on driving values obtained by the monitoring with respect to the detected change in driving mode, at least one driving quantity quantifying the quality of interplay between the autonomous driving unit and a human driver is determined.

Claims

1. A method for improving an operation of a vehicle having an autonomous driving unit configured for autonomously controlling the vehicle and a control takeover management unit configured for passing over control to a human driver, which comprises the steps of: detecting, via a detector disposed in the vehicle, a change in a driving mode from autonomous driving to manual driving, the detecting step including at least one of: determining a point in time a control takeover request is outputted by the control takeover management unit, determining a point in time the human driver takes over control from the autonomous driving unit, and determining a point in time when the human driver assumes safe control over the vehicle; monitoring, via a sensor disposed in the vehicle, at least one driving parameter during the autonomous driving unit and/or the human driver controlling the vehicle; determining at least one driving quantity quantifying a quality of interplay between the autonomous driving unit and the human driver, via an evaluation unit, based on driving values provided by the sensor upon monitoring the at least one driving parameter during a time interval ending at, starting from or including a point in time determined by the detector; configuring, via a configuring unit disposed in a computation facility and based on the at least one determined driving quantity, a risk management model for assessing an effectiveness and/or safety of control takeover events in the vehicle having the autonomous driving unit and the control takeover management unit; outputting, via a computer system disposed in the computation facility and having the risk management model, assessment data obtained by executing a configured risk management model; and operating the vehicle based on the assessment data.

2. The method according to claim 1, which further comprises determining the at least one driving quantity based on an information regarding a software change and/or hardware change in the control takeover management unit.

3. The method according to claim 1, which further comprises monitoring the at least one driving parameter by determining a time interval between the autonomous driving unit outputting the control takeover request and the human driver taking over control.

4. The method according to claim 1, which further comprises determining the at least one driving quantity based on a time interval between the autonomous driving unit outputting the control takeover request and the human driver taking over control, and on a vehicle dynamic variable during the time interval.

5. The method according to claim 4, which further comprises determining the at least one driving quantity based on a product of the time interval and the vehicle dynamic variable during the time interval, wherein the product reaches or exceeds a threshold.

6. The method according to claim 1, which further comprises storing the driving values obtained during the monitoring for a predetermined storage time interval spanning at least from a point in time before the control takeover request is outputted by the autonomous driving unit to the point in time the autonomous driving unit outputs the control takeover request.

7. The method according to claim 1, wherein the at least one driving quantity is a composite quantity.

8. The method according to claim 1, which further comprises determining the at least one driving quantity based on an autonomous driving quantity quantifying a quality of driving of the autonomous driving unit before the change in the driving mode and on a manual driving quantity quantifying a quality of a driving of the human driver after the change in the driving mode.

9. The method according to claim 8, wherein the at least one driving quantity is further based on mileage accumulated in a respective driving mode.

10. The method according to claim 1, which further comprises determining the at least one driving quantity based on an autonomous driving quantity quantifying a quality of the driving of the autonomous driving unit after the control takeover request is outputted by the autonomous driving unit and on a manual driving quantity quantifying a quality of a driving of the human driver subsequently to him taking over control.

11. The method according to claim 1, wherein the at least one driving quantity is further based on accident rate values for autonomous driving or manual driving after the takeover control request is outputted by the autonomous driving unit and subsequently to the human driver taking over control, respectively.

12. The method according to claim 1, which further comprises determining the at least one driving quantity based on driving values obtained by the monitoring during a predetermined first-time interval prior to the autonomous driving unit outputting the control takeover request.

13. The method according to claim 1, wherein the at least one determined driving quantity is interrelated to a quantity distribution of a same driving quantity.

14. The method according to claim 1, which further comprises determining the at least one driving quantity based on a quantity distribution, the quantity distribution being determined based on driving value data provided by a plurality of vehicles.

