Method and device for determining the fill level of a vehicle's windshield washer fluid reservoir

Abstract

Device (101, 120) for determining an estimated value for the fill level of a windshield washer fluid reservoir (105) of a windshield washer system of a vehicle (100); wherein the device (101, 120) is configured, - To determine operating data (121) relating to the operation of the wiper system; wherein the operating data (121) display measured values ​​for the following measured variables - a service life of a pump (103) of the wiper system, which is designed to pump wiper water (107) from the wiper water reservoir (105); and - an electrical power consumed by the pump (103) during a pumping operation; and - a time course of an on-board voltage of an electrical on-board network of the vehicle (100) during a pumping process; and - to determine the estimated value for the fill level of the windshield washer fluid reservoir (105) based on the operating data (121) using a machine-learned estimation model (122).

Description

[0001] The invention relates to a method and a corresponding device for determining, in particular estimating, the fill level of a windshield washer fluid reservoir of a vehicle.

[0002] A (motor) vehicle typically has a wiper system for cleaning one or more windows, in particular a windshield and / or a rear window, and / or one or more headlights and / or one or more vehicle sensors. The wiper system includes a washer fluid reservoir containing washer fluid. Furthermore, the wiper system typically includes a level sensor designed to detect when the washer fluid level in the reservoir falls below a certain threshold.

[0003] The use of a level sensor typically only provides a vehicle user with information on whether the fluid level is higher or lower than a specified threshold. The vehicle user usually receives no information about the actual fluid level in the windshield washer reservoir. US 10,627,245 B2 describes a method for predicting a vehicle's fluid consumption on an upcoming route. DE 10 2008 021 382 A1 describes a method for determining the fluid level of a fluid reservoir. EP 3 650 291 A1 describes a method for predicting windshield washer fluid consumption. DE 10 2015 210 312 A1 describes a method for determining a vehicle's remaining windshield washer fluid range.

[0004] This document addresses the technical task of determining, and in particular estimating, the actual fill level of a windshield washer fluid reservoir in a vehicle's windshield washer system in an efficient and precise manner.

[0005] The problem is solved in each case by the independent claims. Advantageous embodiments are described, inter alia, in the dependent claims. It should be noted that additional features of a claim dependent on an independent claim, without the features of the independent claim itself or only in combination with a subset of the features of the independent claim, can constitute a separate invention independent of the combination of all features of the independent claim, which can be made the subject of an independent claim, a divisional application, or a subsequent application. This applies equally to technical teachings described in the description, which can constitute an invention independent of the features of the independent claims.

[0006] According to one aspect, a device for determining an estimated value for the fill level of a windshield washer fluid reservoir of a (motor) vehicle's windshield washer system is described. The device is configured to determine operating data relating to the operation of the windshield washer system.

[0007] The operating data may display or include measured values ​​for one or more of the following parameters: the time at which the windshield washer fluid reservoir was (last) filled; the operating time of the windshield washer pump designed to pump windshield washer fluid from the reservoir (in particular, the operating time for each individual pumping operation (since the time of filling) and / or the cumulative operating time for all pumping operations (since the time of filling)); the times of one or more temporally isolated pumping operations of the pump (since the time of filling); the electrical power consumed by the pump (in particular as a function of time) during a pumping operation; the time course of the vehicle's electrical system voltage during a pumping operation; and / or the ambient temperature of the vehicle and / or the windshield washer fluid reservoir, in particular the time course of the ambient temperature, since the time of filling.

[0008] The operating data can therefore show how the wiper system has been operated (since the last time it was filled) (especially with regard to the removal of wiper water from the wiper water reservoir).

[0009] The device is further configured to determine the estimated fill level of the windshield washer fluid reservoir based on operating data using a machine-learned estimation model. The estimation model can be configured to take the operating data as input values ​​(to a neural network) and provide the estimated fill level as an output value (to the neural network).

[0010] By using a machine-learned estimation model, the estimated value of a vehicle's windshield washer fluid reservoir can be determined efficiently and precisely. In particular, the washer fluid consumption curve of a windshield washer system can be learned (and described by an estimation model) with precision. The learned consumption curve can then be used to more accurately estimate the washer fluid level.

