Computer-implemented method for operating at least one vehicle

Abstract

The invention relates to a computer-implemented method (100) for operating at least one vehicle (10), comprising at least the following steps: providing (102) at least one estimate (30) indicating a risk of aquaplaning by means of at least one trained machine learning algorithm (28) based on at least one data set (34) comprising at least ultrasonic sensor data (36) from at least one ultrasonic sensor (14) for determining a road surface type in front of the at least one vehicle (10) and / or at least laser sensor data (38) from at least one laser sensor (16) for detecting at least one road groove (24) in front of the at least one vehicle (10); and providing (104) at least one warning signal indicating an existing aquaplaning risk for the vehicle (10) when the at least one provided estimate (30) is within a significance range.The procedure (100) can further increase the safety of driving operations.

Description

[0001] The invention relates to a computer-implemented method for operating at least one vehicle.

[0002] Aquaplaning poses a significant safety risk in road traffic, as it can severely impair vehicle control. The ability to assess the current situation regarding the risk of aquaplaning and its potential consequences is crucial for ensuring safe driving and avoiding potentially life-threatening accidents.

[0003] German patent application DE 10 2019 222 313 A1 discloses a method for collecting data on road surface wetness and weather conditions for assessing aquaplaning situation. The data is processed by a trained machine learning unit, which can determine whether aquaplaning is imminent for the vehicle.

[0004] The object of the invention is to provide a computer-implemented method for operating at least one vehicle in which the safety of the driving operation is further increased.

[0005] The problem is solved by the features of the independent claims. Advantageous further developments are the subject of the dependent claims and the following description.

[0006] According to a first aspect, a computer-implemented method for operating at least one vehicle is described, comprising at least the following steps: providing at least one estimate indicating a risk of aquaplaning, using at least one trained learning algorithm based on at least one data set that includes at least ultrasonic sensor data from at least one ultrasonic sensor for determining a road surface type in front of the at least one vehicle and / or at least laser sensor data from at least one laser sensor for detecting at least one road groove in front of the at least one vehicle; and providing at least one warning signal indicating an existing aquaplaning risk to the vehicle when the at least one provided estimate is within a significance range.

[0007] The computer-implemented method for operating at least one vehicle uses at least one trained machine learning algorithm to determine at least one estimated value for the risk of aquaplaning for the at least one vehicle. This algorithm uses ultrasonic sensor data and / or laser sensor data as input. The ultrasonic sensor data can determine the type of road surface in front of the at least one vehicle. The laser sensor data can further determine road grooves in front of the at least one vehicle. By using the data on road surface type and / or the presence of road grooves, the estimated values ​​provided by the at least one trained machine learning algorithm can indicate the risk of aquaplaning for the at least one vehicle with increased accuracy.The trained, self-learning algorithm can, for example, comprise a neural network that can be provided with at least one data set at its input level and whose output level provides an estimated value for the aquaplaning risk. If the estimated value lies within a significance interval, which can, for example, cover a range between 80% and 100%, preferably between 90% and 100%, or more preferably between 99% and 100%, at least one warning signal can be provided to the driver. In this way, the driver can recognize the aquaplaning situation and adjust their driving accordingly. Alternatively or additionally, the at least one warning signal can also be provided to an assistance or automatic driving function that can adjust the vehicle's controls according to the estimated value. The safety of driving operations can thus be increased by using this computer-implemented method.

[0008] An assistive or automatic driving function can also be referred to as an autonomous driving function in the following.

[0009] According to some embodiments, it is conceivable that the at least one vehicle can have at least one water quantity sensor for detecting at least one quantity of water in at least one surface area of ​​the road surface in front of the at least one vehicle, wherein the at least one data set can further include at least water quantity sensor data from the at least one water quantity sensor.

[0010] By taking into account the amount of water in at least one surface area of ​​the road surface in front of at least one vehicle, the trained, self-learning algorithm can provide a more accurate estimate of the risk of aquaplaning. This can further increase road safety.

