Method for operating a vehicle

Abstract

The invention relates to a method for operating a vehicle in an activated recuperation mode, wherein the vehicle has at least one drive unit designed as an electric machine and a control and / or regulating device (300) for executing a braking torque request by means of recuperation by the at least one electric machine, at least comprising detecting a current driving situation from current sensor values ​​(20); determining a current braking torque request (30) by the control and / or regulating device (300) to the at least one electric machine; determining a current driver request (14) from a current accelerator pedal position and from a current brake pedal position of the vehicle; determining a difference (22) between the current braking torque request (30) by the control and / or regulating device (300) and the current driver request (14);and determining a correction value with the difference (22), with which correction value the value of the braking torque request (30) stored for the current driving situation in the control and / or regulating device (300) is adjusted.;

Description

The invention relates to a method for operating a vehicle in an activated recuperation mode.DE 10 2019 220 196 A1 discloses a method for operating a vehicle having a drive train which has a plurality of drive units, at least one of which is designed as an electric machine, and an energy storage system for the drive units designed as an electric machine. In this case, an operating strategy for the drive train is determined by means of an artificial neural network, in which at least in parts plastic nodes are provided, and the drive train is operated in accordance with the operating strategy. The electric machine can also be operated as a generator, in particular for recuperation of braking energy.An object of the invention is to specify an improved method for operating a vehicle in an activated recuperation mode.The aforementioned object is achieved with the features of the independent claim.Advantageous embodiments and advantages of the invention are evident from the further claims, the description and the drawing.According to one aspect of the invention, a method for operating a vehicle in an activated recuperation mode is proposed, wherein the vehicle has at least one drive unit designed as an electric machine and a control and / or regulating device for executing a braking torque request by means of recuperation by the at least one electric machine, at least comprising capturing a current driving situation from current sensor values; ascertaining a current braking torque request by the control and / or regulating device to the at least one electric machine; ascertaining a current driver request from a current accelerator pedal position and from a current brake pedal position of the vehicle; ascertaining a difference between the current braking torque request by the control and / or regulating device and the current driver request; and ascertaining a correction value with the difference with which correction value the value of the braking torque request stored in the control and / or regulating device for the current driving situation is adapted.The braking torque represents a negative torque which is output to the drive unit for braking. The braking torque may be applied by the electric machine in the recuperation mode. The electric machine in this case feeds electrical energy back to the drive battery of the vehicle.The current implementation of an intelligent regenerative mode of an electrified vehicle provides the driver with a level of brake boost that does not meet all customers. According to the proposed method, braking torque requests, which are specified by a control and / or regulating device of the vehicle, can be corrected after detecting a reaction of the driver, which reaction is ascertained from the current driver requests via the accelerator pedal and brake pedal, in order thus to take into account the wishes of the driver.Advantageously, this allows a higher acceptance of the recuperation mode by the driver to be achieved.According to an advantageous embodiment of the method, the correction value can be ascertained from the difference between the current braking torque request by the control and / or regulating device and the current driver request by means of a self-learning adaptation function.The proposed method introduces a machine learning model in the form of a self-learning adaptation function, which implicitly takes into account the driver behavior. This ensures an individual braking torque behavior of the vehicle, which corresponds to the requirements of the driver. The behavior is modeled by taking into account how the driver overrides the torque of the control and / or regulating device acting on the system, wherein the driver actuates either the accelerator pedal or the brake pedal. Once the machine model has learned the driver behavior, it can supplement the torque request of the control and / or regulating device in such a way that it corresponds to the desired route of travel of the driver if a specific event precedes it.All signals of the vehicle relating to recuperation are recorded at high frequency. A data set contains the torque values from different sources, in particular the torque signal requested by the control and / or regulating device in the given situation and the actual wheel torque. Because of the time taken for the system to transmit the torque request of the control and / or regulating device to the torque coordination system and the torque until it arrives at the wheel, these two recorded values do not directly match. Indeed, there may be a significant delay of, for example, about 400 milliseconds between the torque signal sent by the controller and / or regulator and the resulting torque measured at the wheels. According to the proposed method, a regression model can be used which corrects this time offset.Advantageously, the proposed method can improve the individual driving comfort.Each driver has a different way to brake and accelerate the vehicle in certain situations. This is registered by the speed profile of the vehicle, the accelerator pedal position and the brake pressure over time. These inputs to the self-learning adaptation function are highly tailored to the individual driver. The model learns to predict the prediction of the difference in the requested torques between the control and / or regulating device and the driver. As the torques vary from driver to driver, the model learns online so that it has individual weighting parameters for each driver. As soon as the model with the prediction has reached a sufficient accuracy, the difference of the output torques is added to the output of the control and / or regulating device. The combined torque request acts on the vehicle such that only minimal driver intervention is required during travel.With the use of the proposed method, the intelligent recuperation mode can be used in an amplified manner. By introducing the self-learning adaptation function in the control and / or regulating device, the highly personalized regenerative braking of the vehicle can encourage more drivers to use this mode.According to an advantageous embodiment of the method, methods of machine learning can be used for the self-learning adaptation function. The model may involve past time dependent signals. Therefore, methods of time-series prediction and prediction models of machine learning can be favorably employed.According to an advantageous embodiment of the method, the braking torque request can be corrected by the control and / or regulating device by means of a delay time correction function which takes into account a delay between a braking torque request and a time of the occurrence of a corresponding torque at a vehicle wheel of a driven axle of the vehicle. In this way, the braking request can be adapted to the current needs of the driver in an advantageous manner.According to an advantageous embodiment of the method, the delay time correction function can have a regression model, coefficients of the regression model being predetermined. In particular, the coefficients of the regression model can be predetermined offline. Thus, the time delay between the braking torque request and the time of occurrence of a corresponding torque at the vehicle wheel can be corrected.According to an advantageous embodiment of the method, the braking torque request corrected with the delay time correction function can be used as input for the self-learning adaptation function. Thus, the driver's needs for the driving behavior in the current situation can be anticipated even better.According to an advantageous embodiment of the method, a current vehicle speed and / or a current wheel speed can be determined for detecting the current driving situation. With these values, current torque requests can be advantageously taken into account via a characteristic map which contains predefined brake requests for the measured parameters.According to an advantageous embodiment of the method, signals from cameras, and / or from a navigation system, and / or from at least one radar system can be included for ascertaining the correction value. Requested braking torques can be determined and / or corrected by