Control system for a motor vehicle and method for adjusting a control system

Abstract

The invention relates to a control system (100) for a motor vehicle for outputting a control variable (u) on the basis of which a controlled variable (y) of the motor vehicle can be adjusted by a suitable actuating process in order to adapt the controlled variable (y) to a given variable (w) of the control system (100), comprising a controller (110) which is configured to output a first output variable (u1) on the basis of the controlled variable (y) of the motor vehicle and on the basis of the given variable (w) of the control system (100), and further comprising a prediction model (120) which is trained to output a second output variable (u2) which depicts a deviation of the driving behavior of a driver of the motor vehicle from the first output variable (u1) of the controller (110), wherein the control variable (u) of the control system (100) comprises the sum of the first output variable (u1) and the second output variable (u2).

Description

Technical Field

[0001] This disclosure relates to a control system for outputting a control quantity for a motor vehicle, which, according to the control quantity, can adjust the controlled quantity of the motor vehicle through a suitable operating procedure so that the controlled quantity is suitable for a given quantity of the control system. Background Technology

[0002] Control systems are used in motor vehicles, for example, as driver assistance systems, to support the driver or reduce the driver's workload in specific driving situations.

[0003] To achieve the aforementioned assistance functions, the driver assistance system includes environmental sensors, such as radar sensors, lidar sensors, laser scanners, video sensors, and ultrasonic sensors. If the vehicle is equipped with a navigation system, the driver assistance system also utilizes data from that system. Furthermore, the driver assistance system, preferably connected to the vehicle's onboard electrical network via at least one bus, preferably a CAN bus, also actively intervenes in onboard systems, such as, in particular, the steering system, braking system, transmission system, and alarm system.

[0004] Typically, control systems available within a fleet utilize standardized data. In any situation, the control system can also be configured to either Sport or Comfort modes. Individualized configuration to suit individual driver behavior remains unknown.

[0005] Therefore, it is desirable to provide a control system that can achieve this individualized matching of driving behavior to an individualized driver. Summary of the Invention

[0006] This is achieved through the control system and method according to the independent claims.

[0007] A preferred embodiment relates to a control system for outputting a control quantity for a motor vehicle, which can adjust a controlled quantity of the motor vehicle through a suitable operating process based on the control quantity so that the controlled quantity is adapted to a setpoint of the control system. The control system includes: a controller configured to output a first output quantity based on the controlled quantity of the motor vehicle and a setpoint of the control system; and a prediction module trained to output a second output quantity, the second output quantity describing the deviation between the driving behavior of the driver of the motor vehicle and the first output quantity of the controller, wherein the control quantity of the control system includes the sum of the first output quantity and the second output quantity.

[0008] The controller of the control system includes, for example, conventional control algorithms, such as PID control structures. To enable the control system to match individualized driver behavior, a method is proposed that uses a predictive model to model the difference between the driving behavior and the existing controller. The control system matches the individualized driver behavior by adding a second output variable of the predictive model to a first output variable of the controller. This second output variable describes the deviation between the driver's driving behavior and the first output variable of the controller.

[0009] Control systems are, for example, driving assistance systems that can be used in motor vehicles to support or reduce the driver’s workload in specific driving situations, such as for distance control with reference objects, especially distance keeping assist systems or parking assist systems, or assist systems for integrating at least partially autonomous vehicles into traffic flow.

[0010] For distance control, appropriate maneuvers, such as acceleration and / or braking and / or steering intervention, are typically used to adjust the distance between the vehicle and a reference object to a desired value, i.e., to a given value from the control system. By matching the control quantity to the individualized driving behavior of the driver, the maneuvers can also be adapted to that behavior. Advantageously, this improves the acceptability of such a system.

[0011] In another preferred embodiment, the controlled quantity of the motor vehicle describes the distance between the motor vehicle and a reference object in the environment of the motor vehicle.

[0012] Reference objects in the environment of a motor vehicle include, for example, third vehicles, pedestrians, animals, or other road users, especially those traveling ahead. Alternatively, reference objects can be fixed environmental objects, such as safety barriers, trees, pillars, and buildings. Similarly, road markings, such as lane restrictions and lane lines, can also be considered reference objects.

