Method for operating a vehicle

Abstract

The invention relates to a method for operating a vehicle (1), wherein an in-vehicle computing unit calculates a safe state space using a vehicle dynamics model, comprising a tire model, wherein the vehicle (1) is controllably steerable for each point in the safe state space, the computing unit determines a movement trajectory (2) for the vehicle (1) and controls at least one actuator of the vehicle (1) in order to control the vehicle (1) within the safe state space. The method according to the invention is characterized in that a) the computing unit calculates a permissible driving corridor (3) for the vehicle (1), wherein the computing unit determines, in order to determine a boundary line (4) of the permissible driving corridor (3), which locations the vehicle (1) has reached at the current time increment (t i) under a control for a prediction time window (5) according to the driving dynamics parameters limiting the safe state space; b) the computing unit compares the movement trajectory (2) with the permissible travel corridor (3) and determines whether for the current time increment (t i ) the movement trajectory (2) leaves the permissible travel corridor (3); and if so: c) the computing unit controls the at least one actuator for manipulating the current driving dynamics parameters so that the movement trajectory (2) for the prediction time window (5) runs within the permissible driving corridor (3).

Description

The invention relates to a method for operating a vehicle of the type defined in more detail in the preamble of claim 1.As digitalization increases, assistance systems are also increasingly integrated into vehicles. By providing such assistance systems, the comfort and safety during use of the vehicle can be improved. Corresponding driver assistance systems, such as ABS, ESP, an emergency braking assistant, an avoidance assistant and the like, are normally reactive. This means that the vehicle detects the driving situation by means of a sensor system and responds to respective excursions in the sensor values, for example by actuating an actuator. The vehicle can, for example, detect its environment by means of a surroundings sensor system and examine it for the presence of static and dynamic surroundings objects. Depending on the current driving situation, i.e., in particular taking into account the movement trajectory of the vehicle and the relative position and movement direction of the respective objects with respect to the host vehicle, the vehicle is then able to estimate the risk for a collision with corresponding surrounding objects and to react to this. In response, an emergency braking maneuver or avoidance maneuver may be initiated, for example.It is desirable to be able to provide a driver assistance system which, even in the case of manual control of the vehicle by a person driving the vehicle, is capable of maintaining a safe driving state of the vehicle. This means in particular that the vehicle should not break out, slide or tip over or the controllability should still be given. The earlier the vehicle changes its driving state, the more likely it is also that a hazardous situation can be avoided.DE 10 2019 205 405 A1 discloses the determination of an input variable of a vehicle actuator by means of model-assisted predictive control. The publication describes the actuation of an actuator of an at least partially automatedly controlled vehicle by carrying out a model-based predictive control algorithm for following a predefined trajectory. As a secondary condition, a range of safe or permitted system states is defined in the model predictive control algorithm, which must be observed by the vehicle for following the trajectory. This range of safe system states describes a state space of tolerated driving dynamics model parameters of a driving dynamics model on which the model-predictive control approach is based. Furthermore, the publication describes the estimation of corresponding driving dynamics model parameters. The operating method known from the publication can be applied here exclusively to automatically controlled vehicles for which a following movement trajectory has been predefined.Furthermore, DE 10 2019 006 933 A1 discloses a technique for adapting the model parameter of a dynamic model for the lateral and longitudinal guidance of a motor vehicle. An actual value of a state of motion of the transverse guide and / or longitudinal guide of the motor vehicle is determined by means of a sensor. The actual value of the state of motion is compared with a corresponding setpoint value specification. A corresponding model parameter of the dynamic model is adjusted taking this deviation into account in order to minimize the deviation. To determine the suitable model parameter, artificial intelligence in the form of an artificial neural network can also be used. As input variables, the artificial neural network can receive a lateral acceleration, a steering angle, a steering wheel angle, a yaw rate and / or a curvature of the trajectory of the motor vehicle. As output variable, for example, it is possible to determine: a cornering stiffness at the front and rear axles, at least one center-of-gravity distance of the front and rear axles, a mass of the motor vehicle, tire parameters, a rolling resistance and / or a moment of inertia about a vertical axis of the motor vehicle.In addition, DE 10 2016 011 244 A1 discloses a method