this.According to an advantageous embodiment of the method, a current steering angle and / or acceleration sensor values can be included for ascertaining the correction value, by means of which current driving situation is determined. In this way, the driver behavior can be advantageously registered and taken into account for the correction of the braking torque request proposed by the control and / or regulating device.According to an advantageous embodiment of the method, a driver profile stored in the vehicle, in particular in the control and / or regulating device, can be included for ascertaining the correction value. The control and / or regulating device can thus take into account a learned driver behavior for the output of current braking torque requests.Further advantages are evident from the following description of the drawings. The drawings illustrate an embodiment of the invention. The drawings, specification and claims contain numerous features in combination. The skilled person will expediently also consider the features individually and summarize them to form meaningful further combinations.The following are shown: FIG. 1 shows a determination of a corrected torque from a torque request at the vehicle wheel by means of a delay time correction function according to the method according to the invention for operating a vehicle in an activated recuperation mode; FIG. 2 shows a determination of a differential torque by means of a self-learning adaptation function according to the method according to the invention; and FIG. 3 shows a determination of a personalized torque according to the method according to the invention.In the figures, identical or similar components are denoted by identical reference numerals. The figures merely show examples and should not be understood as limiting.The proposed method can advantageously supplement an existing control and / or regulating device 300 in order to deliver a desired braking torque. In implementing the method, various radar, map, and / or camera-based signals of the system are used to predict the difference 22 between the driver's desired speed profile and the vehicle's speed profile if the controller and / or regulator 300 would apply the braking torque without intervention.The driver's preferred speed profile is learned when the vehicle is in the activated recuperation mode and the driver engages by actuating the accelerator pedal or brake pedal. This learned difference 22 of the torques is added to the output torque 30 requested by the control and / or regulating device 300.The model time-dependently records signals as a time-series prediction model of machine learning. The method comprises two submodels. The first model, delay time correction function 100, is used to correct the system-related delay between the recording of particular signals.Since this is a driver-independent signal delay problem, an offline-learned regression model with previously defined coefficients can be used for this purpose. The second model, the self-learning adaptation function 200, uses this time-shifted signal to predict the difference in torques along with several other signals. The signals used as input to this self-learning adaptation function 200 vary widely from one driver to another. In this way, the model also implicitly learns the driver behavior. The self-learning adaptation function 200 thus learns online, wherein the same architecture is used, but different parameter weighting values are used depending on the driver.FIG. 1 shows a determination of a corrected torque 12 from a torque request 10 at the vehicle wheel by means of a delay time correction function 100 according to the method according to the invention for operating a vehicle in an activated recuperation mode.The vehicle has at least one drive unit designed as an electric machine and a control and / or regulating device 300 for carrying out a braking torque request by means of recuperation by the at least one electric machine.FIG. 1 shows a delay time correction function 100 which takes into account a delay between a braking torque request 10 and a time of occurrence of a corresponding torque at a vehicle wheel of a driven axle of the vehicle. In this case, a required wheel torque 10 is input into the delay time correction function 100 and a corrected torque of the intelligent recuperation is thus determined.The delay time correction function 100 can have, for example, a regression model, coefficients of the regression model being predetermined.The delay time correction function 100 can be trained offline, i.e. not on embedded hardware. In particular, the coefficients of the regression model can be predetermined offline. The training data set of braking torque request 10 and corrected torque can be stored in a data memory 50 in the vehicle.According to the method according to the invention, a current driving situation is detected from current sensor values 20. A current braking torque request 30 to the at least one electric machine is then ascertained by the control and / or regulating device 300. A current driver request 14 is ascertained from a current accelerator pedal position and from a current brake pedal position of the vehicle. From this, a difference 22 between the current braking torque request 30 is determined by the control and / or regulating device 300 and the current driver request 14. A correction value is then determined with the difference 22, with which the value 30 of the braking torque request stored in the control and / or regulating device 300 for the current driving situation is adapted.The correction value can be ascertained from the difference 22 between the current braking torque request by the control and / or regulating device 300 and the current driver request 14 by means of a self-learning adaptation function 200.Methods of machine learning can be used for the self-learning adaptation function 200.FIG. 2 shows a determination of a differential torque 22 by means of a self-learning adaptation function 200 according to the method according to the invention.The self-learning adaptation function 200 generates a differential torque 22 from an input of sensor values 20, which are used for ascertaining the correction value.These sensor values 20 can be, for example, signals from cameras and / or from a navigation system and / or from at least one radar system.Furthermore, the sensor values 20 can represent, for example, data of the current driving situation, such as a current vehicle speed and / or a current wheel speed.Furthermore, data such as a current steering angle and / or acceleration sensor values can be included as sensor values 20.The training of the self-learning adaptation function 200 can take place online on embedded hardware in the vehicle.During training, the braking torque request 16 corrected with the delay time correction function 100 can advantageously be used as input for the self-learning adaptation function 200.In this case, a driver request 14 is used as input for the delay time correction function 100 and a corrected braking torque request 16 is predicted therefrom. This braking torque request 16 can be linked to the braking torque request 30 of the control and / or regulating device 300 as differential torque 18 together with the sensor values 20 stored in the data memory 50.If both models, namely the delay time correction function 100 and the self-learning adaptation function 200, are trained accordingly, a determination of a personalized torque 40 can be carried out according to the method according to the invention during the driving operation of the vehicle, as illustrated in FIG. 3.The personalized torque 40 can thus be determined from the input of the sensor values 20 listed above by way of example both into the control and / or regulating device 300 and into the self-learning adaptation function 200.By means of the sensor values 20, a current braking torque request 30 can be determined with the control and / or regulating device 300. The self-learning adaptation function 200 can also be used to determine a differential torque 22, by means of which a correction value for the braking torque request 30 of the control and / or regulating device 300 is determined. With this correction value, the torque 40 personalized for the current driver can then be determined and applied.In order to ascertain the correction value, a driver profile stored in the vehicle, in particular in the control and / or regulating device 300 and / or the data memory 50, can additionally be included.List of reference characters10 Torque requested at vehicle wheel 12 Corrected torque 14 Driver request 16 Predicted torque 18 Differential torque 20 Sensor value 22 Differential torque 30 Torque request by control and / or regulating device 40 Personalized torque 50 Data memory 100 Delay time correction function 200 Self-learning adaptation function 300 Control and / or regulating deviceReferences included in the specificationThis list of documents cited by the applicant has been produced in an automated manner and is only included for the better information of the reader. The list is not part of the German patent application or utility model application. The DPMA does not take any adhesion for any faults or omissions.Patent Literature citedDE 10 2019 220 196 A1