[0013] In another preferred embodiment, the controller is specified to include conventional control structures, particularly PID control structures and / or predictive models, Gaussian process models, or neural networks.

[0014] In another preferred embodiment, the predictive model is trained to output a second output quantity based on at least one input quantity, wherein the input quantity includes one of the following variables: a setpoint of the control system, a controlled variable of the vehicle, variables representing the vehicle's operating data and / or environmental data of the vehicle. The setpoint of the control system is a desired value, and the controlled variable should be suited to said desired value. The vehicle's operating data includes, for example, speed, acceleration, steering angle, and tilt. The vehicle's environmental data includes, for example, information about street conditions, climate, road gradient, street orientation, etc. By using the aforementioned variables as inputs to the predictive model, it is advantageous to output a second output quantity based on said variables.

[0015] Other preferred embodiments relate to a computer-implemented method for training a predictive model for a control system for a motor vehicle according to the embodiments described, wherein a first training phase includes: determining a deviation between a driver's driving behavior of the motor vehicle in a deactivated state of the control system and a first output of a controller of the control system, and training the predictive model using the determined deviation of the driving behavior.

[0016] The deactivation state of the control system can be understood as the control system no longer being used to control driver assistance functions, but rather being controlled by the driver of the motor vehicle.

[0017] In another preferred embodiment, the first training phase further includes: determining the driver's driving behavior based on the controlled variable of the motor vehicle, and calculating a first output of the controller. Based on the calculated first output of the controller and the driving behavior determined when the control system is deactivated, the deviation between the driving behavior and the first output of the controller can be determined. Advantageously, the prediction model is trained using the deviation of the driving behavior determined based on the controlled variable of the motor vehicle.

[0018] In another preferred embodiment, determining driving behavior includes determining at least one variable representing accelerator pedal intervention and / or brake intervention and / or steering intervention.

[0019] In other preferred embodiments, training is specified based on at least one other variable representing the vehicle's operating data and / or its environmental data. The vehicle's operating data includes, for example, speed, acceleration, steering angle, and tilt. The vehicle's environmental data includes, for example, information about street conditions, weather, road gradient, street orientation, etc.

[0020] In another preferred embodiment, the second training phase of the method includes optimizing the prediction model based on at least one other variable associated with a reference object in the environment of the motor vehicle. The reference object is, for example, a third vehicle traveling ahead. By optimizing the prediction model with respect to the reference object, the prediction model can be advantageously optimized regarding the future position of the reference object.

[0021] In another preferred embodiment, the optimization of the prediction model further includes: determining the state of a motor vehicle, including at least one variable associated with the motor vehicle, at a certain moment determining the state of a reference object, including at least one variable associated with the reference object, determining a distribution with respect to future states, and identifying at least one model parameter that minimizes the expected value of the error in the distribution with respect to future states. It refers to the state of the motor vehicle at time t. This refers to the state of the reference object at time t. The distribution of future states is then derived using the following formula:

[0022] .

[0023] At any moment The error is passed through The model parameters that minimize the expected value of the error solve the following optimization problem:

[0024] ,

[0025] in T max Describes the maximum prediction range. The identified model parameters affect the time step. T max Minimizing the cumulative error is particularly advantageous, as it helps prevent long-term cumulative prediction errors.

[0026] In another preferred embodiment, the third training phase of the method includes examining the predictive model relative to driver intervention while the control system is in an active state. Driver intervention provides information about whether the control system can still depict the driver's driving behavior and / or how well the control system can depict the driver's driving behavior.

[0027] In another preferred embodiment, the first and / or second training phases are repeated based on the check of the predictive model, and / or other steps are implemented, particularly deactivating the control system and / or outputting alarm indications.

[0028] Other preferred embodiments relate to a computer program configured to implement the steps of the method according to the embodiments.

[0029] Other preferred embodiments relate to a machine-readable storage medium on which a computer program according to the embodiments is stored.

[0030] Other preferred embodiments relate to a control device configured to implement the steps of the method according to the embodiments.

