System for controlling the braking behavior of a vehicle

Abstract

The invention relates to a system for controlling the braking behavior of a vehicle, comprising a central or decentralized control unit (13, 33) which is connected to a plurality of vehicle sensors (5, 7, 9, 11), one of which (5) is coupled to the brake pedal (3) for determining its position or movement. In a system that allows the driver to continue influencing vehicle deceleration despite a detected or predictable locking of the brake pedal, the control unit (13, 33) includes an artificial intelligence (15) trained to predict various locking states of the brake pedal (3) from the relationships between the driver, the behavior of the brake pedal (3), and the properties of the foreign object, and to adjust the brake pedal tension accordingly, enabling the driver to continue modulating the deceleration within the remaining travel of the brake pedal (3).

Description

[0001] The invention relates to a system for controlling a braking behavior of a vehicle, comprising a central control unit which is connected to a plurality of vehicle sensors, of which a sensor for determining a position or movement of a brake pedal is coupled thereto.

[0002] DE 10 2008 029 311 A1 discloses a braking system for a motor vehicle and a method for controlling a vehicle brake. The brake pedal unit comprises a sensor with its own intelligence in the form of an electronic processing unit, which directly calculates control signals for brake actuators. Furthermore, the braking system comprises a central electronic control unit, which generates control signals for actuators based on the signals from the sensor in the brake pedal unit or another sensor. The actuators, which also have their own intelligence in the form of an electronic processing unit, check the validity of the signals output by the central control unit. If the control signal output by the central control unit is valid, it is executed by the actuators.If the control signal from the central control unit is not valid, the actuators are controlled based on the control signal calculated by the sensor of the brake pedal unit.

[0003] The object of the invention is to provide a system for controlling the braking behavior of a vehicle, which allows the driver to continue to influence the vehicle deceleration despite a detected or foreseeable blockage of the brake pedal.

[0004] The invention is based on the features of the independent claims. Advantageous developments and refinements are the subject of the dependent claims.

[0005] Further features, possible applications and advantages of the invention will become apparent from the following description and the explanation of embodiments of the invention shown in the figures.

[0006] The problem is solved by the subject matter of patent claim 1.

[0007] In the system explained at the outset for controlling the braking behavior of a production vehicle, comprising a central or decentralized control unit which is connected to a plurality of vehicle sensors, of which a sensor is coupled to determine the position or movement of a brake pedal, the control unit of a production vehicle is designed to detect a currently occurring blockage of the brake pedal or to predict an impending blockage with - a force sensor arranged on a brake pedal plate to check the proportionality of the actuating force to the brake pedal travel, - an accelerator pedal travel sensor for detecting unintentional actuation, - an ultrasonic sensor located in the footwell of the vehicle to detect foreign objects behind or next to the brake pedal via distance detection, - a camera located in the footwell of the vehicle to detect foreign objects behind or next to the brake pedal by identifying objects on an image of the footwell, - a force sensor located on the floor of the vehicle behind the brake pedal to detect foreign bodies behind the brake pedal by means of the brake pedal force transmitted by the sensor to the floor of the footwell, - an interior microphone located near the brake pedal to record noises generated by foreign objects coming into contact with the brake pedal, - a seat mat or a force sensor in a seat cushion for determining a force transfer from the seat cushion to the backrest of the driver's vehicle seat when the brake pedal is pressed, - another interior camera to detect a blocked brake pedal based on the driver's facial expression, - another interior microphone to detect the blocked brake pedal by the driver's voice, - a force sensor on the steering wheel to detect the blocked brake pedal due to the driver gripping the steering wheel more tightly, - a pulse sensor to detect the blocked brake pedal by an increase in the driver's pulse, - at least one environment detection sensor for detecting the blocked brake pedal by reducing the distance of the vehicle to other vehicles or obstacles, wherein the control unit has an artificial intelligence which is trained once in a test vehicle based on the sensor signals determined during a test drive, whereby a predetermined number of tuples are generated, each of which comprises an explanatory feature formed by measurement data from the sensors and an explained feature in the form of the blocking state of the brake pedal prevailing at the time of the measurement, wherein all generated tuples are divided into a training set and a test set, the sets of which do not overlap, with which a mapping function for the prediction of the blocking states of the brake pedal is generated by means of one or more methods of supervised learning, wherein the prediction method to be used by the artificial intelligence model is determined on the basis of a degree of deviation of a predicted unknown tuple from the test set, which is implemented in the control unit of the production vehicle, in which the sensor data transmitted from the sensors to the control unit is processed in such a way that it is used by the implemented artificial intelligence model to predict the brake pedal state for adapting the brake pedal voltage for detecting the driver's braking request even when the brake pedal travel is partially blocked.

