Computer-implemented method for determining at least one friction coefficient profile and vehicle

Abstract

The invention relates to a computer-implemented method (100) for determining at least one friction coefficient profile (30) of at least one surface section (38), in particular a route section, comprising at least the following steps: providing (102) at least one received input signal (24) indicating spatially resolved image information of the at least one surface section (38) to an input level of a first artificial neural network (28) designed to estimate at least one friction coefficient profile (30) at least for the at least one surface section (38); providing (104) the at least one friction coefficient profile (30) to a second artificial neural network (32) designed to derive vehicle parameters, in particular chassis settings, and / or recommended actions, in particular instructions for a vehicle driver, adapted to the at least one friction coefficient profile (30);and providing (106) an output signal (34) indicating the adjusted vehicle parameters and / or recommended actions. This describes a computer-implemented method (100) for further improving the driving experience of a vehicle.

Description

[0001] The invention relates to a computer-implemented method for determining at least one friction coefficient profile and to a vehicle.

[0002] When driving a vehicle, adjusting the chassis to the road conditions is beneficial in order to optimize the vehicle's steering and handling characteristics. The coefficient of friction of a road is particularly important on racetracks, as it directly influences the speed at which the vehicle can travel around the track.

[0003] DE 10 2019 211 052 A1 discloses training a neural network to estimate the coefficient of friction of a vehicle wheel against a surface. For this purpose, images can be provided to the neural network in an input stage. The neural network can estimate a coefficient of friction from the image. If a low coefficient of friction has been estimated for a position, the low coefficient of friction can be provided to a partially automated vehicle so that the vehicle can react accordingly with a modified driving style. Furthermore, a vehicle occupant can be instructed to adjust their driving style to a low coefficient of friction.

[0004] DE 10 2017 123 002 A1 discloses a vehicle that uses data from different types of on-board sensors to determine whether meteorological precipitation is falling near the vehicle. The vehicle may include an on-board camera that captures image data and an on-board accelerometer that collects accelerometer data. The image data may characterize an area in front of or behind the vehicle. The accelerometer data may characterize vibrations of a windshield of the vehicle. An artificial neural network may execute computer hardware carried on-board the vehicle. The artificial neural network may be trained to classify meteorological precipitation in an environment of the vehicle using the image data and the accelerometer data as inputs.The artificial neural network classifications can be used to control one or more functions of the vehicle, such as windshield wiper speed or traction control settings, or the like.

[0005] DE 10 2018 008 788 A1 discloses a method for determining road surface roughness using at least one camera and an artificial neural network. The artificial neural network is trained as a deep artificial neural network in such a way that texture and / or roughness parameters are determined for a given image. Existing texture and / or roughness parameters from map data of a digital map with corresponding data from the image are used as training data. Images are captured by the at least one camera and fed to the trained artificial neural network, and the trained artificial neural network determines road surface roughness based on the training data from the supplied images.

[0006] DE 10 2019 208 234 A1 describes a method for automatically executing a control function of an autonomously or semi-autonomously controlled vehicle to improve safety in autonomous or semi-autonomous driving, which method comprises the following steps: determining sensor data by means of at least one sensor of a vehicle, processing the sensor data by means of an artificial neural network, wherein the sensor data or data derived from the sensor data are fed to the artificial neural network as input variables and the artificial neural network maps the input variables to output variables depending on a parameterizable processing chain, determining at least one value of at least one environmental variable,Classifying the acquired sensor data as a function of the output variables determined by the artificial neural network and as a function of the at least one determined value of the at least one environmental variable, and executing the control function as a function of the classification. Furthermore, the invention provides a control system configured to carry out the method, as well as a vehicle with such a control system.

[0007] The object of the invention is to provide a computer-implemented method with which a vehicle journey is further improved.

[0008] 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.