[0011] The estimation model can comprise one or more trained neural networks. Alternatively or additionally, the estimation model can be trained using a variety of training datasets. A training dataset can contain operational training data from one or more windshield wiper systems of one or more vehicles. Furthermore, a training dataset can include fill level information related to the actual fill level of the windshield washer fluid reservoir of the vehicle's windshield washer system (associated with the operational training data), corresponding to the operational training data.

[0012] The fill level information can include or display the threshold time at which the windshield washer fluid reservoir reaches a specific level (e.g., at which the windshield washer fluid reservoir sensor is triggered). Alternatively or additionally, the fill level information can include or display a measured value of the windshield washer fluid reservoir level.

[0013] By collecting a large number of training data sets during the operation of one or more car washes, a precise and robust estimation model can be trained.

[0014] The device can be configured to take action regarding the windshield washer fluid reservoir based on the determined estimated fill level. This action could, for example, include issuing a notification (regarding the estimated fill level) to a vehicle user.

[0015] The device can be configured to determine, based on the estimated level of the windshield washer fluid reservoir, that a notification to refill the windshield washer fluid reservoir should be issued (e.g., because the level has reached or fallen below a certain level, which may differ from the level threshold of the level sensor).

[0016] Furthermore, the device can be configured to determine that the vehicle is, or will be, in an operating situation suitable for filling the windshield washer fluid reservoir. Examples of such situations include: refueling or charging to take on an energy source for powering the vehicle and / or parking the vehicle.

[0017] In response to this (i.e., in response to the detected operating situation and the detected need to issue a warning), the system can then issue a warning to refill the windshield washer fluid reservoir, particularly selectively, in preparation for and / or during the vehicle's operating situation. This allows the user to be prompted to refill the windshield washer fluid reservoir in a particularly convenient and reliable manner.

[0018] According to another aspect, a device for training and / or adapting a machine-learned estimation model is described, wherein the estimation model is configured to determine an estimated value of the fill level of a windshield washer fluid reservoir of a vehicle based on operating data from a vehicle's windshield wiper system. The device can, for example, be an external server.

[0019] The device is configured to acquire training data using a variety of training datasets. A training dataset can include actual operating data from a windshield wiper system and associated actual fill level information regarding the actual fill level of the wiper system's reservoir. Training datasets can be provided from a variety of different vehicles. The respective vehicle type and / or wiper system type can be specified along with each training dataset to enable the development of an estimation model that is dependent on the vehicle type and / or wiper system type.

[0020] Furthermore, the device is designed to train and / or adapt parameters of the estimation model based on the training data. These parameters can include neuron parameters, in particular weights and / or offsets, of neurons in a neural network of the estimation model. This allows for the efficient provision of an estimation model for the precise determination of the fill level of a windshield washer fluid reservoir.

[0021] As explained above, the operating data can include measured values ​​relating to a set of different measured variables. The device can be configured to adjust the set of measured variables considered by the estimation model. In particular, the device can be configured, based on the trained estimation model, to determine whether a first measured variable from the set of measured variables has an influence on the estimated value determined by the estimation model that is less than an influence threshold. In other words, it can be recognized that a first measured variable has no significant influence on the determined estimated value.

[0022] In response, the first measurement can be removed from the set of measurements considered by the estimation model. This further increases the efficiency of determining estimated windshield washer fluid level.

[0023] According to another aspect, a (road) motor vehicle (in particular a passenger car or a truck or a bus or a motorcycle) is described which includes the device described in this document for determining an estimated value of the fill level of a windshield washer fluid reservoir.

[0024] According to one aspect, a method for training and / or adapting a machine-learned estimation model is described. This model is trained to determine an estimated level of the windshield washer fluid reservoir based on operational data from a vehicle's windshield wiper system. The method includes generating training data from a variety of training datasets. A training dataset can contain actual operational data from a windshield wiper system and associated actual fluid level information related to the reservoir's fill level. The method further includes training and / or adapting parameters of the estimation model based on the training data.