for operating a vehicle. The publication describes a determination of a setpoint trajectory for an automatically controlled vehicle taking into account a difference between the current vehicle position and a roadway marking. As soon as no lane marking can be detected any longer, the position of a vehicle driving ahead can instead be taken into account for the difference formation. If it is simultaneously possible to detect the position of vehicles driving ahead and the roadway marking, a confidence value can be determined for each vehicle driving ahead, describing a fidelity with which the respective vehicle follows the own lane. In the absence of a roadway marking, the setpoint trajectory is then preferably determined with the highest confidence value taking into account the position of the vehicle driving ahead.The object of the present invention is to specify an improved method for operating a vehicle, with the aid of which, in particular also a manually controlled vehicle, can be operated particularly safely.According to the invention, this object is achieved by a method for operating a vehicle having the features of claim 1. Advantageous embodiments and further developments are evident from the claims dependent thereon.In a method of the generic type for operating a vehicle, wherein an in-vehicle computing unit calculates a safe state space using a driving dynamics model comprising a tire model, wherein the vehicle can be controlled in a controlled manner for each point in the safe state space, the computing unit determines a movement trajectory for the vehicle and controls at least one actuator of the vehicle in order to control the vehicle within the safe state space, it is provided according to the invention that a) the computing unit calculates an admissible driving corridor for the vehicle, wherein the computing unit determines, in order to determine a boundary line of the admissible driving corridor, which locations the vehicle can just still reach at the current time increment under control for a prediction time window according to the driving dynamics parameters limiting the safe state space; b) the arithmetic unit matches the movement trajectory with the permissible driving corridor and determines whether the movement trajectory leaves the permissible driving corridor for the current time increment; and if so: c) the arithmetic unit controls the at least one actuator for manipulating the current driving dynamics parameters, so that the movement trajectory for the prediction time window runs within the permissible driving corridor.The permissible driving corridor thus describes a local space surrounding the vehicle at least in sections or adjoining the vehicle, of positions at which the vehicle can be controlled safely and reliably. If the vehicle were to leave the permissible driving corridor, the risk of a loss of safe controllability would be present. This means, for example, that at least one tire of the vehicle spins, the front and / or rear axle breaks out, the vehicle tilts over and the like. In order to be able to evaluate this, corresponding limit values for the safe state space are established, such as a maximum lateral acceleration, a maximum tire slip value and the like. The permissible travel corridor is locally limited by the boundary line. The boundary line thus represents the perimeter of the permissible driving corridor.In order to determine the boundary line for the current time increment, the arithmetic unit varies the respective driving dynamics parameters of the driving dynamics model used for simulative prediction of the vehicle state and checks whether the respective vehicle states determined by the driving dynamics model permit reliable control or whether a loss of control is imminent. The driving dynamics parameters describe those variables which affect the state of motion of the vehicle, such as in particular the speed of motion of the vehicle, an acceleration (positive or negative) acting on the vehicle, a steering angle, a jerk and the like. If at least one vehicle state reveals that the safe operating mode of the vehicle is lost, a point is present for this location for this set of driving dynamics parameters on or outside the boundary line of the permissible driving corridor.In particular, when driving straight on a straight line, it may also be possible for the vehicle to remain controllable in a controlled manner for all driving dynamics parameters that can actually be reached / adjusted from the current time increment within the prediction time window under consideration. In this case, the course of the boundary line ahead of the vehicle in the direction of travel (or, depending on the situation, optionally also at least in sections of the boundary line running laterally next to the vehicle) is obtained from the time duration of the prediction time window taking into account the current travel speed. The prediction window may be, for example, three seconds in length.The driving dynamics model parameters on which the driving dynamics model is based for this calculation can be permanently predefined and / or can be estimated by the vehicle. The driving dynamics model parameters include all influencing variables which affect the calculation of the vehicle state, such as, for example, the vehicle mass, a position of the center of gravity, a mass moment of inertia, a coefficient of friction between a tire and the roadway, an intrinsic steering gradient, a