[0009] According to a first aspect, a computer-implemented method for determining at least one friction coefficient profile of at least one surface section, in particular a route section, is described, comprising at least the following steps: providing at least one received input signal indicating spatially resolved image information of the at least one surface section to an input level of a first artificial neural network configured to estimate at least one friction coefficient profile for at least the at least one surface section; providing the at least one friction coefficient profile to a second artificial neural network configured to derive vehicle parameters adapted to the at least one friction coefficient profile, in particular chassis settings, and / or recommended actions, in particular instructions for a vehicle driver;and providing an output signal indicating the adjusted vehicle parameters and / or recommended actions;

[0010] With the computer-implemented method, the driving characteristics of the vehicle or the driver's actions can be directly adjusted to the estimated friction coefficients along a route. The route can be a race track, for example. The race track can extend in a loop. To estimate the friction coefficient profile, spatially resolved image information of the surface section is provided to a first artificial neural network at an input level of the first artificial neural network. The spatially resolved image information shows, for example, the asphalt of a route. The first artificial neural network is designed to evaluate the spatially resolved image information of the at least one surface section. At least one friction coefficient profile of the surface section is then provided at the output level of the first artificial neural network.The friction coefficient profile can also be provided with spatial resolution. Furthermore, the friction coefficient profile can have the same spatial resolution as the image information. This means that a friction coefficient can be assigned to each point on the surface section. The friction coefficient profile is provided to a second artificial neural network, which is designed to determine adjusted vehicle parameters and / or recommended actions from the friction coefficient profile. The vehicle parameters can, in particular, be chassis settings for a vehicle's chassis. Furthermore, the adjusted vehicle parameters can, for example, be an adjustment of a stability control system or an engine control system. The recommended actions can, for example, be instructions for a vehicle driver. Instructions can, for example, be: brake earlier at curve X, more camber on the front axle, 0.2 bar more air pressure at the rear, etc.This allows the grip potential for the corresponding surface section to be optimally utilized, so that the ride of a vehicle can be further improved.

[0011] According to some embodiments, it is conceivable that the first artificial neural network can be designed as an encoder-decoder network.

[0012] By using an encoder-decoder network, relationships between the spatially resolved image information are taken into account with increased accuracy when estimating the friction coefficient profile for the surface section.

[0013] According to some embodiments, it is conceivable that the first artificial neural network can be designed to estimate at least one friction coefficient profile for a route section that is larger than the at least one surface section.

[0014] The first artificial neural network can then estimate a friction coefficient profile for a route section based on the spatially resolved image information of the surface section. The route section can have the surface section and extend beyond the surface section. The route section can, for example, be an entire route. For use of the computer-implemented method, it is then sufficient to provide spatially resolved image information of a comparatively small part of the route, namely the surface section, in order to obtain an estimated friction coefficient profile for the entire route. High-resolution image information can also be provided as an input signal for several surface sections at different positions on the route in order to increase the accuracy of the friction coefficient profile for the entire route or the route section.

[0015] According to some embodiments, it is conceivable that further information, preferably temperature, position, time of day and / or precipitation conditions, can be provided to the first artificial neural network for deriving the adapted vehicle parameters and / or recommended actions.

[0016] This information can be used to improve the estimation of the friction coefficient profile by the first artificial neural network. For example, precipitation on the route or a low temperature can cause a reduction in the friction coefficient of the route.

[0017] According to some embodiments, it is conceivable that the second artificial neural network can be designed as a large language model for outputting the adapted vehicle parameters and / or recommended actions as plain text.

[0018] This allows a driver to adjust the vehicle, particularly the chassis, and / or their driving style to the estimated friction coefficients along the route. This can increase the speed at which the vehicle can travel along the route.

[0019] According to some embodiments, it is conceivable that the method may further comprise at least the following step: generating a control signal based on the adjusted vehicle parameters, wherein the control signal indicates control values for adjustable systems of a vehicle.

[0020] In this exemplary embodiment, the method can be used to directly influence the vehicle, in particular the chassis, in order to adapt the vehicle to the friction coefficient profile. For example, the adjustable systems can be repeatedly adjusted using the control signal to react to changes in the friction coefficient profile along the route or surface section. The driver can then concentrate on controlling the vehicle, allowing the vehicle's speed to be further increased.

[0021] In some embodiments, the method can be carried out in real time, so that when driving on a race track, a local change in the coefficient of friction on the track can be reacted to and the systems can be adapted in real time to the locally applicable coefficient of friction.

[0022] According to some embodiments, it is conceivable that the input signal can display spatially resolved image information of at least two surface sections of a route, wherein the surface sections extend at least partially parallel to one another and / or overlapping along the route.