[0025] According to another aspect, a method for determining an estimated fill level of a windshield washer fluid reservoir in a vehicle's windshield washer system is described. The method includes determining operational data related to the operation of the windshield washer system. Furthermore, the method includes determining the estimated fill level of the windshield washer fluid reservoir based on this operational data using a machine-learned estimation model.

[0026] According to another aspect, a software (SW) program is described. The SW program can be set up to run on a processor (e.g., on a vehicle's control unit or on an external vehicle unit) and thereby execute at least one of the procedures described in this document.

[0027] According to another aspect, a storage medium is described. The storage medium can include a software program that is configured to run on a processor and thereby execute at least one of the procedures described in this document.

[0028] It should be noted that the methods, devices, and systems described in this document can be used both alone and in combination with other methods, devices, and systems described in this document. Furthermore, any aspect of the methods, devices, and systems described in this document can be combined with one another in a variety of ways. In particular, the features of the claims can be combined with one another in a variety of ways.

[0029] The invention will now be described in more detail using exemplary embodiments. Fig. 1a Exemplary components of a vehicle; Fig. 1b an exemplary device for determining an estimation model for the fill level of a windshield washer fluid reservoir; Fig. 2a an example neural network; Fig. 2b an example neuron; Fig. 3a a flowchart of an exemplary procedure for training an estimation model; Fig. 3b a flowchart of an exemplary procedure for determining an estimated value of the fill level of a windshield washer fluid reservoir.

[0030] As stated at the outset, this document deals with the efficient and precise estimation of the actual fill level of a windshield washer fluid reservoir in a vehicle's windshield washer system. In this context, it shows Fig. 1a Exemplary components of a vehicle 100 with a wiping system for a disc 109, e.g. for a windshield, of the vehicle 100. The disc 109 can be wiped with one or more wipers 108, which are driven by one or more actuators (e.g. motors).

[0031] A washer fluid 107 can be applied to the windshield 109 via a nozzle 102 from a washer fluid reservoir 105 of the washer system. The washer fluid 107 can be pumped from the washer fluid reservoir 105 to the nozzle 102 by means of a pump 103. A level sensor 106 can be arranged on the washer fluid reservoir 106, which is designed to detect whether the fill level of the washer fluid reservoir 105 is greater or less than a certain fill level threshold.

[0032] The vehicle 100 can have a control device 101 which is configured to control the various components 103, 104 of the wiper system. In particular, in response to a request to spray washer fluid onto the windshield 109 to be cleaned, the pump 103 can be caused to spray washer fluid 107 from the washer fluid reservoir 105 onto the windshield 109 via the nozzle 102.

[0033] The device 101 can further be configured to determine, based on the sensor data from the level sensor 106, whether the fill level of the windshield washer fluid reservoir 105 is higher or lower than the fill level threshold. The actual fill level cannot typically be determined solely based on the sensor data from the level sensor 106.

[0034] Fig. Figure 1b shows an exemplary system 150 with a plurality of vehicles 100, which are trained to send operating data 121 relating to the wiper system of the respective vehicle 100 to a vehicle-external unit 120, e.g., a backend server. For this purpose, each vehicle 100 can include a communication unit 110, which enables the transmission of data via a (possibly wireless) communication link, such as 3G, 4G, or 5G.

[0035] The operating data 121 for the wiper system of a vehicle 100 can display, • the filling time at which the washer fluid reservoir 105 was (completely) filled with washer fluid 107; • the threshold time (following the filling time) at which the level sensor 106 (for the first time) indicated that the level threshold had been reached; • the duration of one or more pumping operations (between the filling time and the threshold time) during which the pump 103 was operated to pump washer fluid 107 from the washer fluid reservoir 105; • the time course of the vehicle's electrical system voltage 100 during the operation of the pump 103; • the time course of the electrical power consumed by pump 103 during a pumping operation; and / or • one or more operating conditions of the wiper system between the filling time and the threshold time, such as the ambient temperature of the vehicle 100 or of the wiper fluid reservoir 105, the temperature of the vehicle's drive motor 100, etc.