[0008] An artificial intelligence model is defined as a software function that can make classifications and value predictions based on input information. This mapping function is trained using machine learning methods.

[0009] A brake pedal tension is understood to mean an adjustment of the braking deceleration power over the usable brake pedal travel that can be controlled by the driver.

[0010] A brake pedal blockage describes a restriction in the movement of the brake pedal caused by a foreign object or mechanical component defects.

[0011] The evaluation of the multitude of sensor signals in an artificial intelligence model allows for the reliable detection of an existing brake pedal blockage as well as the prediction of an impending blockage. This not only uses signals from a brake pedal sensor, but also incorporates signals from sensors that monitor the brake pedal environment, the driver, and the vehicle's surroundings. This increases the accuracy of the detection. In addition to information about the brake pedal blockage, the detection function of the artificial intelligence model also provides information about the position at which the brake pedal is blocked.

[0012] The high sensitivity of the brake pedal blockage detection allows objects of different sizes and material properties to be detected.

[0013] The process for detecting brake pedal blockages is divided into the individual steps: generating training data, training the artificial intelligence model, and applying the model in the vehicle.

[0014] Information about the blocking position is used to dynamically adjust the brake pedal tension so that the driver can request the full deceleration power of the brake despite the limited brake pedal travel. If the brake pedal's zero position is blocked, the brake pedal tension is adjusted so that the vehicle can no longer be braked unintentionally. Adjusting the brake pedal tension allows the driver to continue to control the deceleration power within the remaining brake pedal travel.

[0015] In one embodiment, training data for the artificial intelligence model is generated using a simulation process that combines the physical relationships between the driver's behavior, the behavior of the brake pedal, and the properties of the foreign object into a software function. The simulation recreates a real driving maneuver of the vehicle and generates data from virtual sensors. This allows more data to be provided for training the artificial intelligence model in the shortest possible time and with fewer resources.

[0016] In a further embodiment, a method of linear and polynomial regression and / or nearest neighbor classification and / or neural networks can be used to train the artificial intelligence model. Using all of these methods, a mapping function for predicting the locking states of the brake pedal under The method, which calculates the best possible approximation to an ideal mapping function, is determined using metrics that calculate the degree of deviation of the respective artificial intelligence model when predicting unknown tuples from the test block of sensor data.

[0017] In a further embodiment, the control unit is designed to cyclically calculate the brake pedal state at a high frequency. This allows a change in the brake pedal state to be detected very quickly and a corresponding substitute reaction to be triggered.

[0018] In a further embodiment, the control unit classifies the brake pedal and distinguishes between "brake pedal locked" and "brake pedal not locked." If the brake pedal is classified as locked, a further distinction is made as to whether the lock is at the end stop or at the zero position of the brake pedal. An end stop is detected if, for example, an object is behind the brake pedal. A zero position is assumed if the object is on the brake pedal. Regardless of the type of lock, the deceleration power is adjusted to the remaining free travel of the brake pedal.

[0019] In a further embodiment, a transition between different brake pedal tensions occurs abruptly upon release of the brake pedal or in a time-based ramp with continuous brake pedal application. This allows the driver to manage transitions to a different brake pedal tension.