[0031] Other preferred embodiments involve the application of a control system according to the embodiments and / or a predictive model trained by means of a method according to the embodiments and / or a computer program according to the embodiments and / or a machine-readable storage medium according to the embodiments and / or a control device according to the embodiments, for use in a motor vehicle control system to match the driver's individualized driving behavior.

[0032] Other preferred embodiments involve the application of the control system according to the embodiments and / or a predictive model trained by means of the methods according to the embodiments and / or the methods according to the embodiments and / or a computer program according to the embodiments and / or a machine-readable storage medium according to the embodiments and / or a control device according to the embodiments in a driver assistance system for motor vehicles, particularly for spacing control, also known as Adaptive Cruise Control (ACC). Attached Figure Description

[0033] Other features, applications, and advantages of the invention will become apparent from the following description of embodiments of the invention, which are illustrated in the accompanying drawings. Herein, all features described or shown, either alone or in any combination, form the subject matter of the invention, regardless of their generalization in the claims or references thereof, or their expression or illustration in the description or drawings.

[0034] In the attached diagram:

[0035] Figure 1 A schematic diagram of a control system for a motor vehicle is shown;

[0036] Figure 2 A schematic diagram illustrating the steps of the first training phase of a computer-implemented method for training a prediction model;

[0037] Figure 3 A schematic diagram illustrating the steps of the second training phase of a computer-implemented method for training a prediction model;

[0038] Figure 4 A schematic diagram illustrating the steps of the third training phase of a computer-implemented method for training a prediction model;

[0039] Figure 5 Showing according to Figure 3 A schematic overview of the second training phase, and

[0040] Figure 6 A schematic diagram of a control device for a motor vehicle is shown. Detailed Implementation

[0041] Figure 1 A control system 100 for outputting a control quantity u for a motor vehicle (not shown) is schematically shown. The controlled quantity y of the motor vehicle can be adjusted by a suitable operating process according to the control quantity so that the controlled quantity y is suitable for a given quantity w of the control system.

[0042] To achieve the aforementioned control process, the control system is preferably connected to the vehicle's onboard electrical network via at least one bus, preferably a CAN bus (not shown), so that by effectively intervening in the onboard systems, such as, in particular, the steering system, braking system, transmission system, and alarm system, the controlled variable y can be adapted to the given variable w of the control system.

[0043] The control system includes a controller 110 configured to output a first output quantity u1 based on the controlled variable y of the vehicle and the given variable w of the control system. The controller 110 of the control system 100 may include, for example, a conventional control algorithm, such as a PID control structure.

[0044] The control system 100 also includes a predictive model 120, which can be trained to output a second output quantity u2, the second output quantity describing the deviation between the driving behavior of the driver of the motor vehicle and the first output quantity u1 of the controller. The control quantity u of the control system 100, according to the illustrated embodiment, includes the sum of the first output quantity u1 and the second output quantity u2.

[0045] In order to match the control system 100 to the individualized driving behavior of the driver, the difference between the driving behavior and the existing controller 110 is modeled by means of the prediction model 120, and the control system 100 matches the individualized driving behavior of the driver by adding the second output u2 of the prediction model 120 to the first output u1 of the controller 110, the second output describing the deviation between the driving behavior of the driver of the motor vehicle and the first output u1 of the controller 110.

[0046] The control system 100 is, for example, a driving assistance system that can be used in a motor vehicle to support or reduce the burden on the driver in specific driving situations, such as for distance control from a reference object, especially a distance maintaining assist system or a parking assist system, or an assist system for integrating at least partially autonomous vehicles into traffic flow.

[0047] For distance control, appropriate maneuvers, such as acceleration and / or braking and / or steering intervention, are typically used to adjust the distance between the vehicle and a reference object to a desired value, i.e., to a given value from the control system. By matching the control quantity to the individualized driving behavior of the driver, the maneuvers can also be adapted to that behavior. Advantageously, this improves the acceptability of such a system.

[0048] In another preferred embodiment, the controlled quantity of the motor vehicle describes the distance between the motor vehicle and a reference object in the environment of the motor vehicle.