float angle gradient and the like. The driving dynamics model parameters have an effect on the driving dynamics parameters that are established. If, for example, the vehicle mass is greater, the vehicle takes a longer period of time at a predefined drive power in order to accelerate to a desired speed.In order to estimate driving dynamics model parameters, information currently collected by the vehicle by means of a sensor system can be taken into account, in particular. Driving dynamics model parameters can thus be measured directly or also estimated indirectly. For example, a model parameter of the tire model included in the driving dynamics model can be adapted as a function of a currently measured tire slip value or a wheel speed. Furthermore, by means of a camera-based or acoustic roadway monitoring, moisture on the roadway can be detected, for example. This can be used accordingly to change driving dynamics model parameters, such as reducing an adhesion coefficient to the roadway for the route section lying ahead. Corresponding information can also be obtained from a source external to the vehicle, such as, for example, via the Internet or via Car2Car or Car2X information. Information can thus be retrieved from a weather service. To evaluate the current driving situation, the vehicle can use a wide variety of sensors, such as, in particular, the on-board sensor system, the surroundings sensor system or even a map-related sensor system. The on-board sensor system includes, for example, acceleration sensors, yaw rate sensors, a wheel speed sensor, and the like. The environment sensor system includes, for example, cameras, radar systems, LiDAR systems, ultrasonic sensors, acoustic sensors and the like. The map sensor system includes, in particular, map information and a vehicle position determined by means of a global navigation satellite system.If the movement trajectory of the vehicle leaves the permissible driving corridor, this would mean that the vehicle would leave the safe driving state after passing over the boundary line. Thus, for example, said wheel spin threatens, vehicle axle break-out or even vehicle rollover. In order to proactively prevent this, the computing unit intervenes in the operation of the vehicle and correspondingly controls a respective actuator. The actuator could be, for example, a wheel-specific brake, a servomotor for adjusting the steering angle or a power requirement for the one to more drive engines of the vehicle in order to accelerate the vehicle even further. As a further actuator, a differential or locking differential could likewise be actuated. By means of such a locking differential, the drive torque output by the drive engine of the vehicle can be divided in a targeted manner into the individual wheels of the vehicle. This makes it possible to more reliably maintain the adhesion of the respective tires to the road surface. If the vehicle is a purely battery-electrically driven vehicle, then a separate electric motor could be provided for each wheel of the vehicle or at least each axle of the vehicle, so that a setpoint torque for the respective wheels can likewise be predefined by means of power control of the respective electric motors. With the aid of the method according to the invention, manually controlled vehicles, at least partially automatically controlled vehicles and also autonomously controlled vehicles can be operated equally safely.Whether or not the method according to the invention is to be carried out can be manually defined in the respective vehicle. Thus, a vehicle occupant or the vehicle-guiding person can switch on and off a corresponding driving mode by actuating a corresponding operating element. Carrying out the method according to the invention by the vehicle can also be described with the provision of a special safety driver assistant system.In order to cause the movement trajectory for the prediction time window to run within the permissible travel corridor in step c), two different solution strategies are generally possible. Thus, an advantageous development of the method according to the invention provides that the computing unit actuates the at least one actuator in step c) for reducing the current driving dynamics parameters in order to increase the area of the permissible driving corridor. According to a further advantageous embodiment of the method according to the invention, it would also be possible to actuate the at least one actuator in order to place the course of the movement trajectory within the permissible driving corridor. Of course, both approaches may be combined. Alternatively or additionally to enlarging the area of the permissible driving corridor, the shape of the area of the permissible driving corridor can also be adapted, so that the movement trajectory again runs within the permissible driving corridor.In particular, by reducing the acceleration or the speed of the vehicle, it is possible to achieve an increase in the area of the permissible driving corridor. If a specific steering angle is set, the vehicle threatens to break down from exceeding a specific speed. The smaller the steering angle, the greater the just-acceptable speed of travel. Thus, if the vehicle brakes, greater steering angles can be adjusted without risking loss of control over the vehicle. Accordingly, the area of the allowable travel corridor increases. However, it should be noted that a tire has physical limits; in