Method and apparatus for increasing the proportion of autonomous driving in at least a partially autonomous vehicle

Abstract

The present invention relates to a method for increasing the proportion of autonomous driving in an at least partially autonomous vehicle in which conditions for autonomous driving are monitored, the method comprising the steps of: (a) predicting, using a model, that the autonomous driving will be terminated by manual vehicle control when the conditions for autonomous driving are met; (b) outputting information to the driver informing him that the autonomous driving is in good control of the driving situation, in order to take advantage of the autonomous driving best designed for traffic flow, road safety and energy balance;

Description

Technical Field

[0001] The present invention relates to a method for increasing the proportion of autonomous driving and an apparatus for implementing this method in at least a partially autonomous vehicle in which conditions for autonomous driving are monitored.

Background Art

[0002] German Patent Application Publication No. 10335900 discloses a driver assistance system having a module connected to sensors for monitoring conditions for autonomous driving. In particular, in a partially autonomous (semiautonomous) vehicle, due to the lack of trust and uncertainty of the vehicle driver, the vehicle driver often takes over the control of the vehicle to manual driving. As a result, the traffic flow is negatively affected, which has an adverse effect on traffic safety and the energy balance of the vehicle.

Summary of the Invention

Problems to be Solved by the Invention

[0003] An object of the present invention is to provide a method and an apparatus for increasing the proportion of autonomous driving, in which the number of handovers to manual driving by the driver is reduced in order to utilize the advantages of autonomous driving that are optimally designed in accordance with traffic flow, traffic safety, and energy balance.

Means for Solving the Problems

[0004] The present invention will become apparent from the features of the independent claims. Advantageous developments and embodiments are the subject matter of the dependent claims. Other features, applicability, and advantages of the present invention will become apparent from the following description and the description of the exemplary embodiments of the present invention shown in the drawings.

[0005] This object is solved by the method described at the beginning. That is, when the conditions for autonomous driving are met, it is predicted using a model that the autonomous driving will end by manual driving control, and when a takeover by the driver is predicted, information is output to the driver notifying the driver that the autonomous driving, i.e., the autonomous vehicle, is reliably controlling the driving situation. The driver is informed that there is no need to take over the driving and switch to the manual driving mode. For example, the model is a trained neural network. The model is pre-trained by, for example, a representative driver and installed in the vehicle. Preferably, the model is trained while being individualized for each individual driver so that the manual takeover of vehicle control can be predicted in a driver-specific manner. The model predicts a manual takeover immediately when the probability exceeds a predetermined value. In a particular driving situation, if a manual takeover is rarely performed, the probability is low, and if the takeover is frequent or always performed, the probability is high. Since the driver's trust in autonomous driving is improved, the number of manual takeovers of the driving mode by the driver is reduced. The longer the autonomous driving is, the higher the traffic safety and the better the traffic flow are improved.

[0006] Preferably, the functional quality of autonomous driving is determined by monitoring the conditions for autonomous driving, and when a takeover by the driver is predicted, information is output to the driver only when the functional quality of autonomous driving exceeds a predetermined limit value. Thereby, it is guaranteed that the driver is not affected by his own intention to take over the manual driving mode when the functional quality of autonomous driving is low. In order to support the driver regarding the intention to take over, in such a case, a takeover request can be activated by the vehicle.

[0007] In one embodiment, in the case of an unsolicited manual vehicle takeover and when the conditions for autonomous driving are met, especially when the functional quality exceeds the limit value, the driver receives a notification regarding what the driving operation data would have been in a hypothetical autonomous driving without a manual vehicle takeover. Thereby, the driver can compare the manual driving maneuvers with the simulated autonomous driving maneuvers and check his own decision.

[0008] In one variant, the model is trained with data including driver behavior, autonomous vehicle behavior, driving situation, and / or environmental conditions so that, using the model, the time when the driver takes over manual vehicle control without a request via the vehicle can be predicted. With a large amount of data, the takeover of vehicle control by the driver can be predicted very accurately. In particular, the model is personalized, stored, and evaluated for the driver, that is, the prediction can always be carried out using the model assigned to the current driver.

[0009] In one embodiment, the model is stored in a central server and trained with data from all owned vehicles. For this purpose, each driver of all owned vehicles is classified with respect to their behavior, and an individual model used as a criterion for predicting a manual takeover according to the driving type assigned to a driver is trained with the data of each classification.

[0010] The information is preferably output to the driver by means of functional graphics, voice output, and / or text output. Thereby, it is ensured that the driver is reliably informed, depending on each information format and the amount of data to be transmitted.