[0002]

Claims

Method for operating a vehicle in an activated recuperation mode, wherein the vehicle has at least one drive unit designed as an electric machine and a control and / or regulating device (300) for executing a braking torque request by means of recuperation by the at least one electric machine, at least comprising - detecting a current driving situation from current sensor values (20); - ascertaining a current braking torque request (30) by the control and / or regulating device (300) to the at least one electric machine; - ascertaining a current driver request (14) from a current accelerator pedal position and from a current brake pedal position of the vehicle; - ascertaining a difference (22) between the current braking torque request (30) by the control and / or regulating device (300) and the current driver request (14); and - determining a correction value with the difference (22), wherein the value (30) of the braking torque request stored in the control and / or regulating device (300) for the current driving situation is adapted with the correction value.Method according to Claim 1, wherein the correction value is ascertained from the difference (22) between the current braking torque request by the control and / or regulating device (300) and the current driver request (14) by means of a self-learning adaptation function (200).Method according to Claim 2, wherein machine learning methods are used for the self-learning adaptation function (200).Method according to one of Claims 2 to 3, wherein the braking torque request (30) is corrected by the control and / or regulating device (300) by means of a delay time correction function (100) which takes into account a delay in time between a braking torque request (10) and a time at which a corresponding torque occurs at a vehicle wheel of a driven axle of the vehicle.Method according to claim 4, wherein the delay time correction function (100) comprises a regression model, wherein coefficients of the regression model are predetermined, in particular wherein the coefficients of the regression model are predetermined offline.Method according to Claim 4 or 5, wherein the braking torque request (16) corrected with the delay time correction function (100) is used as input for the self-learning adaptation function (200).Method according to one of the preceding claims, wherein a current vehicle speed and / or a current wheel speed are determined for detecting the current driving situation.Method according to one of the preceding claims, wherein signals from cameras, and / or from a navigation system, and / or from at least one radar system are included for determining the correction value.Method according to one of the preceding claims, wherein a current steering angle and / or acceleration sensor values are included for determining the correction value, by means of which current driving situation is determined.Method according to one of the preceding claims, wherein a driver profile stored in the vehicle, in particular in the control and / or regulating device (300), is included for determining the correction value.

Images