15. The method according to claim 1, wherein the driving value data is generated by driving value statistic units provided in each of a plurality of vehicles, the driving value statistic units monitoring at least one driving parameter of each of the plurality of vehicles and transmitting corresponding driving values to a server.

16. The method according to claim 1, wherein the monitoring of the at least one driving parameter includes monitoring at least: vehicle speed; vehicle acceleration; and vehicle deceleration.

17. A method for improving operation of a vehicle having an autonomous driving unit configured for autonomously controlling the vehicle and a control takeover management unit configured for passing over control to a human driver, which comprises the steps of: detecting, via a detector disposed in the vehicle, a change in a driving mode from autonomous driving to manual driving, the detecting step including at least one of: determining a point in time a control takeover request is outputted by the control takeover management unit, determining a point in time the human driver takes over control from the autonomous driving unit, and determining a point in time the human driver assumes safe control over the vehicle; monitoring, via a sensor disposed in the vehicle, at least one driving parameter during the autonomous driving unit and/or the human driver controlling the vehicle; determining at least one driving quantity quantifying a quality of interplay between the autonomous driving unit and the human driver, via an evaluation unit, based on driving values provided by the sensor upon monitoring the at least one driving parameter during a time interval ending at, starting from or including a point in time determined by the detector; performing, via a computer, based on the at least one driving quantity, at least one of the following: outputting remedial information used in development of the autonomous driving unit; outputting remedial information used in human driver educational purposes; and determining whether manual driving associated with a control takeover event has degraded compared to earlier control takeover events, and outputting a warning if a driving safety associated with the at least one driving quantity falls below a predetermined threshold.

18. A system for improving operation of a vehicle, the system comprising: an autonomous driving unit disposed in the vehicle, said autonomous driving unit configured for autonomously controlling the vehicle; a control takeover management unit disposed in the vehicle, said control takeover management unit being configured for passing over vehicle control to a human driver, a sensor disposed in the vehicle, said sensor configured to monitor at least one driving parameter during said autonomous driving unit and/or the human driver controlling the vehicle; a detector disposed in the vehicle, said detector configured to detect a change in a driving mode from an autonomous driving to manual driving, said detector performing at least one of the following: determining a point in time a control takeover request is outputted by said control takeover management unit, determining a point in time the human driver takes over control from said autonomous driving unit, and including a point in time the human driver assumes safe control of the vehicle; an evaluator configured to determine, based on driving values provided by said sensor upon monitoring the at least one driving parameter during a time interval ending at, starting from or including a point in time determined by said detector, at least one driving quantity quantifying the quality of interplay between said autonomous driving unit and the human driver; a computer system disposed in a computation facility and having a risk management model for assessing an effectiveness and/or safety of a control takeover event in the vehicle having said autonomous driving unit and said control takeover management unit, said computer system configured to output assessment data obtained by means of said risk management model and used for operating the vehicle; and a configuring unit disposed in the computation facility, said configuring unit configured to configure the risk management model based on the at least one driving quantity.

Description

BRIEF DESCRIPTION OF THE FIGURES

[0149] FIG. 1 is an illustration showing an example of a process of transfer of control of a vehicle;

[0150] FIG. 2 is a block diagram showing an example of a method for analyzing the control of a vehicle having an autonomous driving unit;

[0151] FIG. 3 is a graph showing an example of determining a driving quantity based on driving values obtained during a driving period;

[0152] FIG. 4 is an illustration showing an example of determining a driving quantity based on a time interval between an autonomous driving unit outputting a control takeover request and a human driver taking over control; and

[0153] FIG. 5 is an illustration showing an example of a system for analyzing the control of a vehicle comprising an autonomous driving unit.

DETAILED DESCRIPTION OF THE INVENTION

[0154] Referring now to the figures of the drawings in detail and first, particularly to FIG. 1 thereof, there is shown an example of a process of transfer of control of a vehicle 1 from an autonomous driving unit of the vehicle 1 to a human driver, wherein the thick black arrow indicates time t. The transfer of control is grouped into four stages A, B, C, D each representing a different phase of the transfer of control. The different stages A-D are separated by dotted lines.