[0020] In a further embodiment, the control unit is connected to a warning device that issues a warning when a critical brake pedal condition is detected. The driver is thus warned in advance, even before a dangerous driving situation occurs. They can park the vehicle in a safe location in a timely manner to remove any object near the pedal.

[0021] Further advantages, features, and details will become apparent from the following description, which describes at least one exemplary embodiment in detail. Described features may form the subject matter of the invention alone or in any meaningful combination, possibly even independently of the claims, and may, in particular, also be the subject matter of one or more separate applications.

[0022] Showing: Fig. 1 a schematic diagram of the system according to the invention (sensor + actuator), Fig. 2 an embodiment for generating training data for the artificial intelligence model, Fig. 3 an example of training the artificial intelligence model, Fig. 4 an embodiment of the application of the trained artificial intelligence model in the vehicle, Fig.5 examples of how to adjust the brake pedal clamping.

[0023] In Fig. Figure 1 shows a schematic diagram of the system according to the invention, which is arranged in a vehicle. The vehicle 1 comprises a plurality of sensors, only a few of which are shown as examples in the drawing. A brake pedal sensor 5 is arranged on a brake pedal 3, and a footwell sensor 7 is arranged near the brake pedal 3. Table 1 contains a detailed list of sensors suitable for detecting a brake pedal lock.

[0024] These sensors 5, 7, as well as a driver monitoring sensor 9 and a vehicle environment sensor 11, are connected to a central control unit 13 of the vehicle 1, which includes an artificial intelligence model 15 that evaluates the sensor signals for a blockage of the brake pedal 3. Depending on the results provided by the artificial intelligence model 15, the central control unit 13 controls a braking device 17 of the vehicle 1. Advantageously, the sensors 5, 7, 9, 11 and the braking device 17 communicate with each other via a data bus 19 of the vehicle 1. Table 1 Vehicle sector sensor Measured quantity Reference to the blockade brake pedal Brake pedal travel sensor Brake pedal travel -absolute brake pedal travel remains permanently below maximum value or above minimum value during the driving cycle, -occurrence of characteristic brake pedal travel gradients, -occurrence of characteristic actuation frequencies (panic repeated pressing of the brake pedal Force sensor on brake pedal plate Force acting on the brake pedal plate -Brake pedal force no longer proportional to the brake pedal travel Accelerator pedal travel sensor Accelerator pedal travel If the accelerator pedal is inadvertently depressed while the brake pedal is depressed in a panic situation, -Accelerator pedal travel >0%, -Occurrence of characteristic accelerator pedal gradients Footwell Ultrasonic sensor Distance to objects in the footwell Ultrasonic sensor detects foreign objects behind or next to the brake pedal camera Object identification on a video image of the footwell The image shows foreign objects behind and next to the brake pedal Force sensor on the footwell floor behind the brake pedal Force acting on the footwell floor If there is a foreign object behind the brake pedal, it transfers the brake pedal force to the floor of the footwell Interior microphone Noises in the footwell When the brake pedal comes into contact with the foreign body, a characteristic noise is produced driver Seat mat / force sensor in seat cushion Force acting on the seat cushion When the driver presses the brake pedal hard, he leans into the backrest and the seat cushion is relieved Interior camera Recognition of the driver's facial expressions on the driver's video image The driver's facial expression can indicate when the driver is frightened or panicking because of a blocked brake pedal Interior microphone Noises in the interior, driver's comments Characteristic noises / exclamations from the driver, such as "Ah!", "Oh my God!", which indicate a technical defect Force sensor on the steering wheel Force with which the steering wheel is gripped When the driver panics due to a blocked brake pedal, he grips the steering wheel with greater force Heart rate sensor / fitness bracelet Driver's pulse If the driver panics because of a blocked brake pedal, his pulse increases rapidly Vehicle environment Environment detection with radar sensor, camera, LIDAR sensor, ultrasonic sensor Distance to other vehicles and obstacles, e.g. as output signal of a passive adaptive cruise control If the brake pedal is blocked, the brake pedal travel may be smaller than expected in the respective driving situation in order not to undercut the distance to other vehicles and obstacles

[0025] To detect the blockage, the artificial intelligence model 15 must be trained using various machine learning methods before being deployed in vehicle 1. The training is divided into the steps "generating training data," "training the model," and "applying the model."