[0036] The operating data 121 for a vehicle 100 can be provided together with type information relating to the type of vehicle 100, in particular to the type of wiper system.

[0037] The vehicle-external unit 120 can be configured to determine a machine-learned estimation model 122 based on the operating data 121 of one or more vehicles 100. This model is configured to determine an estimated value of the fill level of the windshield washer fluid reservoir 105 of the windshield washer system based on the operating data 121 of the windshield washer system. The determined estimation model 122 can be made available to the one or more vehicles 100 (e.g., via a communication link, possibly wireless) to enable them to estimate the actual fill level of the windshield washer fluid reservoir 105 of the respective vehicle 100 in a precise and efficient manner.

[0038] The estimation model 122 can include one or more trained neural networks. Fig. 2a and Fig. Figure 2b shows exemplary components of a neural network 200, in particular a feedforward network. In the example shown, the network 200 comprises two input neurons or input nodes 202, each of which receives a current value of an input variable as an input value 201 at a specific time t. The one or more input nodes 202 are part of an input layer 211. In general, the network 200 can be configured to receive an input data set with one or more input values ​​201. For example, operating data 121 from a vehicle's windshield wiper system 100 can be passed to the neural network 200 as input values ​​201.

[0039] The neural network 200 further comprises neurons 220 in one or more hidden layers 212 of the neural network 200. Each of the neurons 220 can have as input values ​​the individual output values ​​of the neurons of the preceding layer 212, 211 (or at least a part thereof). Processing takes place in each of the neurons 220 to determine an output value of the neuron 220 depending on the input values. The output values ​​of the neurons 220 of the last hidden layer 212 can be processed in an output neuron or output node 220 of an output layer 213 to determine the one or more output values ​​203 of the neural network 200. In general, the network 200 can be configured to provide output data with one or more output values ​​203. For example, the estimated fill level of the windshield washer fluid reservoir 105 can be provided as the initial value 203 for the windshield washer system.

[0040] Fig. Figure 2b illustrates exemplary signal processing within a neuron 220, in particular within the neurons 202 of one or more hidden layers 212 and / or the output layer 213. The input values ​​221 of the neuron 220 are weighted with individual weights 222 to determine a weighted sum 224 of the input values ​​221 in a sum unit 223 (possibly taking into account a bias or offset 227). An activation function 225 can map the weighted sum 224 to an output value 226 of the neuron 220. The activation function 225 can, for example, limit the range of values. For a neuron 220, for example, a sigmoid function, a hyperbolic tangent (tanh) function, or a rectified linear unit (ReLU), e.g., f(x) = max(0, x), can be used as an activation function 225. If necessary, the value of the weighted sum 224 can be shifted by an offset 227.

[0041] A neuron 220 thus has weights 222 and / or, if applicable, an offset 227 as neuron parameters. The neuron parameters of the neurons 220 of a neural network 200 can be trained in a training phase to cause the neural network 200 (i.e., the estimation model comprising the neural network 200) to approximate a specific function and / or model a specific behavior.

[0042] Training a neural network 200 can be done, for example, using the backpropagation algorithm. For this purpose, in a first phase of a q ten In the epoch of a learning algorithm, corresponding output values ​​203 are determined at the output of the one or more output neurons 220 for the input values ​​201 at the one or more input nodes 202 of the neural network 200. Based on the output values ​​203, the error value of an optimization or error function can be determined.

[0043] In a second phase of the qten In each epoch of the learning algorithm, the error or error value is propagated back from the output to the input of the neural network to modify the neuron parameters of neurons 220 layer by layer. The resulting error function at the output can be partially derived for each individual neuron parameter of neural network 200 to determine the magnitude and / or direction for adjusting the individual neuron parameters. This learning algorithm can be repeated iteratively for a large number of epochs until a predefined convergence and / or termination criterion is reached.

[0044] In the present context, training data 121 can be determined from a variety of training datasets based on the operational data 100 provided by one or more vehicles 100. This training data can be used to train an estimation model 122 with one or more neural networks 200, enabling the estimation model 122 to determine an estimated value for the fill level of the windshield washer fluid reservoir 105 of vehicle 100 based on the operational data 121 of a vehicle 100.