[0049] Reference objects in the environment of a motor vehicle include, for example, third vehicles, pedestrians, animals, or other road users, especially those traveling ahead. Alternatively, reference objects can be fixed environmental objects, such as safety barriers, trees, pillars, and buildings. Similarly, road markings, such as lane restrictions and lane lines, can also be considered reference objects.

[0050] To detect the distance between a vehicle and a reference object, the vehicle advantageously includes environmental sensors (not shown), such as radar sensors, lidar sensors, laser scanners, video sensors, and ultrasonic sensors. Data from a navigation system can also be used if the vehicle is equipped with one.

[0051] In another preferred embodiment, the controller 110 includes a conventional control structure, particularly a PID control structure and / or a predictive model 120, a Gaussian process model, or a neural network.

[0052] In another preferred embodiment, the prediction model 120 is trained to output a second output u2 based on at least one input, wherein the input includes one of the following variables: a setpoint w of the control system, a controlled variable y of the vehicle, variables representing the vehicle's operating data and / or environmental data. The setpoint w of the control system 100 is a desired value, and the controlled variable y should be suited to said desired value. The vehicle's operating data includes, for example, speed, acceleration, steering angle, and tilt. The vehicle's environmental data includes, for example, information about street conditions, climate, road gradient, street orientation, etc. By using the mentioned variables as inputs to the prediction model 120, the second output u2 can be advantageously output based on said variables. Advantageously, said variables are detected by suitable sensors, such as environmental sensors, and / or provided to the control system by suitable mechanisms for data transmission.

[0053] Figure 2 The steps of a first training phase of a computer-implemented method 200 for training a prediction model 120 for a motor vehicle control system 100 according to the embodiment are schematically shown. The first training phase includes the steps of: determining 220 the deviation between the driving behavior of the driver of the motor vehicle and the first output quantity u1 of the controller 110 of the control system 100 in a deactivated state of the control system 100, and training 230 the prediction model 120 with the determined deviation of the driving behavior.

[0054] The deactivation state of the control system 100 can be understood as the control system 100 not being used to control driving assistance functions, but rather being controlled by the driver of the motor vehicle.

[0055] In another preferred embodiment, the first training phase of method 200 further includes the steps of: determining the driver's driving behavior 210a based on the controlled variable y of the motor vehicle, and calculating the first output u1 of the controller 110. Based on the calculated first output u1 of the controller 110 and the driving behavior determined when the control system 100 is deactivated, the deviation between the driving behavior and the controller's first output u1 can be determined. Advantageously, the prediction model 120 is trained using the deviation of the driving behavior determined based on the controlled variable y of the motor vehicle.

[0056] In another preferred embodiment, determining the driving behavior 210a includes determining at least one variable representing accelerator pedal intervention and / or brake intervention and / or steering intervention.

[0057] In other preferred embodiments, the predictive model 120 is trained based on at least one other variable representing the vehicle's operating data and / or its environmental data. The vehicle's operating data includes, for example, speed, acceleration, steering angle, and tilt. The vehicle's environmental data includes, for example, information about street conditions, weather, road gradient, street orientation, etc.

[0058] In another preferred embodiment, the second training phase of the method includes optimizing the prediction model based on at least one other variable associated with a reference object in the environment of the motor vehicle. The reference object is, for example, a third vehicle traveling ahead. By optimizing the prediction model 120 with respect to the reference object, the prediction model 120 can advantageously optimize for the future position of the reference object.

[0059] In another preferred embodiment, the optimization of the prediction model 120 further includes: determining the state of a motor vehicle, including at least one variable associated with the motor vehicle, at a time t; determining the state of a reference object, including at least one variable associated with the reference object, at that time t; determining a distribution with respect to future states; and identifying at least one model parameter that minimizes the expected value of the error in the distribution with respect to future states. The model parameters characterize the connection between the inputs and outputs of the prediction model 120.

[0060] Advantageously, this prevents the occurrence of long-term cumulative prediction errors. Long-term cumulative errors can occur particularly when the prediction model 120 fails to accurately depict deviations in driving behavior.