[0026] To train the artificial intelligence model 15, the method of supervised learning is used. First, a set of tuples 21 is generated, as shown in Fig.2. Each tuple 21 consists of an explanatory feature 21a (feature) and an explained feature 21b (label). The measurement data from sensors 5, 7, 9, 11 serve as features 21a, for which the blocking state of the brake pedal 3 prevailing at the time of measurement is stored as a label 21b in a tuple 21. To generate representative tuples 21, vehicle tests are conducted. Typical blocking states of the brake pedal 3 are deliberately induced by successively positioning various objects in the footwell so that a blockage of the brake pedal 3 occurs. At the same time, typical driving maneuvers are performed with the prepared vehicle on a closed test track, and measurement data are recorded. The identical driving maneuvers are also performed and measured with a freely moving brake pedal 3.To best represent typical real-life situations, both the driving maneuvers and the properties of the foreign objects are evenly distributed within defined limits. Table 2 shows possible foreign objects that could block the brake pedal 3. Table 2 Foreign body Soft material Medium-hard material Hard material Small flat shape doormat Pack of tissues Wallet, mobile phone Large, thick shape Safety vest, cuddly toy Flip-flops. Empty soda bottle. Ice scraper, hand brush, full drink bottle, umbrella

[0027] To increase the efficiency of data generation, simulation methods are also used in which the relevant physical relationships between the driver 23, the behavior of the brake pedal 3, and the properties of the foreign object are modeled in a software function. The simulation is then used to recreate real driving maneuvers in the test vehicle and generate data from virtual sensors.

[0028] According to Fig.3a, the set of all tuples 21 is divided into a training set 25 and a test set 27. The two sets must be disjoint so that an objective evaluation of the performance of the artificial intelligence model 15 can be performed using the test set 27.

[0029] If training data is inadvertently included in the test set 27, the model's performance may be overestimated. For example, the artificial intelligence model 15 may not have captured the general relationship between the sensor data and the brake pedal lock, but instead only correctly predict the learned training data and thus only predict known states.

[0030] The actual performance of the artificial intelligence model 15 could be worse than expected with unknown sensor data, e.g., with deviating driving maneuvers or with new geometries or material properties of the foreign objects. This behavior of the artificial intelligence model 15 is referred to as "overfitting" and can be prevented by correctly dividing the training and test sets 25, 27.

[0031] In Fig.Figure 3 shows an embodiment for training the artificial intelligence model 15. Several supervised learning methods are available to generate a suitable mapping function for predicting the locking states of the brake pedal 3. These include, for example, linear and polynomial regression, "nearest neighbor classification," or artificial neural networks (ANN). These methods have in common that they are based on the training data of the training set 25 and mathematical optimization methods. calculate the best possible approximation of the ideal mapping function ( Fig. 3b).

[0032] To determine the most efficient of the selected methods for the present detection of the locking state of the brake pedal 3 and the sensors 5, 7, 9, 11 to be selected for this purpose, all methods to be tested are examined and validated with different configurations (hyperparameters of the method). Metrics that calculate the degree of deviation of the artificial intelligence model 15 in predicting unknown tuples 29 from the test set 27 serve as an objective evaluation criterion ( Fig. 3c).

[0033] The previously once and offline trained Artificial Intelligence model 15 is, after successful validation, Fig.3c is implemented as a software function on the central control unit 13 of the vehicle 1. Alternatively, this software function can also be implemented on a decentralized control unit, e.g., the control unit 33 of the brake system 17, which also implements the brake pedal voltage in order to minimize the reaction times when adjusting the brake pedal voltage. The control unit 33 of the brake system 17 receives all sensor information from a drive control unit 35, the interior control unit 37, and driver assistance control units 39 via the data bus 19. According to Fig. 4, the driver assistance control unit 39 is connected to the vehicle environment sensor 9, while the interior control unit 37 is coupled to the driver monitoring sensors 7. The drive control unit 35 is connected to an accelerator pedal sensor, and the control unit 33 of the braking system 17 is connected to the brake pedal sensor 3.