[0155] In the first stage A, the vehicle 1 is under autonomous control. When the autonomous driving unit realizes that an upcoming traffic scenario is too complex to safely control the vehicle in the scenario, the autonomous driving unit outputs a control takeover request R, for example in form of an optical signal, an acoustic signal, a haptic signal and/or the like. From the point in time at which the request is outputted (also referred to as the control takeover request event R) up to the point in time at which the human driver actually takes over control (also referred to as the control takeover event T), the autonomous driving unit still controls the vehicle 1. This period corresponds to the second stage B. During this stage, the control of the vehicle 1 by the autonomous driving unit is key for the situation in which the human driver finds himself in when taking over control at the control takeover event T.

[0156] After the human driver has taken over control, the third stage C begins. During this stage, the human driver controlling the vehicle 1 adjusts to the current traffic scenario, for example by executing a particular driving maneuver such as braking (decelerating) or evading (steering). The reactions of the human driver during this stage may be highly dependent on the driving behavior of the autonomous driving unit in the previous stages, in particular in the second stage B. Usually, adjusting to the current traffic scenario, i.e. thoroughly grasping the current situation and performing the necessary maneuvers, takes up to 30 seconds, in particular up to 60 seconds. For this reason, the third stage C preferably extends from the control takeover event T for a predetermined second time interval t2.

[0157] In the fourth stage D, starting at the end of the predetermined second time interval t2, the human driver still controls the vehicle 1. Because a significant amount of time since taking over control has passed, no impact of the driving behavior of the autonomous driving unit, in particular driving decisions made by the autonomous driving unit, on the driving of the human driver is to be expected.

[0158] In order to evaluate the driving quality of the driving of the autonomous driving unit, in particular in the first and second stage A, B, and/or the driving quality of the human driver, in particular in the third and fourth stage C, D, driving quantities Q1, Q2, Q3, Q4 may be derived from a plurality of driving values obtained by monitoring different driving parameters. For example, vehicle speed, vehicle acceleration, vehicle deceleration and erratic driving may be monitored and taken as the basis to determine the respective driving quantity Q1-Q4, wherein erratic driving is defined as the amount of acceleration and deceleration during a defined time interval. In FIG. 1, the driving quantities Q1-Q4 are indicated by horizontal bars extending through the different stages A-D.

[0159] Preferably, to evaluate the quality of interplay between the autonomous driving unit and the human driver, e.g. to rate how well the autonomous driving unit controls the vehicle prior to handing over control to the human driver and/or how well the human driver copes with the traffic scenario the autonomous driving unit maneuvered him into, each of the driving quantities Q1-Q4 may be combined across at least two of the different stages A-D into a composite driving quantity. Alternatively or additionally, a driving quantity for a particular stage A-D may be at least partially based on driving values obtained in a different stage. For example, determining an autonomous driving quantity Q1-Q4 quantifying the quality of driving of the autonomous driving unit during the second stage B may be determined by considering not only driving values obtained by monitoring a driving parameter during the second stage B, but also driving values obtained by monitoring the same driving parameter during at least part of the preceding first stage A, this part being indicated by reference numeral A′. In particular, the autonomous driving quantity Q1-Q4 may be determined based on driving values obtained during a predetermined first-time interval t1 extending from the control takeover request event R back into the first stage A, i.e. covering driving values obtained prior to outputting the control takeover request R.

[0160] FIG. 2 shows an example of a method 2 for improving operation of a vehicle having the autonomous driving unit and a control takeover management unit. In step S1, a change in driving mode from autonomous driving to manual driving is detected. For example, a point in time may be determined at which the autonomous driving unit outputs a control takeover request signal, and a point in time may be determined at which a human driver takes over control.

[0161] In a further step S2, at least one driving parameter as speed, acceleration, deceleration and/or the like is monitored. This monitoring may occur before and/or after the change in driving mode detected in step S1, i.e. during autonomous driving and/or manual driving.