particular, high longitudinal forces are caused by strong braking, for example, and high transverse forces are caused by strong steering, for example, mutually. In this case, in particular, the lateral limits of the permissible driving corridor can become greater (since higher steering angles can be tolerated) and the limit lying ahead of the vehicle can approach the vehicle (since, at low speed and with the prediction time window remaining the path travelled by the vehicle decreases), as a result of which the shape of the permissible driving corridor is distorted. By skillfully distributing the torque output by the drive unit of the vehicle to the individual wheels of the vehicle, the area of the permissible travel corridor can likewise be distorted. In addition to a corresponding setpoint value specification for the respective actuator by the computing unit, alternatively the space of the characteristic diagram of a respective actuator that can be selected manually by the user can also be limited. Thus, for example, when the accelerator pedal is fully depressed, only a lower drive power can be called up.It is likewise possible to adapt the course of the movement trajectory of the vehicle, such that the vehicle does not leave the permissible driving corridor. This is likewise possible by actuating the respective actuators. The vehicle can thus be accelerated or braked in a targeted manner and / or a different steering angle can be set than originally planned / estimated. The changing of the course of the movement trajectory and the adaptation of the area of the permissible driving corridor can thereby blend smoothly into one another.The computing unit determines concrete value specifications or tolerable value ranges for a respective time horizon ahead, which a respective actuator is / is intended to assume. In other words, a setpoint value trajectory for a respective actuator or an admissible restricted characteristic field range is determined. The time horizon ahead can be of the same size, smaller or even larger than the prediction time window. The time horizon ahead can also be of the same size, smaller or even greater than a time interval between two successive time increments (see below).A further advantageous embodiment of the method according to the invention further provides that the computing unit repeats steps a) to c) at a future time increment following the current time increment, in particular while maintaining a fixed repetition frequency. In general, the method according to the invention can only be carried out once at a specific point in time. This time is referred to as a (current) time increment. However, it is particularly advantageous if steps a) to c) are carried out repeatedly. Thus, the safe operation of the vehicle during use thereof can be maintained. In this case, the method steps a) to c) are carried out again at a plurality of successive time increments. For example, time increments may follow one another with a fixed time duration, for example every 20 milliseconds. The time interval between two successive time increments can thus be smaller than the prediction time window under consideration. However, the time interval can also be exactly the same as the prediction time window or greater. The time interval between successive time increments can also vary, so that, for example, a second time increment follows a first time increment after 50 ms, but a third time increment follows the second time increment only after 180 ms. The more frequently time increments follow one another, i.e. the method steps a) to c) are carried out, the more seamless the operation of the vehicle can also be ensured. The repetition frequency can be selected to be correspondingly of any desired magnitude and can be, for example, 0.2 Hz, 1 Hz, 10 Hz or else fractions or multiples thereof.The instant at which method steps a) to c) are carried out, i.e. when a respective time increment is present, can also be linked to the arrival of an event. Thus, the vehicle can monitor different variables or parameters and provide a time increment exactly when one or more of these parameters reach a defined value or value range. For example, the speed of travel of the vehicle could be used as a reference parameter. For example, the execution of method steps a) to c) is linked to the fact that the vehicle must travel rapidly at least 30 km / h. Thus, the accident risk decreases accordingly, or, should an accident occur, the potential for damage, so that it is possible to dispense with carrying out the method according to the invention at low driving speeds. The parameter considered could also be an odometry parameter, so that, for example, a time increment is provided again after a specific travel path travelled by the vehicle. For example, method steps a) to c) could be carried out every three meters. Further parameters triggering the execution of the method steps according to the present invention could be, for example: a specific tire slip value, a wheel speed, an acceleration acting on the vehicle in a specific spatial direction or a deviation of a measured driving dynamics parameter from a driving dynamics parameter calculated by the driving dynamics model, and the like.A further advantageous embodiment of the method according to the invention further provides thatwhen the vehicle is manually controlled, the computing unit estimates an estimation trajectory dependent on the control behavior of a vehicle-guiding