[0011] In another embodiment, when a takeover by the driver without a takeover request is predicted, the autonomous driving is adjusted to reduce the takeover frequency by changing the behavior of the autonomous vehicle. At this time, as soon as a manual takeover is predicted with a correspondingly high probability, driving parameters such as speed and inter-vehicle distance are changed iteratively, experimentally, or according to a stored characteristic map. Since the training of the model for predicting manual vehicle takeover continues, the frequency of actual and predicted takeovers, and the driver information associated therewith, also decrease in response to changes in the driving behavior of the vehicle. After the behavior of the autonomous vehicle is changed so that the information output and the change in vehicle behavior reduce the probability of a manual takeover at the same time, the information output to the driver is maintained. In an ideal case, since the driving behavior has been changed and the driver's trust has increased, as soon as the driver stops receiving information due to the reduced probability of a takeover, a manual vehicle takeover is no longer performed in the corresponding driving situation. In another variant, the information output is suppressed after the driving behavior is changed and is activated again as soon as a manual vehicle takeover is performed in the corresponding driving situation despite the change in driving behavior.

[0012] In yet another variant, the driving parameters in a certain vehicle in autonomous driving mode, which are determined to be effective for reducing the takeover frequency, are transmitted to other vehicles through a server. This measure can be used at any time when the driver frequently terminates the autonomous driving mode by a manual takeover without a vehicle request. This measure helps to further strengthen the driver's trust in autonomous driving.

[0013] Another aspect of the present invention relates to an apparatus for increasing the proportion of autonomous driving in at least a partially autonomous vehicle, including a module for monitoring conditions for autonomous driving. In order to take advantage of the benefits of autonomous driving that are optimally designed according to traffic flow, traffic safety, and energy balance, in an apparatus for reducing the number of manual handovers by a driver, a monitoring module stores a model for predicting the intention of a driver to end autonomous driving by manual vehicle control. When a handover by the driver is predicted when the conditions for autonomous driving are met, an action unit, i.e., a control device, outputs to the driver information notifying the driver that autonomous driving, i.e., the autonomous vehicle is reliably controlling the driving situation.

[0014] Preferably, the monitoring module and / or the action module of each vehicle in all owned vehicles are wirelessly connected to a vehicle external server for classifying groups having similar characteristics regarding the handover from autonomous driving to manual driving mode. In this way, models trained for different drivers in all owned vehicles can be uniformly evaluated, and if necessary, output to the vehicle to which the driver belongs to one of these classifications.

[0015] Other advantages, features, and details will become apparent from the following description in which at least one exemplary embodiment is described in detail, optionally with reference to the drawings. The components described and / or illustrated can form the subject matter of the invention, either by themselves or as any logical combination, and in some cases, independently of the claims, and in particular, can also be the subject of one or more separate applications. The same, similar, and / or functionally identical parts are labeled with the same reference numerals.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Best Mode for Carrying Out the Invention

[0017] In FIG. 1, an autonomous vehicle 1 having a device 3 according to the present invention is shown. The device 3 is configured as a vehicle assistant and includes a monitoring module 5 in which a model 7 for predicting the intention of a driver who attempts to terminate the autonomous driving of the vehicle 1 by manual vehicle control is stored. The monitoring module 5 is connected to an action module 9, and when it is predicted that the driver will take over when the conditions for autonomous driving are satisfied, the action module 9 outputs to the driver information notifying that the autonomous driving is reliably controlling the driving situation. To output such information, the action module 9 is connected to a display unit 11 for displaying the information as functional graphics or text, and is also connected to a speaker 13 for audio output. To check whether the conditions for autonomous driving are satisfied by the vehicle 1, a module 15 for monitoring the conditions for autonomous driving, which also communicates with the action module 9, is provided. The action module 9, the module 15, and the monitoring module 5 are each an independent control device or are incorporated into one control device.

[0018] The monitoring module 5 continuously determines the state, attention, and behavior of the driver using sensors 17 such as a camera, a steering sensor, and a biometric sensor of the vehicle 1, and outputs this to the model 7. At the same time, traffic data, weather, a route map, a driving situation, time, etc. are supplied to the model 7. The model 7 is continuously trained by all these data to predict the timing or characteristics of a manual takeover of the driving mode that is carried out autonomously by the driver, i.e., without a request from the vehicle, and this corresponds to the end of autonomous driving. The model 7 uses a statistical method or a machine learning method. Preferably, the model 7 may be configured as a neural network. The model 7 is incorporated into a separate control device or into a control device having at least one of the above-described modules.