[0011] According to some embodiments, it is conceivable that the method, prior to the step of providing at least one estimate, may include at least the following steps: Training at least one learning algorithm using at least one training data set, which includes at least ultrasonic sensor data from at least one ultrasonic sensor of at least one training vehicle for determining a road surface type and / or at least laser sensor data from at least one laser sensor of the at least one training vehicle for detecting at least one road groove in front of the at least one training vehicle, such that the at least one learning algorithm outputs at least one estimate indicating a risk of aquaplaning for the at least one training vehicle.

[0012] To obtain a trained, self-learning algorithm, a self-learning algorithm can first be trained using a training data set.

[0013] The training dataset can be acquired, for example, using a training vehicle equipped with at least one ultrasonic sensor to detect the road surface type in front of the training vehicle and / or at least one laser sensor to detect at least one road groove in front of the training vehicle. Furthermore, the training dataset can contain information about the aquaplaning risk associated with the provided sensor data. In this way, a trained, self-learning algorithm can be obtained that can provide at least one highly accurate estimate of the risk of aquaplaning. Using this trained, self-learning algorithm can further enhance road safety.

[0014] According to some embodiments, it is conceivable that the at least one training data set may further include at least water quantity sensor data from at least one water quantity sensor of the at least one training vehicle for detecting at least one quantity of water in at least one surface area of ​​the road surface.

[0015] The trained, self-learning algorithm can then be further trained to consider at least a certain amount of water in at least one surface area on the road surface in front of the vehicle when making its estimate. This can further increase the accuracy of the estimate.

[0016] According to some embodiments, it is conceivable that the at least one training data set may further contain at least parameter data relating to tire properties of the at least one training vehicle, at least one vehicle type of the at least one training vehicle and / or at least one driving situation in which at least the ultrasonic sensor data of the at least one training vehicle were determined.

[0017] By considering parameter data relating to tire characteristics, vehicle type, and / or driving situation—for example, the speed of the training vehicle, cornering, braking behavior, the presence of vehicles ahead, etc.—the training of the machine learning algorithm can be further improved. A machine learning algorithm trained in this way can further enhance driving safety when used in a vehicle.

[0018] According to some embodiments, it is conceivable that the method, after the step of providing at least one warning signal, may further comprise at least the following steps: receiving the at least one warning signal by means of at least one autonomous driving function; and adjusting at least one setpoint for the vehicle's driving behavior by means of the autonomous driving function based on the displayed existing aquaplaning risk.

[0019] This allows an assistive or automatic driving function to adapt its driving behavior to the risk of a potential aquaplaning situation in front of the vehicle, based on the warning signal. This can increase the driving safety provided by the assistive or automatic driving function.

[0020] According to some embodiments, it is conceivable that the step of providing at least one estimated value and the step of providing at least one warning signal can be executed in real time.

[0021] The determination of the estimated value using the trained machine learning algorithm can be performed in real time. This means that the at least one ultrasonic sensor of the at least one vehicle can provide ultrasonic sensor data, or the at least one laser sensor of the at least one vehicle can provide laser sensor data, and the trained machine learning algorithm can display the at least one estimated value for the risk of aquaplaning before the corresponding surface area for which the ultrasonic sensor data or the laser sensor data were determined is crossed by the at least one vehicle.

[0022] According to a second aspect, a computer program product is described, comprising instructions that, when the program is executed by a computer, cause it to perform the steps of the procedure according to the preceding description.

[0023] The advantages, effects, and further developments of the computer program product result from the advantages, effects, and further developments of the method described above. Therefore, reference is made to the preceding description in this regard. A computer program product can be understood, for example, as a data carrier on which a computer program element is stored, containing instructions executable by a computer. Alternatively or additionally, a computer program product can also be understood, for example, as a permanent or volatile data storage medium, such as flash memory or main memory, that contains the computer program element. However, this does not exclude other types of data storage media that contain the computer program element.