[0061] Figure 5A schematic overview of a second training phase for optimizing prediction model 120 based on at least one additional variable associated with a reference object in the environment of the motor vehicle is shown. The variable associated with the reference object in the environment of the motor vehicle is provided by a second prediction model 130, which is adapted to predict the state of the reference object. A third prediction model 140 integrates controller 110 and prediction model 120 and is thus adapted to predict the state of the motor vehicle.

[0062] This indicates the state of the motor vehicle at time t, which is advantageous. This includes all variables provided to the prediction model 120 and the controller 110. This indicates the state of a reference object, such as a third vehicle traveling ahead, at time t, particularly information about its position and / or speed. The distance from the reference object at time t is also provided to the prediction model 120 and the controller 110.

[0063] If at least one predictive model 120, 130, 140 or controller 110 is a stochastic model, then a distribution about the future state can be derived from it, which is obtained by the following formula:

[0064] .

[0065] In the future state at time The error is passed through Errors are derived, such as the difference between the measurement and a given quantity and / or the excess or deficiency of the maximum or minimum permissible difference. The model parameters that minimize the expected value of the error solve the following optimization problem:

[0066] ,

[0067] in T max Describes the maximum prediction range. The identified model parameters affect the time step. T max The cumulative error is minimized. Advantageously, the prediction model 120 is optimized based on this.

[0068] In another preferred embodiment, the third training phase of method 200 includes: examining the prediction model 250 relative to driver intervention while the control system is in an active state. A schematic diagram of the steps of the third training phase of the computer-implemented method 200 is shown in... Figure 4 As shown in the image.

[0069] In another preferred embodiment, the first and / or second training phases are repeated based on the check 250 of the prediction model 120, and / or other steps are implemented, particularly deactivating the control system 100 260a and / or outputting the alarm indication 200b.

[0070] Other preferred embodiments relate to a computer program configured to implement the steps of method 200 according to the embodiments.

[0071] Other preferred embodiments relate to a machine-readable storage medium on which a computer program according to the embodiments is stored.

[0072] Other preferred embodiments relate to a control device 300 configured to implement the steps of the method 200 according to the embodiments described above. The control device 300 includes a computing device 310 and at least one storage device 320 on which the control system 100 is stored. Furthermore, the control device 300 includes an input terminal 330 for receiving information about variables of the control system, such as setpoints and controlled variables, and other variables representing operating data of the motor vehicle and / or environmental data of the motor vehicle. Advantageously, the variables are detected by suitable sensors, such as environmental sensors, and / or provided to the control system by suitable mechanisms for data transmission. Additionally, the control device 300 includes output terminals 340 for actuators used to operate the vehicle's onboard systems, particularly the steering system, braking system, transmission system, and alarm system.

[0073] Other preferred embodiments involve the application of a control system 100 according to the embodiments and / or a predictive model 120 trained by means of a method 200 according to the embodiments and / or a computer program according to the embodiments and / or a machine-readable storage medium according to the embodiments and / or a control device 300 according to the embodiments, for use in a motor vehicle control system 100 to match the driver's individualized driving behavior.

[0074] Other preferred embodiments involve the application of the control system 100 according to the embodiments and / or the predictive model 120 trained by the method 200 according to the embodiments and / or the method 200 according to the embodiments and / or the computer program according to the embodiments and / or the machine-readable storage medium according to the embodiments and / or the control device 300 according to the embodiments in a driver assistance system for motor vehicles, particularly for spacing control, also known as Adaptive Cruise Control (ACC).

Claims

1. A control system (100) for a motor vehicle for outputting a control quantity (u) from which a controlled quantity (y) of the motor vehicle can be adjusted by a suitable manipulation process in order to adapt the controlled quantity (y) to a given quantity (w) of the control system (100), the control system comprising: a controller (110) configured to output a first output quantity (ul) based on the controlled quantity (y) of the motor vehicle and based on the given quantity (w) of the control system (100), and further comprising a prediction model (120) trainable to output a second output quantity (u2) depicting a deviation of a driving behavior of a driver of the motor vehicle from the first output quantity (ul) of the controller (110), wherein the control quantity (u) of the control system (100) has an addition of the first output quantity (ul) and the second output quantity (u2).