[0019] A module 15 for monitoring the automated driving monitors the conditions required for the automated driving with respect to its functional quality. The surroundings of the vehicle 1 are observed by another sensor 19 with respect to, for example, traffic participants, obstacles, intersections, etc. The action module 9 determines the current functional quality of the automated driving based on the monitoring and compares this with a threshold value. If the determined functional quality falls below the threshold value, since the automated driving is no longer safe, a takeover request to start the manual driving mode is automatically output to the driver by the action module 9.

[0020] Since data from the monitoring module 5 and the module 15 are combined in the action module 9 for monitoring the automated driving, the action module 9 activates the actuators necessary for communication with the vehicle driver to compare the manual driving maneuvers with the simulated (partial) automated driving maneuvers. The information is output to the driver for a predetermined period before the predicted takeover time. This period is determined according to the difference between the determined functional quality and the threshold value. That is, the smaller the difference between the functional quality and the threshold value, the shorter the period before the information is output. Thereby, it is possible to ensure that the functional quality of the automated driving still exceeds the threshold value even at the predicted takeover time, thereby always guaranteeing the driving safety.

[0021] When it is predicted that the driver will manually take over the driving mode without a request from the vehicle 1 and when the conditions for the automated driving are met, that is, when the functional quality exceeds the threshold value, the action module 9 outputs to the driver information regarding what the driving operation data were in the hypothetical automated driving without manual takeover. These information may include the distance to be automatically driven without manual intervention, and may also include a guideline regarding the increase in safety when continuing the automated driving in comparison with the manual driving. For this purpose, the detected approach to the preceding vehicle caused by the manual driving, speeding, inattentiveness, etc. can be used.

[0022] In addition to this, the monitoring module 5 is wirelessly connected to an external vehicle server 21 such as an OEM data center or a cloud application, and communicates with a large number of vehicles of all owned vehicles. The monitoring module 5 of each vehicle 1 transmits the data processed by each model 7 and the result of the prediction to the external vehicle server 21. The external vehicle server 21 classifies the drivers of all owned vehicles with reference to these data. These drivers can be evaluated, for example, as sporty, cautious, highly safety-conscious, etc. With reference to this classification, various models 7 suitable for the drivers of each classified driver type are trained.

[0023] In order to modify, optimize, or adjust the rules of autonomous driving, the feedback from the monitoring module 5 to the external vehicle server 21 can be utilized. In particular, the predictive feature values that form the basis of the model 7 used can be extracted. Further, groups with similar driving behaviors during the use of autonomous driving can be extracted from the individual models 7 of the vehicle 1 in all owned vehicles. This makes it possible to develop group-specific characterizations of autonomous driving.

[0024] FIG. 2 shows an exemplary embodiment of the method according to the present invention. During autonomous driving, the method stored in the vehicle assistant (device 3 ) starts at block 100. In block 110, the monitoring module 5 determines the probability p of a manual takeover of the driving operation by the driver. When the high probability p (exceeding the probability threshold p s ) in block 120, and accordingly an imminent manual takeover by the driver is determined, a corresponding signal is transmitted to the action module 9. At the same time, the module 15 for monitoring autonomous driving determines environmental data and determines the functional quality FG of autonomous driving using this environmental data, and these are also transmitted to the action module 9 in block 140. In block 150, the functional quality FG is compared with the limit value G FG . In block 150, the functional quality FG is compared with the limit value G FGIf it is determined that it is below, in block 160, a handover request for manual vehicle control is automatically activated, or the automatic driving ends to prevent a system failure of the automatic driving.

[0025] When the functional quality FG exceeds the limit value G FG , in block 170, communication with the vehicle driver is started by the action module 9. At this time, an appropriate communication path is selected to notify the driver regarding the performance of the automatic driving. Warning symbols such as green lamps and icons on the display instrument 13 can be used, and this can be supplemented additionally by text information. As an addition or alternative thereto, in order to reduce the probability p that the driver intends to manually take over the vehicle operation, the driving behavior of the automatic driving can be adjusted, for example, by reducing the speed or increasing the safety distance from the preceding vehicle.

[0026] In block 180, it is checked whether the manual driving mode has been started. If not, the process returns to block 110. In the case of manual handover, in block 190, the action module 9 calculates how the energy balance, traffic flow balance, and safety balance in the current manual handover differ from the hypothetical continued automatic driving (shadow mode). In block 200, the driver can be notified retrospectively of how the driving performance during the automatic driving should have been. This information can also be used in the future to determine whether communication with the driver should be started depending on the predicted intention of the manual handover, for example, when the difference between the automatic driving and the manual handover regarding the energy balance, traffic flow balance, and safety balance is considered too small. This information can also be used when the probability p determined by the monitoring module 5 is too high and the manual handover cannot be avoided even by using communication. Then, the method ends at block 210.