person and uses the estimation trajectory in step b) for the movement trajectory; orwhen the vehicle is controlled at least partially automatically, the computing unit uses a setpoint trajectory provided for guiding the vehicle in step b) for the movement trajectory.The movement trajectory taken into account in step b) therefore does not necessarily have to be a setpoint trajectory predefined for the vehicle by a computer system. An estimation trajectory can thus be ascertained on the basis of the driving behavior of a vehicle-guiding person. For this purpose, for example, the currently set steering wheel angle or steering angle can be linked to the current speed of movement of the vehicle. Taking into account the rest of the sensor system comprised by the vehicle, surrounding objects in the environment of the vehicle can also be detected. The computing unit can thus estimate whether an object is blocking the current lane, whereupon the vehicle-guiding person is likely to steer around the corresponding object. The computing unit can then estimate such an avoidance trajectory for the person driving the vehicle. The computing unit can determine a plurality of possible avoidance trajectories and assign a probability value to each avoidance trajectory, expressing a probability with which the vehicle-guiding person will choose precisely this avoidance trajectory. That avoidance trajectory with the highest probability can then be taken into account as the movement trajectory in step b).The setpoint trajectory, on the other hand, can be predefined for the vehicle by any desired computing unit, such as the control unit of a vehicle subsystem. Such a control device can be used to form a driver assistance system. The setpoint trajectory can therefore result in connection with the provision of the functions of a driver assistance system. For example, it is a target trajectory for following a predetermined navigation route or an avoidance route for avoiding a collision.The advantages of the method according to the invention are especially evident when the movement trajectory is determined in agreement with the estimation trajectory. This is because, by executing the method according to the invention, it is possible to provide a driver assistance system which improves the safety of the vehicle and which predictively avoids critical driving states instead of merely a reactive prevention. The driving behavior can thus be adapted early to potential dangerous situations, which allows even more reliable control of the vehicle, even when driving manually.A further advantageous embodiment of the method according to the invention further provides that the computing unit for determining the movement trajectory determines a route profile of the roadway being traveled by the vehicle and a position of the vehicle relative to the roadway and plans the movement trajectory such that a fixed relation of the vehicle position to the route profile is observed. In this case, the computing unit can both read the route profile from a digital road map and determine the route profile from sensor data which have been generated by at least one environment sensor and / or at least one vehicle state sensor. Both variants are suitable both for the case in which the movement trajectory is an estimation trajectory or said setpoint trajectory.The term "drivable roadway" may be more generally formulated as follows. A drivable surface can also be referred to as drivable subgrade. A drivable subgrade is not finally limited to any type of fixed, flat roads and subgrades. These can be detected directly by environmental sensors such as cameras, radar, lidar or can be derived by other vehicles detected by the environmental sensors that are travelling or driven on the ground. In contrast, non-drivable or conditionally drivable subgrades may be identified, such as gravel beds, subgrades, ditches, populated or grown or populated terrain, etc. Additionally, stationary and moving objects may be detected by environmental sensors. Based on the detected objects, a currently drivable roadway or a future drivable roadway or drivable subsoil can be determined with the aid of the prediction of the movement paths of the detected movable objects.The defined relation of the vehicle position to the route profile can be, for example, central. Accordingly, the vehicle is to travel centrally in the respective lane or parallel to the roadway boundary. A current relation selected by the vehicle-guiding person could also be detected and this offset of the vehicle to the driving lane could be maintained. The digital road map can be stored, for example, in a navigation unit such as a navigation system. Different digital road maps may be maintained for different regions of this earth. The position of the vehicle relative to the roadway can be determined using a navigation satellite-assisted position determination unit, for example, by GPS.With the aid of the surrounding sensors of the vehicle, it may likewise be possible to record the route profile of the road or roadway lying ahead of the vehicle and the relationship of the vehicle to the respective roadway. For example, the route profile can be derived from correspondingly generated camera images. Proven image recognition algorithms can be used for this purpose.The route profile can also be derived from sensor data generated by vehicle state sensors. Such a vehicle state sensor may be, for example, a wheel speed