[0162] The driving values obtained by the monitoring are preferably grouped into different stages of the change in control (see FIG. 1). Additionally or alternatively, the driving values are stored at least for a predetermined storage time interval.

[0163] Based on the obtained driving values, at least one driving quantity quantifying the quality of interplay between the autonomous driving unit and the human driver is determined in a further step S3. Therein, the driving quantity is determined with respect to the detected change in driving mode, e.g. by separately evaluating driving values obtained during at least two of the different stages of the change in control and comparing the evaluation results, or by differently weighting driving values from different stages.

[0164] In a further step S4, a risk management model is modified based on the determined driving quantity. This risk management model may be hosted by a computer system and designed for assessing the effectiveness and/or safety of control takeover events in a vehicle comprising an autonomous driving unit and a control takeover management unit. For example, the risk management model may be initialized based on the determined driving quantity, i.e. parameter values may be determined based on the determined driving quantity. Alternatively or additionally, the risk management model may be updated based on the determined driving quantity, i.e. the structure and/or mathematical functions of the model may be altered based on the determined driving quantity.

[0165] In a further step S5, assessment data may be outputted by the risk management model. To this end, the model may be executed on or by the computer system. The assessment data may be indicative of the security and/or safety of the control takeover event(s) associated with the autonomous driving unit and the human driver. For example, the assessment data may comprise a rating value allowing a comparison to other combinations of autonomous driving units and human drivers. Alternatively, the executing the risk management model includes this comparison, and the assessment data is indicative of an insurance premium. Accordingly, the assessment data is used for operation of the vehicle in further step S6.

[0166] In another variant of the method 2, in step S4 a remedial information used in development of the autonomous driving unit and/or used in human driver educational purposes is determined and outputted in step S5. Alternatively or additionally, in step S6 it is determined whether manual driving associated with a control takeover event has degraded compared to earlier control takeover events, and a warning is outputted if a driving safety associated with the determined at least one driving quantity falls below a predetermined threshold.

[0167] FIG. 3 shows an example of determining a driving quantity based on driving values obtained during a driving period. In particular, FIG. 3A schematically indicates how an autonomous driving quantity quantifying the quality of driving of the autonomous driving unit is determined, while FIG. 3B schematically indicates how a manual driving quantity quantifying the quality of driving of a human driver is determined. To eventually rate the overall combined driving of the entity comprising both the autonomous driving unit and the human driver, the autonomous and the manual driving quantity may be combined into a composite driving quantity.

[0168] Both the autonomous and manual driving quantities are based on driving values which are obtained by monitoring a driving parameter, for example speed, during autonomous or manual operation of the vehicle, respectively. These driving values are preferably collected during a driving period, wherein a driving period may correspond to a single journey or a plurality of journeys of the vehicle, in particular during a predetermined time span as ten days, a month or quarter of the year. The driving values may be obtained with respect to the current location of the vehicle and compared to a driving value associated with the location, e.g. a statutory speed limit, an average speed or the like. The result of the comparison, e.g. a difference, may give a driving quality value.

[0169] Thus, each monitoring of the driving parameter during a part of the driving period, e.g. during separate journey “events”, in particular with regard to different stages of the change in driving mode of the vehicle between autonomous control of the autonomous driving unit and the human driver, results in a particular quality value distribution Vi of driving quality values, the index i=1 . . . n indicating the ith part of the driving period or the ith event or stage, respectively.

[0170] From each of these quality value distributions Vi, a normal distribution N may be obtained, in particular by averaging each of the quality value distributions Vi. This reflects the central limit theorem according to which a large number of independent random variables (the driving values) asymptotically forms a stable distribution, in particular the normal distribution.