[0027] Regarding the working mode of the vehicle assistant, the following application examples will be taken and explained in detail. In this application example, the automatic driving is regularly deactivated by individual drivers on the highway before the exit. Each time point ti of the driver's request to shift from automatic driving to manual control by the control of the vehicle assistant is recorded. In addition to this, the characteristic values at each time point ti and in the period tn-i before each time point ti is saved in the monitoring module 5 and the module 15 for monitoring the automatic driving are saved. In the model 7, during training, the recorded characteristic values are combined with the high probability p of the shift request. Here, the model 7 learns, for example, that the highway driving mode and the geographical position immediately before the exit increase the probability of the request for manual shift. And when the vehicle 1 of an individual driver approaches immediately before the exit on the highway, the action module 9 notifies the driver that no manual takeover is required before the vehicle assistant transmits information to the driver. This is done at a point in time before the driver consciously makes a decision on the request for manual takeover by himself. In addition to this, in a driving situation where such a manual takeover is predicted, the vehicle can test various variations of the driving behavior of the vehicle 1 by, for example, selecting a wider or narrower safety distance from the preceding vehicle, or the center lane or the left lane of the highway, or by transmitting additional communication to the surroundings such as traffic participants moving around the vehicle. When the driver changes his behavior so that it can be observed whether the variation leads to a manual takeover continuously or whether the manual takeover can be successfully avoided, it is preferable not to output information to the driver. If a certain variation leads to successfully avoiding the manual takeover, this variation is added to the driving profile of the driving assistant. As a result, the model learns that the probability of manual takeover is low in this modified driving mode, and accordingly, in this driving mode, the prediction of manual takeover and the steps associated therewith are not implemented in the driving situation.

[0028] Such variations in the driving behavior in vehicle 1 can be predefined by the vehicle assistant or can be learned heuristically by the system within predefined limits.

Prior Art Documents

Patent Documents

[0029]

Patent Document 1

Claims

1. A method for increasing the proportion of autonomous driving in a vehicle in which conditions for autonomous driving are monitored and an at least partially automated driving can be executed, comprising: When the conditions for autonomous driving are met, it is predicted using a model that autonomous driving will end by manual vehicle control, and when a takeover by the driver is predicted, information is output to the driver that a takeover to manual driving is not necessary because autonomous driving is reliably controlling the driving situation. A method characterized by the above.

2. A functional quality for evaluating the safety of autonomous driving is determined based on information obtained by a sensor of the vehicle by monitoring the conditions for autonomous driving, and when a takeover by the driver is predicted, the information is output to the driver only when the functional quality exceeds a predetermined limit value of the functional quality and thus the safety of the autonomous driving is guaranteed. The method according to claim 1, characterized by the above.

3. In the case of a manual takeover of the vehicle by the driver without a request, and when the conditions for autonomous driving are met and the functional quality further exceeds the limit value, the driver receives a notification regarding what the driving operation data was in a hypothetical autonomous driving without taking over the vehicle. The method according to claim 2, characterized by the above.

4. The model is trained with data including the behavior of the driver, the behavior of the vehicle, the driving situation, and / or environmental conditions so that the time when the driver takes over the manual vehicle control without a request can be predicted via a vehicle capable of executing the at least partially automated driving using the model. The method according to claim 1, characterized by the above.

5. The model is stored in a central server and is trained with data including the behavior of the driver, the behavior of the vehicle, the driving situation, and / or environmental conditions of all the owned vehicles connected to the server. The method according to claim 1, characterized in that.

6. The information is output to the driver by functional graphics, voice output, and / or text output. The method according to claim 1, characterized in that.

7. The automatic driving is adjusted to reduce the handover frequency by changing the behavior of the vehicle when a handover by the driver is predicted. The method according to claim 1, characterized in that.

8. The driving parameters in a certain vehicle during the automatic driving mode, which are determined to be effective for reducing the handover frequency, are transmitted to other vehicles through the server. The method according to claim 5, characterized in that.

9. An apparatus for increasing the proportion of automatic driving in a vehicle capable of executing at least partially automated driving, including a module (15) for monitoring the conditions for automatic driving, equipped with a monitoring module (5) storing a model (7) for predicting the intention of the driver to end the automatic driving by manual vehicle control, and when a handover by the driver is predicted when the conditions for automatic driving are satisfied, the action module (9) outputs to the driver information notifying the driver that there is no need for a handover to manual driving because the automatic driving surely controls the driving situation. The apparatus, characterized in that.

10. The monitoring module (5) and / or the action module (9) provided in each vehicle (1) in the fully-owned vehicle is wirelessly connected to a vehicle external server (21) for classifying groups having similar characteristics regarding the transfer from automatic driving to manual driving mode The device according to claim 9, characterized in that

Images