sensor, a sensor for detecting the steering wheel position, a speed sensor, an acceleration sensor and the like. As already mentioned, it is possible, taking into account the current speed of movement of the vehicle and the steering wheel angle which is turned, to estimate how the course of a curve driven by the vehicle fails.According to a further advantageous embodiment of the method according to the invention, the computing unit formulates for step c) the departure from the movement trajectory of the permissible driving corridor as a cost function, wherein the computing unit finds the at least one actuator to be controlled and the respective control amount by minimizing the cost function by means of a model-based predictive control. Different variables are possible, which can be used as a corresponding cost measure or, in particular, also a weighted sum of different variables. For example, this can be the Euclidean or orthogonal distance of the movement trajectory for delimiting the permissible driving corridor. It could also be the time duration with which the vehicle follows the movement trajectory outside the permissible driving corridor. It could also be a quadratic error of a corresponding distance value or the maximum distance of the vehicle from the boundary line of the permissible driving corridor and the like.A corresponding model-based predictive control algorithm then varies the corresponding driving dynamics parameters of the vehicle under the control of the actuators, so that the vehicle follows a corresponding movement trajectory within the permissible driving corridor while maintaining the cost measure. For this purpose, various actuator settings are simulated by the control algorithm until a suitable profile for respective actuators with respective respective activation values has been found, so that the underlying cost function is minimized. In the extreme case, it is thus achieved that the movement trajectory runs at least on the edge, i.e. the boundary line, of the permissible driving corridor. Advantageously, it is established as a further boundary condition that the vehicle should have a certain distance from the boundary line of the permissible driving corridor within the permissible driving corridor following the movement trajectory. The vehicle thus has a greater distance from the perimeter of the permissible driving corridor, which can be understood as a safety buffer. If the underlying driving dynamics model have errors, so that the determined course of the boundary line of the permissible driving corridor does not completely correspond to reality, the risk can be reduced here, which, despite a simulatively determined location position of the vehicle within the permissible driving corridor, does not threaten loss of control over the vehicle.A further advantageous embodiment of the method according to the invention further provides that the computing unit causes an indication message to be output in the vehicle when step c) is carried out. The notification message can be output visually, acoustically and / or haptically. For example, a warning light may illuminate in the vehicle interior, an instruction text may be displayed on a display device, and / or an animation may be played back, a warning message may be output by computer-generated speech via speakers, a warning sound may be sounded, the vehicle seat or the steering wheel may vibrate, and the like. By outputting the notification message, the vehicle occupants or the vehicle-guiding person are notified of the intervention of a driver assistance system provided by the method according to the invention. This improves the acceptance for the execution of the method according to the invention for the respective vehicle occupants. Thus, the execution of method step c) results in the corresponding actuator being actuated. This can be surprising for the vehicle-guiding person. This surprising effect can be attenuated by outputting the notification message. The person guiding the vehicle is thus also less distracted and can devote himself to the driving situation with greater attention, which ultimately likewise improves the safety in road traffic.According to a further advantageous embodiment of the method according to the invention, the control for the at least one actuator determined by the computing unit in step c) is overwritten by a manual control input made by a vehicle-guiding person. This can also improve the acceptance for the execution of the method according to the invention for vehicle occupants. The actuation of the actuator can be understood as a loss of control for the person driving the vehicle. Advantageously, the vehicle-guiding person can thus be given the option of manually overwriting the control specifications for the actuator. For example, a specific steering wheel angle can be predefined, which is overwritten by the vehicle-guiding person by turning the steering wheel. The vehicle could also be braked automatically, which can be overwritten by the person driving the vehicle pressing the accelerator pedal.By manually overwriting any actuator specifications, at least following the simulation derived by the driving dynamics model, the loss of control over the vehicle is imminent. However, the case may occur that, for example, due to a limited detection capability of a vehicle sensor, the current driving situation has been incorrectly evaluated by the computing unit, so that actually no loss of control