[0171] From the normal distribution N, the autonomous (FIG. 3A) or manual (FIG. 3B) driving quantity may be determined, respectively. In particular, the driving quantity may be determined based on a property of the normal distribution N. For example, the driving quantity may be associated with the expected value of the normal distribution N. Additionally or alternatively, it is also conceivable that the driving quantity is associated with or determined based on another property, as e.g. the variance or the full width at half maximum (FWHM), or possibly a combination thereof.

[0172] In particular, the driving quantity may be determined based on a comparison of a property of the normal distribution N with the same property of an averaged distribution, the averaged distribution being preferably based on a plurality of normal distributions obtained from a plurality of driving periods of different vehicles driven by an autonomous driving unit or different human driver, respectively. For example, if the expected value of the average distribution with regard to vehicle speed during autonomous driving amounts to a first speed and the expected value of the determined normal distribution N with regard to vehicle speed during autonomous driving amounts to a second speed, the driving quantity may be determined as the difference between the first and the second speed.

[0173] FIG. 4 shows an example of determining a driving quantity based on the time interval Δt between an autonomous driving unit outputting a control takeover request R (also referred to as control takeover request event R) and human driver 3 taking over control of a vehicle 1 (also referred to as control takeover event T). By basing the driving quantity on the time interval Δt, it is possible to quantify the driver's degree of attentiveness as well as the effectiveness of the design of the critical man-machine aspects of the autonomous driving unit particularly with respect to alerting the driver 3 for imminent transfer of control.

[0174] The time interval Δt may be determined as the time lag between the control takeover request event R and the detection of a control signal generated by an interaction of the human driver 3 with the vehicle 1, in particular the control system of the vehicle 1. For example, the time interval Δt may be the time lag between the control takeover request event R and a contact being detected between the driver's hands and the steering wheel, a pressure exerted on a pedal being detected, a change in posture of the driver being detected, and/or the like. Accordingly, the driving quantity may be a time lag metric.

[0175] Preferably, the driving quantity is determined based further on a vehicle speed the vehicle 1 exhibits during the time interval Δt. In particular, the time interval Δt may be weighted by the vehicle speed. This results in a distance d the vehicle 1 covers between the control takeover request event R and the control takeover event T. Because the vehicle speed may change during the time interval Δt, it is preferred to base the driving quantity on an average speed.

[0176] Further preferably, the change in driving mode may be grouped for a particular locale as e.g. urban, rural or highway. This may help to obtain a more differentiated assessment of the intertwined impact of both handover alert design and driver alertness, because the speed of the vehicle 1 during control handovers in urban areas is significantly lower than during control handovers on a highway.

[0177] From a plurality of such changes in driving mode, occurring for example during a driving period of a plurality of journeys of the vehicle 1 e.g. over ten days, a month or a quarter of a year and/or with respect to a particular locale category, a distribution of the weighted time interval, i.e. the distance d, may be obtained (see FIG. 3). The driving quantity is preferably associated with a property of this distribution, for example its expected value. In particular, the driving quantity may be associated with an average of a particular part of the distribution, e.g. the fourth quartile. In that way, only the 25% worst cases are taken into account, the rationale being here that not too many changes of control are to be expected and thus the focus should be on the most meaningful cases in terms of causation of AI assisted driving accidents.

[0178] This driving quantity may be rated, e.g. by comparing it to a corresponding property of an average distribution of the weighted time interval obtained for e.g. a plurality of other vehicles or from records of the same vehicle and/or driver. Alternatively, the driving quantity may be determined by the comparison, e.g. as the difference between the property of the distribution of the present vehicle 1 and the corresponding property of the average distribution obtained from the plurality of other vehicles or records, respectively. If the driving quantity, e.g. said difference, exceeds a certain threshold, the driver 3 may be informed of a takeover attention deficit and/or autonomous navigation, at least for a certain locale category as urban or rural, may be suspended. Alternatively or additionally, the autonomous driving unit or the corresponding autonomous driving solution provider, respectively, may be assigned an increased premium from its insurer to reflect the ineffectiveness of its transfer of control solution relative to that of other autonomous driving solution providers.