[0023] For example, spatially resolved image information can be recorded with a vehicle traveling along a first surface section of the route. When the vehicle is again close to the first surface section, further spatially resolved image information can be recorded. Since the vehicle generally does not travel along the route on exactly the same trajectory, the further spatially resolved image information relates to a second surface section that runs at least partially parallel and / or overlapping along the route to the first surface section. Alternatively or additionally, the further image information for the second surface section can be determined by a second vehicle traveling along the second surface section. This can increase the accuracy of the estimation for the friction coefficient profile of the entire route.

[0024] According to a second aspect, a computer program product is described, comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of the method according to the preceding description.

[0025] Advantages and effects as well as further developments of the computer program product arise from the advantages and effects as well as further developments of the method described above. In this regard, reference is therefore made to the preceding description. A computer program product can be understood, for example, as a data carrier on which a computer program element is stored that has 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 device, such as flash memory or RAM, that has the computer program element. However, this does not exclude other types of data storage devices that have the computer program element.

[0026] According to a third aspect, a vehicle is described, comprising at least one processor unit, wherein the processor unit is designed to carry out the steps of the computer-implemented method according to the preceding description, wherein the processor unit is designed to receive at least one input signal comprising spatially resolved image information of the at least one surface section.

[0027] Advantages and effects, as well as further developments of the vehicle, arise from the advantages and effects, as well as further developments of the method described above. To avoid repetition, reference is made to the previous description in this regard.

[0028] According to some embodiments, it is conceivable that the vehicle may further comprise at least one adjustable system, and the adjustable system is configured to receive a control signal from the processor unit and to adjust according to the control signal.

[0029] The adjustable system may include, for example, a stability control system, a chassis, an air-filled wheel or an engine control system.

[0030] According to some embodiments, it is conceivable that the vehicle may further comprise at least one display device which is designed to display the adjusted vehicle parameters and / or recommended actions of the output signal from the processor unit.

[0031] The display device can be, for example, a screen inside the vehicle, a head-up display, or a helmet display for the driver. When using a helmet display, the driver can receive recommended actions without having to take their eyes off the road. Both the helmet display and the display device can be connected to the processor unit via a signal connection. The signal connection can be wireless or wired.

[0032] The invention is described below using an exemplary embodiment with the aid of the accompanying drawings. They show: Fig. 1 a flowchart of the computer-implemented method; Fig. 2 a schematic representation of a route; Fig. 3 a schematic representation of a vehicle; Fig. 4 a schematic representation of the processor unit, at least one display device and at least one adjustable system.

[0033] The computer-implemented method is carried out according to Fig. 1 is hereinafter referred to in its entirety by the reference numeral 100.

[0034] The computer-implemented method 100 is designed to determine at least one friction coefficient profile 30 of at least one surface section 38. The surface section 38 can be part of a route section of a route 36, which in Fig. 2 is shown as an example.

[0035] According to a first step 102, at least one input signal is provided, which was received, for example, from an input interface. The input interface can be, for example, a signal connection of a processor unit on which the computer-implemented method is executed. The input signal indicates spatially resolved image information of the at least one surface section 38, which is schematically shown in Fig. 3 are designated by reference symbol 24. The spatially resolved image information 24 can, for example, comprise pixels of a digital image that was captured, for example, with a camera. The image information 24 can, for example, display an asphalt surface, which can form the surface section 38 of the route for 36.

[0036] The input signal is provided to an input level of a first artificial neural network 28. The neural network 28 can be trained to estimate at least one friction coefficient profile 30 for at least the surface section 38. Accordingly, the first artificial neural network 28 provides a friction coefficient profile 30 at its output level that is associated with at least the surface section 38 from which the spatially resolved image information 24 originates.

[0037] Alternatively or additionally, the input signal may comprise the spatially resolved image information of multiple surface sections 38 of a route 36. The surface sections 38 may be arranged parallel to one another and / or overlap in some areas.

[0038] The first artificial neural network 28 can be configured, for example, as an encoder-decoder network. The first artificial neural network 28 can be configured such that the spatial resolution of the friction coefficient profile 30 corresponds to the spatial resolution of the spatially resolved image information 24. Thus, for example, the first artificial neural network 28 can determine a friction coefficient for each pixel of an image of a surface section 38.