2. Control system (100) according to claim 1, wherein the controlled quantity (y) of the motor vehicle depicts a distance of the motor vehicle to a reference object in an environment of the motor vehicle.

3. Control system (100) according to claim 1 or 2, wherein the controller (110) comprises a control structure and / or the prediction model (120) comprises a Gaussian process model or a neural net.

4. Control system (100) according to claim 3, wherein the control structure comprises a PID control structure.

5. Control system (100) according to claim 1 or 2, wherein the prediction model (120) is trainable to output the second output quantity (u2) depending on at least one input quantity, wherein input quantities comprise one of the following variables: the given quantity (w) of the control system (100), the controlled quantity (y) of the motor vehicle, a variable representing operational data of the motor vehicle and / or environmental data of the motor vehicle.

6. A computer-implemented method (200) for training a prediction model for use in a control system according to any one of claims 1 to 5, wherein a first training phase comprises: In a deactivated state of the control system, a deviation of a driving behavior of a driver of the motor vehicle from a first output quantity of a controller of the control system is determined (220), and the prediction model is trained (230) by means of the determined deviation of the driving behavior.

7. Method (200) according to claim 6, wherein the first training phase further comprises: a driving behavior of the driver is determined (210a) depending on the controlled quantity of the motor vehicle, and a first output quantity of the controller is calculated (210b).

8. Method (200) according to claim 7, wherein determining (210a) the driving behavior comprises determining at least one variable representing an acceleration pedal intervention and / or a brake intervention and / or a steering intervention.

9. Method (200) according to any one of claims 6 to 8, wherein training (230) is performed depending on at least one variable representing operational data of the motor vehicle and / or environmental data of the motor vehicle.

10. Method (200) according to any one of claims 6 to 8, wherein the second training phase comprises: the prediction model (120) is optimized (240) depending on at least one variable associated with a reference object in an environment of the motor vehicle.

11. Method (200) according to claim 10, wherein optimizing (240) the prediction model (120) further comprises: a state of the motor vehicle comprising at least one variable associated with the motor vehicle is determined (240a) at one point in time; determining (240b) a state of the reference object comprising at least one variable associated with the reference object at the instant; and determining (240c) a distribution about a future state and identifying (240d) at least one model parameter that minimizes an expected value of an error in the distribution about the future state.

12. The method (200) according to any one of claims 6 to 8, wherein the third training phase comprises: checking (250) the prediction model (120) with respect to an intervention of the driver in the active state of the control system (100).

13. The method (200) according to claim 10, wherein in accordance with the checking (250) of the prediction model (120), the first training phase and / or the second training phase are repeated and / or further steps are implemented.

14. The method (200) according to claim 13, wherein the further steps comprise deactivating (260a) the control system (100) and / or outputting (260b) a warning indication.

15. A computer program product comprising a computer program configured to implement the steps of the method (200) according to any one of claims 6 to 14.

16. A machine-readable storage medium having stored thereon a computer program configured to implement the steps of the method (200) according to any one of claims 6 to 14.

17. A control device (300) configured to implement the steps of the method (200) according to any one of claims 6 to 14.

18. Use of the control system (100) according to any one of claims 1 to 5 and / or of a prediction model (120) trained by means of the method (200) according to any one of claims 6 to 14 and / or of the method (200) according to any one of claims 6 to 14 and / or of the computer program product according to claim 15 and / or of the machine-readable storage medium according to claim 16 and / or of the control device (300) according to claim 17 for matching a control system (100) for a motor vehicle to an individualized driving behavior of a driver.

19. Use of the control system (100) according to any one of claims 1 to 5 and / or of a prediction model (120) trained by means of the method (200) according to any one of claims 6 to 14 and / or of the method (200) according to any one of claims 6 to 14 and / or of the computer program product according to claim 15 and / or of the machine-readable storage medium according to claim 16 and / or of the control device (300) according to claim 17 in a driver assistance system of a motor vehicle for adaptive cruise control.

Images