[0179] FIG. 4 shows an example of a system 4 for improving operation of a vehicle 1 having an autonomous driving unit 5 and a control takeover management unit 11. The system 4 comprises a detection unit 6 configured to detect a change in the driving mode from autonomous driving by the autonomous driving unit 5 to manual driving by a human driver and a sensor unit 7 configured to monitor at least one driving parameter before and/or after the change was detected. Further, the system 4 comprises an evaluation unit 8 configured to determine, based on driving values provided by the sensor unit 7 upon monitoring with respect to the detected change driving mode, at least one driving quantity quantifying the quality of interplay between the autonomous driving unit 5 and the human driver. The evaluation unit 8 may further include position data indicative of the vehicle's position provided by e.g. a GPS sensor 9 in the determination of the at least one driving quantity, e.g. to compare a current driving value to a distribution of driving values obtained for other vehicles at the same position or to a statutory speed limit. In the present example, the evaluation unit 8 has access to a map database 10 containing map data, which provides e.g. information on such statutory speed limits with respect to location.

[0180] Additionally, the map data may provide information on GPS signal quality. The GPS signal quality may be considered by the evaluation unit 9 with regard to the determination of the at least one driving quality. If e.g. the GPS signal is indicated as weak and thus considered unreliable, accelerometer of an accelerometer of the vehicle 1 or a mobile device carried along in the vehicle 1 may be used instead to determine the position of the vehicle 1.

[0181] In the shown example, the evaluation unit 8 is configured as a central evaluation unit 8, e.g. a software module running on a processing unit communicatively coupled to the vehicle 1 by means of a wireless connection, for example via the Internet. However, in another embodiment (not shown), the evaluation unit 8 may be arranged in or be part of the vehicle 1, respectively.

[0182] The evaluation unit 8 may be part of a computation facility or data processing center. Particularly, the evaluation unit 8 may be established or hosted by a computer system 14 of the computation facility or data processing center. The computer system 14 preferably also has a risk management model 12 configured for assessing the effectiveness and/or safety of a control takeover event in a vehicle having an autonomous driving unit and a control takeover management unit and is configured to output to output assessment data obtained by means of the risk management model and used for operation of the vehicle. The risk management model 12 is preferably configurable by means of a configuration unit 13, which might also be a part of the computer system 14.

[0183] The evaluation unit 8 may evaluate the driving values provided by the sensor unit 7 during driving periods, e.g. a single journey or a plurality of journeys, with respect to the driving mode, in particular one of different stages A, A′, B, C, D of the change in driving mode (see FIG. 1). From the driving values, the evaluation unit 8 may determine a normal distribution, at least one property of which may be used to derive the driving quantity (see FIG. 2). In particular, the evaluation unit 8 may evaluate data in form of driving values obtained from a plurality of different vehicles as well or have at least access to a databank containing these data, and determine the driving quantity based on a comparison of at least one property of the normal distribution obtained for the present vehicle 1 and corresponding distributions obtained for the plurality of other vehicles using normal or non-parametric methods of inference.

[0184] In particular, the evaluation unit 8 may be configured to determine, based on the driving values provided by the sensor unit 7, an autonomous driving quantity quantifying the driving of the autonomous driving unit 5 during a first stage A and/or a second stage B of the transfer in control and a manual driving quantity quantifying the driving of the human driver during a third stage C and/or a fourth stage D of the transfer in control (see FIG. 1). The autonomous and manual driving quantity may then be combined into the driving quantity quantifying the quality of the interplay between the autonomous driving unit 5 and the human driver, i.e. the quality of driving of the entity comprising both the autonomous driving unit 5 and the human driver.