[0034] This sensor data is processed in the control units 33, 35, 37, and 39 so that it can be used as features 21a for predicting the brake pedal state by the artificial intelligence model 15, which is implemented in the control unit 33 of the braking system 17. The brake pedal state is calculated cyclically at a high frequency.

[0035] Two examples for the adjustment of the brake pedal tension by shifting the brake pedal characteristic curve, which represents the deceleration performance of the braking system over the brake pedal travel, depending on the detected brake pedal state are shown in Fig. 5. The result of the artificial intelligence model 15 classifies the brake pedal state, for example, into "brake pedal is released" and "brake pedal is locked." In the case of an unlocked brake pedal 3, the deceleration performance can be adjusted by the driver 23 based on the existing brake pedal characteristic curve 41.

[0036] If the brake pedal 3 is classified as blocked, two blockage scenarios can be predicted. In Fig. 5a, a blockage position is detected in which an object, e.g. a drinking bottle, a flip-flop sandal, a toy, etc., jams the brake pedal 3 from behind. In this case, the artificial intelligence model predicts a brake pedal value of 80%, beyond which brake pedal 3 can no longer be pressed. To still achieve maximum deceleration (braking) performance, the existing brake pedal characteristic curve 41 is adjusted. Since the prediction is imprecise and the object can continue to move, the maximum deceleration would be advanced to a brake pedal value of 50% (adjusted brake pedal characteristic curve 43) and spanned over the resulting shortened brake pedal travel with a residual controllability.

[0037] In another blockage scenario, which can be predicted by the artificial intelligence model 15, an object appears to prevent the complete release of the brake pedal 3. This can occur, for example, if a floor mat rests on the brake pedal 3. To prevent a constant and unintentional demand for a slight deceleration force from the braking system, the minimum deceleration force is shifted to the predicted brake pedal value, which reflects the contact of the floor mat with the brake pedal 3, for example, to 20%. From this point, the brake pedal characteristic curve 43 is redrawn up to the maximum deceleration force.

[0038] In order for the transition to the changed brake pedal tension to remain manageable for the driver 17, this change must occur abruptly ( Fig. 5a) or with continuous pedal operation ( Fig. 5b) with a time ramp.

[0039] As soon as the artificial intelligence model 15 has detected a blocked brake pedal, the driver 23 is informed directly via a display device 47 ( Fig. 1) and requested to drive to a safe location promptly and remove any object near the brake pedal 3. As additional measures, other road users are informed of a potentially dangerous situation via a Car2Car interface and the hazard warning lights of vehicle 1 are activated.

[0040] After the object has been removed, the driver is prompted via the display unit 47 to fully depress the brake pedal 3 from 0% to 100% three times in succession. Once the brake pedal is free again, the brake pedal tension returns to its initial state (brake pedal characteristic curve 41) and the warning message disappears. As an additional safety measure, an entry is permanently stored in the error log of the brake control unit 33 so that specialist personnel can check the brake pedal 3 for damage during a subsequent workshop visit. QUOTES CONTAINED IN THE DESCRIPTION

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

[0000] DE 10 2008 029 311 A1

[0002]