[0039] For example, several vehicles can determine spatially resolved image information 24 about several surface sections 38 along the route 36 in advance. In doing so, surface sections 38 are used or traveled over that can extend parallel to one another at least in sections and / or can overlap. Alternatively or additionally, a vehicle can determine spatially resolved image information for several surface sections 38 in several passes along the route 36. In this exemplary embodiment, too, the surface sections can then be parallel to one another at least in sections and / or overlap at least in regions.

[0040] Furthermore, the first artificial neural network 28 can provide the friction coefficient profile 30 for a route section that may be larger than the surface section 38. For example, the first artificial neural network 28 can provide a friction coefficient profile 30 for the entire route 38.

[0041] Furthermore, according to an optional step 108, at least one further piece of information can be provided to the first artificial neural network 28, which can be independent of the spatially resolved image information. For example, at least one piece of information 26 about the temperature, position, time of day, and / or precipitation can be provided. It is also conceivable to provide additional information.

[0042] This information can also be provided to an input level or part of an input level of the first artificial neural network 28. With this information, the first artificial neural network 28 can provide a more accurate friction coefficient profile 30 for the surface section 38.

[0043] According to a further step 104, the friction coefficient profile 30 can be provided to a second artificial neural network 32. The second artificial neural network 32 can be trained to provide vehicle parameters adapted from a friction coefficient profile and / or recommended actions for controlling the vehicle on the surface section 38.

[0044] The vehicle parameters can be, for example, chassis settings. Vehicle parameters can be, for example, parameters of the suspension and / or shock absorbers. The air pressure of a wheel, for example, can also be a vehicle parameter that can be adjusted.

[0045] According to a further optional step 110, the second artificial neural network 32 can be configured to generate a control signal indicating control values for adjustable systems of a vehicle. The control values can be generated from the adjusted vehicle parameters and / or recommended actions.

[0046] Recommendations for action can, for example, be instructions to a driver on how to steer the vehicle. For example, a recommendation for action could be to brake before a certain curve and / or to manually adjust the chassis inclination to a specific value for a specific section of the route, especially the surface section.

[0047] For this purpose, the second artificial neural network 32 can be configured, for example, as a large language model. The large language model can provide adjusted vehicle parameters and / or recommended actions in text form based on the friction coefficient profile 30. The text containing the adjusted vehicle parameters and / or recommended actions can optionally be converted into an audio signal to be provided to a driver alternatively or additionally in spoken form.

[0048] The adjusted vehicle parameters and / or recommended actions can be provided as output signal 34 according to step 106. Furthermore, output signal 34 can include the control signal explained above.

[0049] If the output signal 34 contains recommendations for action for the driver, the output signal 34 can be sent, for example, to a display device 14 of a vehicle 10, which is shown, for example, in Fig. 4. Alternatively or additionally, the output signal 34 can be transmitted to a helmet display 20 of the vehicle driver. The vehicle driver can then, for example, read the recommended actions directly from the visor of a helmet he is wearing.

[0050] If the output signal 34 comprises, for example, a control signal with control values for adjustable systems of a vehicle, the output signal 34 can be transmitted to the corresponding adjustable systems 16, 22. The adjustable systems 16, 22 of the vehicle 10 can be, for example, the suspension and / or shock absorbers 16. Alternatively or additionally, an adjustable system of the vehicle 10 can also be a vehicle wheel 22 that is filled with air. In the former example, the control value can indicate a suspension force or a damping force, and in the latter example, for example, an air pressure that is adapted to the friction coefficient profile of the surface sections 36 arranged in front of the vehicle.

[0051] Further adjustable systems 16, 22 can be, for example, stability management or engine control.

[0052] The vehicle 10 may further comprise an adjustable system 12 on which the first artificial neural network 28 and / or the second artificial neural network 32 may be implemented. Alternatively or additionally, the processor unit 12 may be associated with an external computer or server, in which case at least one of the display devices 14, 20 and / or at least one of the adjustable systems 16, 22 may be connected to the processor unit 12 via a wireless signal connection 18.