[0185] For example, the evaluation unit 8 may be configured to determine a combined driving quantity Q.sub.C+D=Q′.sub.M for manual driving during stages three (C) and four (D) of the change in driving mode as follows:

[0186] When

[00001]$Q_D>Q_C\mathrm{then}{Q}_M^{\prime }=Q_D+{\%}_{\mathrm{Acc}C}.Math.\left[Q_D+\frac{S_D^2}{S_C^2}\right].Math.\left(Q_C-Q_D\right),$

and when

[00002]$Q_D<Q_C\mathrm{then}{Q}_M^{\prime }=Q_D+{\%}_{\mathrm{Acc}C}.Math.\left[Q_D+\frac{S_C^2}{S_D^2}\right].Math.\left(Q_C-Q_D\right).$

Therein, Q.sub.D and Q.sub.C are the driving quantities quantifying the quality of driving during the fourth and third stage D and C, respectively, and %.sub.Acc C is the accident rate value for, e.g. nominal percentage of accidents occurring during, the third stage C. Further, S.sub.D.sup.2 and S.sub.C.sup.2 are the variances of the normal distributions determined from the driving values for driving during the fourth and third stage D and C, respectively (see FIG. 3).

[0187] Similarly, the evaluation unit 8 may be configured to determine a combined driving quantity Q.sub.A+B=Q′.sub.A for autonomous driving during stages one (A) and two (B) of the change in driving mode as follows:

[0188] When

[00003]$Q_A>Q_B\mathrm{then}{Q}_A^{\prime }=Q_A+{\%}_{\mathrm{Acc}B}.Math.\left[Q_A+\frac{S_A^2}{S_B^2}\right].Math.\left(Q_B-Q_A\right),$

and when

[00004]$Q_A<Q_B\mathrm{then}{Q}_A^{\prime }=Q_A+{\%}_{\mathrm{Acc}B}.Math.\left[Q_A+\frac{S_A^2}{S_B^2}\right].Math.\left(Q_B-Q_A\right).$

Therein, Q.sub.A and Q.sub.B are the driving quantities quantifying the quality of driving during the first and second stage A and B, respectively, and %.sub.Acc B is the accident rate value for, e.g. the nominal percentage of accidents occurring during, the second stage B. Further, SA and SB are the variances of the normal distributions determined from the driving values for driving during the first and second stage A and B, respectively.

[0189] From the combined driving quantities for autonomous driving Q.sub.A+B=Q′.sub.A and manual driving Q.sub.C+D=Q′.sub.M, the composite driving quantity Q.sub.A+B+C+D quantifying the quality of interplay between the autonomous driving unit 5 and the human driver may then be obtained as follows:

[00005]$\mathrm{When}{Q}_M^{\prime }>Q_A^{\prime }\mathrm{then}{Q}_{A+B+C+D}=Q_M^{\prime }.Math.{\%}_M+{\%}_A.Math.\left[Q_M^{\prime }+\frac{S_D^2}{S_A^2}\right].Math.\left(Q_A^{\prime }-Q_M^{\prime }\right),$$\mathrm{and}\mathrm{when}{Q}_M^{\prime }<Q_A^{\prime }\mathrm{then}{Q}_{A+B+C+D}=Q_M^{\prime }.Math.{\%}_M+{\%}_A.Math.\left[Q_M^{\prime }+\frac{S_A^2}{S_D^2}\right].Math.\left(Q_A^{\prime }-Q_M^{\prime }\right).$

[0190] Therein, %.sub.M and %.sub.A are the relative mileages accrued under manual and autonomous control, respectively.

[0191] Alternatively or additionally, the sensor data of the GPS sensor 9 and/or the GPS signal quality provided by the map database 10 may be used to determine the at least one driving quantity as follows: if the GPS signal is weak or indicated as weak, e.g. due to the vehicle passing through a tunnel, the driving values may be obtained by an extrapolation of previously obtained driving values. In particular, the speed of the vehicle 1 may be determined based on the locations of strong or at least reliable GPS signal, e.g. ahead and behind the tunnel, and the time needed by the vehicle 1 to travel between these locations. If speeding is detected this way, this may have a direct effect on the determined driving quantity.