Claims

[1] System for controlling a braking behavior of a vehicle, comprising a central or decentralized control unit (13, 33) which is connected to a plurality of vehicle sensors (5, 7, 9, 11), of which a sensor (5) for determining a position or movement of a brake pedal (3) is coupled thereto, characterized by that the control unit (13, 33) is designed to detect a currently occurring blockage of the brake pedal (3) or to predict an impending blockage with - a force sensor arranged on a brake pedal plate to check the proportionality of the actuating force to the brake pedal travel, - an accelerator pedal travel sensor for detecting unintentional actuation, - an ultrasonic sensor arranged in the footwell of the vehicle (1) for detecting foreign bodies behind or next to the brake pedal (3) via distance detection, - a camera arranged in the footwell of the vehicle (1) for detecting foreign objects behind or next to the brake pedal (3) by identifying objects on an image of the footwell, - a force sensor arranged on the floor of the vehicle (1) behind the brake pedal (3) for detecting foreign bodies behind the brake pedal (3) by means of the brake pedal force transmitted by the latter to the floor of the footwell, - an interior microphone arranged near the brake pedal (3) for recording noises generated by foreign bodies coming into contact with the brake pedal (3), - a seat mat or a force sensor in a seat cushion for determining a force transfer from the seat cushion to the backrest of the driver's vehicle seat when the brake pedal is operated (3), - another interior camera for detecting a blocked brake pedal (3) based on the driver's facial expression, - another interior microphone for detecting the blocked brake pedal (3) by the driver's voice, - a force sensor on the steering wheel to detect the blocked brake pedal (3) due to the driver gripping the steering wheel more tightly. - a pulse sensor for detecting the blocked brake pedal (3) by an increase in the driver's pulse, - at least one environment detection sensor (11) for detecting the blocked brake pedal (3) by reducing the distance of the vehicle (1) to other vehicles or obstacles wherein the control unit (13, 33) has an artificial intelligence (15) which is trained once in a test vehicle using the sensor signals determined during a test drive, wherein a predetermined number of tuples (21) are generated, each comprising an explanatory feature (21a) formed by measurement data from the sensors (5, 7, 9, 11) and an explained feature (21b) in the form of the blocking state of the brake pedal (3) prevailing at the time of the measurement, wherein all generated tuples (21) are divided into a training set (25) and a test set (27), the sets of which do not overlap, with which a mapping function for predicting the blocking states of the brake pedal (3) is generated by means of one or more methods of supervised learning, wherein the prediction method to be applied by the artificial intelligence model (15) is determined based on a degree of deviation of a predicted unknown tuple (29) from the test set (27), which is implemented in the control unit (13, 33) of a production vehicle (1), in which the sensor data transmitted from the sensors (5, 7, 9, 11) to the control unit (13, 33) are processed in such a way that they are used by the implemented artificial intelligence model (15) to predict the brake pedal state for adapting the brake pedal voltage for detecting the driver's braking request, even when the brake pedal travel is partially blocked. become. [2] Vehicle system according to claim 1, characterized by that the control unit of the test vehicle is designed in such a way that, for training the artificial intelligence model (15), training data is generated using a simulation method which summarizes the physical relationships between the driver's behavior, the behavior of the brake pedal (3) and the properties of the foreign body in a software function by simulating a real driving maneuver of the test vehicle with the simulation and measuring data from virtual sensors. [3] Vehicle system according to claim 1 or 2, characterized by that a method of linear and polynomial regression and / or nearest neighbor classification and / or neural networks can be used to train the artificial intelligence model (15). [4] Vehicle system according to claim 1, 2 or 3, characterized bythat the control unit (13, 33) of the series vehicle (1) is designed to calculate the brake pedal state cyclically at high frequency. [5] Vehicle system according to at least one of the preceding claims, characterized by that the control unit (13, 33) classifies the brake pedal (3) and distinguishes between "brake pedal blocked" and "brake pedal not blocked" during the classification, wherein if the brake pedal (3) has been classified as blocked, a distinction is again made as to whether the blockage is at the end stop or at the zero position of the brake pedal (3). [6] Vehicle system according to claim 5, characterized by that a transition between different brake pedal voltages occurs abruptly when the brake pedal (3) is released and in a time ramp when the brake pedal is continuously actuated. [7] Vehicle system according to at least one of the preceding claims, characterized bythat the control unit (13) is connected to a warning device (47) for issuing a warning when a critical condition of the brake pedal (3) is detected.

Images