[0024] According to a third aspect, a vehicle is described comprising at least one control unit for carrying out the steps of the procedure according to the preceding description, as well as at least one ultrasonic sensor for determining a road surface type in front of the vehicle and / or at least one laser sensor for detecting at least one road groove in front of the vehicle, wherein the control unit is configured to receive ultrasonic sensor data from the at least one ultrasonic sensor and / or laser sensor data from the at least one laser sensor and to output at least one warning signal indicating an existing aquaplaning risk for the vehicle.

[0025] According to some embodiments, it is conceivable that the vehicle may further have at least one water quantity sensor for detecting at least one quantity of water in at least one surface area of ​​the road surface in front of the vehicle.

[0026] The advantages, effects, and further developments of the vehicle result from the advantages, effects, and further developments of the procedure described above. To avoid repetition, reference is therefore made to the preceding description in this regard.

[0027] The invention is described below with reference to an exemplary embodiment and the accompanying drawing. The drawing shows: Fig. 1. A flowchart of the process; Fig. 2 a schematic representation of a driving situation with possible aquaplaning risk for the vehicle; Fig. 3 a schematic representation of a data set for the trained machine learning algorithm; and Fig. 4 A schematic representation of the training step of the learning algorithm.

[0028] The computer-implemented method for operating at least one vehicle is according to Fig. 1 in its entirety is referenced by the reference number 100.

[0029] The procedure 100 can be carried out in a vehicle 10 moving along a road, as in Fig. Figure 2 shows the direction of movement of vehicle 10, which is marked with an arrow.

[0030] The vehicle 10 can have at least one control unit 12 configured to execute the computer-implemented method 100. Furthermore, the vehicle 10 can have at least one ultrasonic sensor 14 and / or at least one laser sensor 16. Optionally, the vehicle 10 can also be equipped with a water quantity sensor 18. The sensors 14, 16, and 18 can be connected to the control unit 12 via signal connections.

[0031] The ultrasonic sensor 14 can be configured to determine the type of road surface of the roadway 20 in front of the vehicle 10 and to provide the road surface type as ultrasonic sensor data 36 via a first signal connection to the control unit 12.

[0032] The laser sensor 16 can be configured to detect at least one road groove 24 or track groove in front of the vehicle 10 and to provide the result of the detection as laser sensor data 38 via a second signal connection to the control unit 12.

[0033] The optional water quantity sensor 18 can be configured to determine the quantity of water in at least one surface area 22 of the road surface 20 in front of the vehicle 10. Furthermore, the water quantity sensor 18 can be configured to provide the determined quantity of water as water quantity sensor data 40 to the control unit 12 via a third signal connection.

[0034] According to the computer-implemented method 100, at least one estimated value 30 indicating a risk of aquaplaning for the vehicle 10 can be provided in a step 102 using a trained, self-learning algorithm 28. The trained, self-learning algorithm 28 can, for example, be implemented as a neural network.

[0035] An input for the trained, self-learning algorithm 28 can be according to Fig. 3. There must be at least one data set 34 that includes at least the ultrasonic sensor data 36 and / or the laser sensor data 38. The output of the trained machine learning algorithm 28 can then be at least one estimate 30 for the vehicle's aquaplaning risk.

[0036] If at least one estimate 30 lies outside a significance range, the aquaplaning risk can be classified as low or non-existent. Step 102 can then be repeated with a new dataset 34 containing new ultrasonic sensor data 36 and / or new laser sensor data 38.

[0037] If at least one estimated value 30 is within a significance range, at least one warning signal can be provided according to step 104. The warning signal can indicate the existing aquaplaning risk for the vehicle 10. After providing the warning signal, step 102 can be performed again with a new data set 34.

[0038] The at least one estimated value 30 can, for example, be expressed as a probability for the aquaplaning risk. Accordingly, the significance range can, for example, be between 80% and 100%, preferably between 90% and 100%, and more preferably between 99% and 100%.