[0192] Alternatively or additionally, the evaluation unit 8 may be configured to interrelate the driving quantity, in particular for the different stages A-D of the change in driving mode, to a quantity distribution of the same driving quantity obtained from a plurality of other vehicles with an autonomous driving unit and different human drivers or from records of the present vehicle 1. By this means, it becomes possible for the evaluation unit 8 to assess the responsibility of either the autonomous driving unit 5 or the human driver in the case of an accident during one of the stages A-D of the change in driving mode. In other words, the determined driving quantity may be used as a forensic tool in case of accidents occurring during the change in driving mode, in particular during the second or third stage B, C.

[0193] For example, if a driving quantity quantifying the quality of driving of the autonomous driving unit 5 during a part A′ of the first stage A, e.g. a first time interval prior to a control takeover request event, is substantially equal to the average of the same driving quantity obtained for the plurality of other vehicles during the first stage A, this may indicate responsibility of the human driver. That is because substantially equal quantities indicate no deterioration of the autonomous driving and thus a regularly operating autonomous driving unit 5.

[0194] If, however, the driving quantity is higher than the average driving quantity obtained for the plurality of other vehicles during the first stage A, this may indicate deterioration of the autonomous driving, and thus suggest responsibility of the autonomous driving unit. That is, the relative increase of the driving quantity, e.g. due to higher speed, stronger and/or more erratic acceleration and/or deceleration, can be associated with the (unsuccessful) attempt of the autonomous driving unit 5 to cope with the upcoming complex traffic scenario.

[0195] Examples for the different driving quantities which may be determined for the different stages A-D of the change in driving mode are given in the table below as follows:

TABLE-US-00001 Period Driving quantity notation during first stage A Q.sub.A (quantifying the quality of unimpaired autonomous driving) during a predetermined first Q.sub.pre-B (quantifying the quality of time interval prior to the autonomous driving in view of an second stage B (stage A′) upcoming complex traffic scenario) during second stage B Q.sub.B (quantifying the quality of autonomous driving in view of immanent handover) Q.sub.Δt (quantifying the quality of handover alert design and driver alertness, see FIG. 3) during third stage C Q.sub.C (quantifying the quality of manual driving in view of a complex traffic scenario) during fourth stage D Q.sub.D (quantifying the quality of unimpaired manual driving)

[0196] Further, the different driving quantities with exemplary interrelationships or relationships to their average, respectively, as well as the possible inferred responsibility is given in the table below as follows:

TABLE-US-00002 Driving quantity Responsible Q.sub.Δt > Q.sub.Δt, average human driver Q.sub.B ≈ Q.sub.A Q.sub.pre-B ≈ Q.sub.A Q.sub.C > Q.sub.D Q.sub.Δt ≈ Q.sub.Δt, average autonomous driving unit Q.sub.B > Q.sub.A Q.sub.pre-B > Q.sub.A Q.sub.C ≈ Q.sub.D

[0197] This responsibility may be assigned as described in the above table particularly if an accident occurs after the autonomous driving unit has attempted to re-engage the driver, i.e. if the autonomous driving unit has outputted a control takeover request. If, on the other hand, the accident occurs during autonomous driving with no attempt to re-engage the driver, the responsibility may be assigned to the autonomous driving unit or its manufacturer, respectively.

[0198] The following is a summary list of reference numerals and the corresponding structure used in the above description of the invention: [0199] 1 vehicle [0200] 2 method [0201] 3 driver [0202] 4 system [0203] 5 autonomous driving unit [0204] 6 detection unit [0205] 7 sensor unit [0206] 8 evaluation unit [0207] 9 GPS sensor [0208] 10 map database [0209] 11 control takeover management unit [0210] 12 risk management model [0211] 13 configuring unit [0212] 14 computer system [0213] A-D stages of the change in driving mode [0214] A′ part of the first stage [0215] N normal distribution [0216] R control takeover request event [0217] T control takeover event [0218] Vi quality value distribution [0219] d distance [0220] t time [0221] t1, t2 first, second time interval [0222] Δt time interval