[0045] Thus, a system 150 and / or a device 101 is described that enables an efficient estimation of the windshield washer fluid level. Operating data 121 of a vehicle fleet are collected, e.g., in a database 120.

[0046] The operating data 121 can include (temporally) continuous and / or (temporally) discrete values, such as e.g. • the time course of the on-board voltage while pump 103 was running; • the duration of the pumping process, possibly including an absolute timestamp (this allows, for example, the influence of pauses between pumping processes to be taken into account); • the ambient temperature; • the engine temperature; • the filling time of the windshield washer fluid tank 105; and / or • the time of falling below the residual fill level (based on the detection of the fill level sensor 106).

[0047] The operational data 121 collected in the individual vehicles 100 are sent to a backend system 120. This transmission can occur, for example, at periodic time intervals and / or based on events. An example of an event is when the residual fill level falls below the threshold. In other words, the operational data 121 for a vehicle 100 can be provided at the threshold point at which the fill level sensor 106 detects (for the first time after a filling event) that the fill level has reached or fallen below the threshold value.

[0048] The collected operating data 121 can be extended by one or more additional input signals, such as static vehicle characteristics, for example the vehicle type, the production date, the type of the installed pump 103, etc. Based on this information, it is possible to adapt the estimation model 122 to different vehicle variants and / or vehicle types and / or wiper system types.

[0049] Using the collected operational data 121 of a vehicle fleet, an estimation model 122 can be trained by means of a machine learning algorithm. This model maps the relationship between the operational data 121 of a windshield wiper system and the fill level of the tank 105 or the consumption of the windshield washer fluid 107. An estimation model 122 can be implemented, for example, by one or more artificial neural networks 200.

[0050] The estimation model 122 can be retrained and / or adjusted as soon as new operational data 121 is provided from the vehicle fleet. This allows for repeated training of the estimation model 122, thereby further improving its accuracy.

[0051] The currently trained model 122 can be distributed from the backend system 120 to the vehicles 100. The operational data 121 used to train the estimation model 122 can be collected in a vehicle 100. Locally in the vehicle 100, an estimated fill level can be determined based on the model 122 and the operational data 121. Alternatively, the estimated fill level can be determined in the backend system 120 or on a mobile device (such as a smartphone) if the operational data 121 currently collected in the vehicle 100 is available there.

[0052] The fill level estimation can be used to alert the driver of a vehicle (100) that the windshield washer fluid reservoir (105) needs refilling even before the remaining fill level (i.e., the fill level threshold) is reached. This notification can be issued at a convenient time for the driver, e.g., during refueling at a gas station.

[0053] As part of the training of the estimation model 122, at least some measured values ​​relating to the actual fill level of windshield washer fluid reservoirs 105 can be used. For example, a (relatively small) proportion of vehicles 100 (e.g., 1%) may be equipped with additional sensors that provide more accurate measurements of the current fill level of the windshield washer fluid reservoir 105. Using these measurements, it is possible to further increase the accuracy of the training data and thus the accuracy of the trained estimation model 122.

[0054] For example, the processing unit 120 can check whether a measured value provided in the operating data 121 is relevant for determining the estimated fill level. The number of measured values ​​provided and / or considered in the operating data 121 can then be adjusted, e.g., reduced or increased.

[0055] In other words, the signals, i.e., measured values, to be acquired by the vehicles 100 can be modified if necessary. It may become apparent during the training of the estimation model 122 that a particular signal, i.e., a particular measured value, has no or only a relatively minor effect on determining the estimated value. The backend system 120 can then issue a control command to the vehicle fleet to no longer acquire this signal, i.e., this measured value. Similarly, a signal, i.e., a measured value, can be subsequently included in the data acquisition.

[0056] Fig. Figure 3a shows a flowchart of a (possibly computer-implemented) procedure 300 for training and / or adapting a machine-learned estimation model 121, which is trained to determine an estimated value of the fill level of a windshield washer fluid reservoir 105 of the windshield washer system based on operating data 121 of a vehicle's windshield washer system 100. The procedure 300 can be executed by a vehicle-external unit 120, in particular by a backend server.