[0039] The at least one data set 34 can also contain water quantity sensor data 40. The trained machine learning algorithm 28 can then use the water level sensor data 40, in addition to the ultrasonic sensor data 36 and the laser sensor data 38, to determine the at least one estimated value 30.

[0040] Steps 102 and 104 can be performed in real time, so that at least one estimate 30 is output by the trained learning algorithm 28 before the area of ​​the roadway for which the ultrasonic sensor data 36 and / or laser sensor data 38 used for the output estimate 30 were determined is driven on by the vehicle 10.

[0041] To provide the trained machine learning algorithm 28, step 106 can optionally be performed before step 102, which is explained above. In the optional step 106, a machine learning algorithm 28 can be trained with a training dataset 26, as exemplified in Fig.Figure 4 is shown. The training data set 26 can contain at least ultrasonic sensor data 36 and / or laser sensor data 38, which were acquired by a training vehicle (not shown) in a training situation. Furthermore, the at least one training data set 26 can also contain water volume sensor data 40, which were likewise acquired by the training vehicle in the training situation. The training situation can have specific parameters for, for example, the type of road surface, the speed of the training vehicle, cornering by the training vehicle, braking behavior of the training vehicle, the amount of water on the road surface 20 and / or road grooves 24. Parameters can also be provided regarding whether and / or when aquaplaning occurred with the training vehicle. These parameters can be summarized as parameters of at least one driving situation 46.

[0042] Furthermore, the training data set 26 can also contain parameters about the tire properties 42 of the at least one training vehicle and / or the vehicle type 44 of the at least one training vehicle. With these parameters, the machine learning algorithm 28 can be specifically trained on a vehicle type 44 or on specific tire properties 42.

[0043] For training, at least one training data set 26 can be used, preferably a plurality of training data sets 26. The at least one training data set 26 is provided as input to the at least one machine learning algorithm 28. The estimated value 30 output by the machine learning algorithm 28 is then checked by means of a comparator 32 to see how well the estimated value 30 indicates the aquaplaning risk for the training vehicle in the driving situation assigned to the training data set 26.

[0044] If the agreement of the output estimate 30 with the aquaplaning risk is low or outside a tolerance range, the learning algorithm 28 can be modified or adapted to provide an estimate 30 in a next run with the at least one training data set 26 that is closer to the aquaplaning risk assigned to the corresponding at least one training data set 26.

[0045] The training can be repeated with all training data sets 26 until the estimates 30 output by the learning algorithm 28 are essentially in agreement with the aquaplaning risks derived from the respective driving situation.

[0046] The result of the training can then be referred to as a trained learning algorithm 28.

[0047] The training vehicle and the vehicle 10 in which the trained machine learning algorithm 28 is used do not have to be the same vehicle. The trained machine learning algorithm 28 can also be implemented on at least one control unit 12 of the vehicle 10, for example by copying it from the training vehicle or a separate data carrier.

[0048] In a further optional step 108 of the procedure 100, at least one autonomous driving function can receive at least one warning signal after step 104. The autonomous driving function can then take the warning signal or the aquaplaning risk indicated by it into account for the control of the vehicle 10.

[0049] In a further optional step 110, at least one target value for the vehicle's driving behavior 10 can be adjusted. For example, a target speed value can be reduced so that the vehicle 10 decelerates gradually to reduce the risk of aquaplaning.