[0053] If the processor unit 12 is assigned to an external computer or server, the processor unit 12 can regularly access greater computing resources than in a vehicle 10. The processor unit 12 can then execute the computer-implemented method 100 more quickly; in particular, the friction coefficient profile 30 corresponding to a position of the vehicle 10 on the route 38 can then be provided to the second artificial neural network 32 as needed. Thus, a corresponding action instruction can be provided for each position of the vehicle 10 on the route 38.

[0054] The input signal can also be transmitted to the processor unit 12 via a wireless signal connection 18.

[0055] Furthermore, in a further embodiment, the first artificial neural network 28 can be executed temporally separately from the second artificial neural network 32. Thus, a friction coefficient profile 30 can first be determined from an input signal using the first artificial neural network 28. If necessary, the friction coefficient profile 30 or a required portion of the friction coefficient profile 30 can then be provided to the second artificial neural network 32 in order to obtain at least one instruction or vehicle parameter.

[0056] The example described above does not limit the invention in any way. Rather, the invention can be modified in many ways. All of the features of the invention described above can be essential to the invention alone or in combination with one another. List of reference symbols 10 vehicles 12 Processor unit 14 Display device 16 adjustable system 18 Signal connection 20 Display device 22 adjustable system 24 input signal 26 further information 28 first artificial neural network 30 Friction profile 32 second artificial neural network 34 Output signal 36 route 38 Surface section

Claims

[1] Computer-implemented method (100) for determining at least one friction coefficient profile (30) of at least one surface section (38), in particular a route section, comprising at least the following steps: a. Providing (102) at least one received input signal (24) indicating spatially resolved image information of the at least one surface section (38) to an input level of a first artificial neural network (28) configured to estimate at least one friction coefficient profile (30) at least for the at least one surface section (38); b. Providing (104) the at least one friction coefficient profile (30) to a second artificial neural network (32) which is designed to derive vehicle parameters adapted to the at least one friction coefficient profile (30), in particular chassis settings, and / or recommended actions, in particular instructions for a vehicle driver; and c. Providing (106) an output signal (34) indicating the adjusted vehicle parameters and / or recommended actions. [2] Computer-implemented method (100) according to claim 1, characterized by that the first artificial neural network (28) is designed as an encoder-decoder network. [3] Computer-implemented method (100) according to claim 1 or 2, characterized by that the first artificial neural network (28) is designed to estimate at least one friction coefficient profile (30) for a route section which is larger than the at least one surface section (38). [4] Computer-implemented method (100) according to one of the preceding claims, characterized bythat further information (26), preferably temperature, position, time of day and / or precipitation conditions, is provided (108) to the first artificial neural network for deriving the adapted vehicle parameters and / or recommended actions. [5] Computer-implemented method (100) according to one of the preceding claims, characterized by that the second artificial neural network (32) is designed as a large language model for outputting the adapted vehicle parameters and / or recommended actions as plain text. [6] Computer-implemented method (100) according to one of the preceding claims, characterized by that the method (100) further comprises at least the following step: a. generating (110) a control signal as an output signal (34) based on the adjusted vehicle parameters, wherein the control signal indicates control values for adjustable systems of a vehicle (10). [7] Computer-implemented method (100) according to one of the preceding claims, characterized by that the input signal (24) displays spatially resolved image information of at least two surface sections (38) of a route (36), wherein the surface sections (38) extend at least partially parallel to one another and / or overlapping along the route (36). [8] A computer program product comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method (100) according to any one of claims 1 to 7. [9] Vehicle (10) comprising at least one processor unit (12), wherein the processor unit (12) is designed to carry out the steps of the computer-implemented method (100) according to one of claims 1 to 7, wherein the processor unit (12) is designed to receive at least one input signal (24) comprising spatially resolved image information of the at least one surface section (38). [10] Vehicle (10) according to claim 9, characterized by that the vehicle (10) further comprises at least one adjustable system (16, 22), and the adjustable system (16, 22) is adapted to receive a control signal from the processor unit (12) and to adjust according to the control signal. [11] Vehicle (10) according to claim 9 or 10, characterized bythat the vehicle (10) further comprises at least one display device (14, 20) which is designed to display the adapted vehicle parameters and / or recommended actions of the output signal (34) from the processor unit (12).

Images