[0057] The procedure 300 comprises the determination 310 of training data using a large number of training datasets. The training datasets may have been provided by one or more vehicles 100. A training dataset may include actual operating data 121 of a windshield washer system and associated actual fill level information relating to the actual fill level of the washer fluid reservoir 105 of the windshield washer system. The actual operating data 121 for a training dataset may extend from the time the respective washer fluid reservoir 105 is filled to the threshold time at which the fill level sensor 106 is triggered. The actual fill level information may indicate the threshold time and / or the time interval between the filling time and the threshold time.

[0058] The procedure 300 further comprises the training and / or adaptation 302 of parameters of the estimation model 121, in particular neuron parameters 223, 227 of a neural network 200, based on the training data. In this way, an estimation model 121 can be efficiently provided that enables a precise estimation of the fill level of a windshield washer fluid reservoir 105.

[0059] Fig. Figure 3b shows a flowchart of an exemplary (possibly computer-implemented) procedure 310 for determining an estimated value for the fill level of a windshield washer fluid reservoir 105 of a windshield washer system of a vehicle 100. The procedure 310 can be executed by a control device 101 of the vehicle 100, by a vehicle-external unit 120 (e.g., by a backend server), and / or by an electronic user device (such as a smartphone) of a user of the vehicle 100.

[0060] The procedure 310 comprises determining 311 operational data 121 relating to the operation of the wiper system of the vehicle 100. In other words, operational data 121 can be recorded during the operation of the wiper system, in particular starting from the last time the wiper fluid reservoir 105 was filled.

[0061] Furthermore, procedure 300 includes determining 312 the estimated value for the fill level of the windshield washer fluid reservoir 105 based on the operating data 121 using a machine-learned estimation model 122. The estimation model 122 may have been trained using procedure 300 described in this document.

[0062] The measures described in this document make it possible to determine the fill level of a windshield washer fluid reservoir 105 in an efficient and precise manner (even without the use of a fill level sensor 106).

[0063] The present invention is not limited to the embodiments shown. In particular, it should be noted that the description and the figures are intended only to illustrate the principle of the proposed methods, devices, and systems.