[0050] The example described above does not in any way limit the invention. Rather, the invention can be modified in numerous ways. All features of the invention described above can be essential to the invention, either alone or in combination. Reference symbol list 10 vehicles 12 Control unit 14 Ultrasonic sensor 16 laser sensor 18 Water volume sensor 20 Road surface 22 Surface area 24 Road groove 26 Training data set 28 machine learning algorithms 30 Estimated value 32 comparators 34 data set 36 Ultrasonic sensor data 38 laser sensor data 40 water volume sensor data 42 parameter data on tire properties 44 parameter data for vehicle type 46 parameter data for driving situation QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2019 222 313 A1

[0003]

Claims

A computer-implemented method (100) for operating at least one vehicle (10), comprising at least the following steps: a. Providing (102) at least one estimate (30) indicating a risk of aquaplaning, by means of at least one trained machine learning algorithm (28) based on at least one data set (34) comprising at least ultrasonic sensor data (36) from at least one ultrasonic sensor (14) for determining a road surface type in front of the at least one vehicle (10) and / or at least laser sensor data (38) from at least one laser sensor (16) for detecting at least one road groove (24) in front of the at least one vehicle (10); and b. Providing (104) at least one warning signal indicating an existing aquaplaning risk to the vehicle (10) when the at least one provided estimate (30) is within a significance range. Computer-implemented method (100) according to claim 1, characterized in that the at least one vehicle (10) has at least one water quantity sensor (18) for detecting at least one quantity of water in at least one surface area (22) of the road surface (20) in front of the at least one vehicle (10), wherein the at least one data set (34) further comprises at least water quantity sensor data (40) of the at least one water quantity sensor (18). A computer-implemented method (100) according to one of the preceding claims, characterized in that the method (100) comprises, prior to the step of providing (102) at least one estimate (30), at least the following steps: a. Training (106) at least one machine learning algorithm (28) using at least one training data set (26) comprising at least ultrasonic sensor data (36) from at least one ultrasonic sensor (14) from at least one training vehicle for determining a road surface type (20) and / or at least laser sensor data (38) from at least one laser sensor (16) of the at least one training vehicle for detecting at least one road groove (24) in front of the at least one training vehicle, such that the at least one machine learning algorithm (28) outputs at least one estimate (30) indicating a risk of aquaplaning for the at least one training vehicle. Computer-implemented method (100) according to claim 3, characterized in that the at least one training data set (26) further comprises at least water quantity sensor data (40) of at least one water quantity sensor (18) of the at least one training vehicle for detecting at least one quantity of water in at least one surface area (22) of the road surface (20). Computer-implemented method (100) according to one of claims 3 or 4, characterized in that the at least one training data set (26) further comprises at least parameter data relating to tire properties (42) of the at least one training vehicle, at least one vehicle type (44) of the at least one training vehicle and / or at least one driving situation (46) in which at least the ultrasonic sensor data (36) of the at least one training vehicle were determined. Computer-implemented method (100) according to one of the preceding claims, characterized in that the method (100) after the step: providing (104) at least one warning signal, further comprises at least the following steps: a. receiving (108) the at least one warning signal by means of at least one autonomous driving function; and b. adjusting (110) at least one setpoint for the driving behavior of the vehicle (10) by means of the autonomous driving function based on the displayed existing aquaplaning risk. Computer-implemented method (100) according to one of the preceding claims, characterized in that the step: providing (102) at least one estimate (30), and the step: providing (104) at least one warning signal, are performed in real time. Computer program product comprising instructions which, when the program is executed by a computer, cause it to perform the steps of the method (100) according to any one of claims 1 to 7. Vehicle (10) comprising at least one control unit (12) for performing the steps of the method according to one of claims 1 to 7 and at least one ultrasonic sensor (14) for determining a road surface type in front of the vehicle (10) and / or at least one laser sensor (16) for detecting at least one road groove (24) in front of the vehicle (10), wherein the (12) is configured to receive ultrasonic sensor data (36) from the at least one ultrasonic sensor (14) and / or laser sensor data (38) from the at least one laser sensor (16) and to output at least one warning signal indicating an existing aquaplaning risk for the vehicle (10). Vehicle (10) according to claim 9, characterized in that the vehicle (10) further comprises at least one water quantity sensor (18) for detecting at least one quantity of water in at least one surface area (22) of the road surface (20) in front of the vehicle (10).

Images