Claims

[1] Device (101, 120) for determining an estimated value for the fill level of a windshield washer fluid reservoir (105) of a windshield washer system of a vehicle (100); wherein the device (101, 120) is configured, - To determine operating data (121) relating to the operation of the wiper system; wherein the operating data (121) display measured values ​​for the following measured variables - a service life of a pump (103) of the wiper system, which is designed to pump wiper water (107) from the wiper water reservoir (105); and - an electrical power consumed by the pump (103) during a pumping operation; and - a time course of an on-board voltage of an electrical on-board network of the vehicle (100) during a pumping process; and - to determine the estimated value for the fill level of the windshield washer fluid reservoir (105) based on the operating data (121) using a machine-learned estimation model (122). [2] Device (101, 120) according to claim 1, wherein the operating data (121) display measured values ​​for one or more of the following measured quantities, - a filling time at which the windshield washer fluid reservoir (105) was filled; - Time points of one or more temporally isolated pumping operations of the pump (103); and / or - an ambient temperature of the vehicle (100) and / or the windshield washer reservoir (105), in particular a time course of the ambient temperature since the time of filling. [3] Device (101, 120) according to one of the preceding claims, wherein the estimation model (122) is configured to receive the operating data (121) as input values ​​(201) and to provide the estimated level as output value (203). [4] Device (101, 120) according to one of the preceding claims, wherein - the estimation model (122) comprises one or more trained neural networks (200); and / or - the estimation model (122) was trained on the basis of training data using a large number of training datasets; and / or - a training data set comprising training operational data (121) from one or more wiper systems of one or more vehicles (100); and / or - a training data set comprising level information relating to the actual level of a windshield washer reservoir (105) of a windshield washer system of a vehicle (100), wherein the level information in particular includes a threshold time at which the level of the windshield washer reservoir (105) reaches a level threshold value, and / or includes a measured value of the level of the windshield washer reservoir (105). [5] Device (101, 120) according to one of the preceding claims, wherein - the device (101, 120) is set up to take action with respect to the windshield washer fluid reservoir (105) depending on the determined estimated value for the fill level of the windshield washer fluid reservoir (105); and - the measure includes in particular issuing a notice to a user of the vehicle (100). [6] Device (101, 120) according to claim 5, wherein the device (101, 120) is configured, - to determine, based on the estimated value for the fill level of the windshield washer reservoir (105), that a notification to refill the windshield washer reservoir (105) should be issued; - to determine that the vehicle (100) is or will be in an operating situation suitable for filling the windshield washer fluid reservoir (105); wherein the operating situation includes, in particular, a process for taking on an energy carrier for propulsion of the vehicle (100) and / or a parking operation; and - in response to this, to cause the instruction to fill the windshield washer reservoir (105), in particular selectively, to be issued in preparation for and / or during the operating situation of the vehicle (100). [7] Device (150) for training and / or adapting a machine-learned estimation model (122) configured to determine, based on operating data (121) of a wiper system of a vehicle (100), an estimated value of the fill level of a wiper fluid reservoir (105) of the wiper system; wherein the operating data (121) display measured values ​​for the following measured variables - a service life of a pump (103) of the wiper system, which is designed to pump wiper water (107) from the wiper water reservoir (105); and - an electrical power consumed by the pump (103) during a pumping operation; and - a time course of an on-board voltage of an electrical on-board network of the vehicle (100) during a pumping process; wherein the device (150) is configured, - to determine training data using a variety of training datasets; wherein a training dataset comprises actual operating data (121) of a wiper system and associated actual fill level information relating to the actual fill level of the wiper system's washer fluid reservoir (105); and - To train and / or adapt parameters of the estimation model (122) based on the training data. [8] Device (150) according to claim 7, wherein the parameters comprise neuron parameters (222, 227), in particular weights (222) and / or offsets (227), of neurons (220) of a neural network (200) of the estimation model (122). [9] Device (150) according to one of claims 7 to 8, wherein - the operational data (121) include measured values ​​relating to a set of different measured quantities; and - the device (150) is set up to adjust the quantity of measured variables taken into account by the estimation model (122). [10] Device (150) according to claim 9, wherein the device (150) is configured, - to determine, based on the trained estimation model (122), that a first measurement from the set of measurements has an influence on the estimate determined by the estimation model (122) that is smaller than an influence threshold; and - in response to this, to remove the first measurement from the set of measurements considered by the estimation model (122). [11] Method (310) for determining an estimated value for the fill level of a windshield washer fluid reservoir (105) of a windshield washer system of a vehicle (100); wherein the method (310) comprises, - Determining (311) operational data (121) relating to the operation of the wiper system; wherein the operational data (121) indicate measured values ​​for the following measured variables - a service life of a pump (103) of the wiper system, which is designed to pump wiper water (107) from the wiper water reservoir (105); and - an electrical power consumed by the pump (103) during a pumping operation; and - a time course of an on-board voltage of an electrical on-board network of the vehicle (100) during a pumping process; and - Determining (312) the estimated value for the fill level of the windshield washer fluid reservoir (105) based on the operating data (121) using a machine-learned estimation model (122). [12] Method (300) for training and / or adapting a machine-learned estimation model (122) which is trained to determine an estimated value of the fill level of a windshield washer fluid reservoir (105) of the windshield washer system based on operational data (121) of a windshield washer system of a vehicle (100); wherein the method (300) comprises, - Determining (301) training data using a plurality of training datasets; wherein a training dataset comprises actual operating data (121) of a wiper system and associated actual fill level information relating to the actual fill level of the wiper system's washer fluid reservoir (105); wherein the operating data (121) display measured values ​​for the following measured variables - a service life of a pump (103) of the wiper system, which is designed to pump wiper water (107) from the wiper water reservoir (105); and - an electrical power consumed by the pump (103) during a pumping operation; and - a time course of an on-board voltage of an electrical on-board network of the vehicle (100) during a pumping process; and - Training and / or adapting (302) parameters of the estimation model (122) based on the training data.

Images