at all is imminent for the vehicle. This can be correspondingly recognized by the vehicle-guiding person, so that the vehicle-guiding person can continue to control the vehicle as usual.In particular, the possibility of overwriting an activation of an actuator in vehicles with a partially automated operating mode carried out according to SAE levels 1 to 3 is provided. If a vehicle with an SAE level 4 and higher is used, the method described can be integrated into the system as a safety function and, if appropriate, can adapt or overwrite trajectories proposed by the driving system.A further advantageous embodiment of the method according to the invention further provides that the computing unit executes steps a) to c) only if a sensor data analysis carried out by the computing unit reveals that a hazardous situation is present. This allows the method according to the invention to be carried out only when it can also contribute to increasing safety. In situations deviating from this, the method according to the invention does not have to be carried out, as a result of which the load on the computing unit can be reduced. This allows the energy consumption of the computing unit to be reduced.To detect potential hazardous situations, the vehicle can rely on its environment detection and / or state sensor system. The computing unit thus recognizes a dangerous situation when the vehicle exceeds a threshold that is characteristic of a driving-dynamic extreme situation. This is the case, for example, when the amount of lateral or longitudinal acceleration, the amount of yaw rate, the amount of steering angular velocity, the amount of wheel slip, a deviation of a driving dynamics calculated by the driving dynamics model from a real measured driving dynamics or a quantity derived therefrom, a deviation of the vehicle from the driving lane or lane, or the like, exceeds a specific amount. This specific extent can depend on further boundary conditions, such as in particular the respective location of the vehicle, the calendar date, the time, the weather, the driving situation and the like. If, for example, the vehicle is travelling on a track, higher tolerable values can be set for corresponding lateral and longitudinal accelerations than when travelling in a road. Also, lower tolerable values may be set for the respective quantities, for example, when it is raining, instead of being sunny.A further advantageous embodiment of the method according to the invention further provides thatthe arithmetic unit determines a deviation between the result of the driving dynamics model and the actual driving dynamics actually occurring; andthe arithmetic unit adjusts at least one model parameter of the driving dynamics model while minimizing the deviation.As already mentioned, the driving dynamics parameters describe the driving dynamics. The driving dynamics model parameters, on the other hand, enter into the driving dynamics model and influence the determination of the driving dynamics parameters. If deviations occur between the driving dynamics determined by the driving dynamics model and the actually measured actual driving dynamics, this can be an indication that one or more driving dynamics model parameters do not correspond to reality. By adapting the respective driving dynamics model parameters, the quality of the driving dynamics model and thus ultimately its prediction accuracy can thus be improved.Since the driving dynamics model comprises the tire model, this means that adapting a driving dynamics model parameter also means adapting a corresponding tire model parameter. To find suitable driving dynamics model parameters, optimization methods can be used which modify at least one underlying driving dynamics model parameter in such a way that the deviation of the calculated driving dynamics from the measured actual driving dynamics is minimized. To solve this problem, artificial intelligence, such as "physics-induced networks", can also be used. Variables which are established in reality, such as a location-specific coefficient of adhesion / coefficient of friction with respect to the roadway, in particular reduced as a result of weathering, can also be read out from an online database. Corresponding values can be determined on a sensor basis, for example by infrastructure and / or measuring vehicles, and can be stored live in the database, in particular. The host vehicle can access this database and thus determine driving dynamics model parameters that are suitable for the respective location.Artificial intelligence used to estimate driving dynamics model parameters can be continued on-line permanently, for example on a backend of the vehicle manufacturer, or else off-line in the own vehicle. The training can take place continuously or else once or multiple times in one or more training campaigns. For example, such a training campaign can be started when a training data set required for this has reached a certain size. A new training iteration can also be started when a fixed period of time has elapsed, for example once per month. The updating of the corresponding AI model can also take place during a workshop stay. In this case, a quality of the underlying AI model, which is stored in the vehicle, can also be matched to a reference AI model, such that the AI model stored in the vehicle is updated or further trained only if the deviation from the reference exceeds a threshold value.In particular, the training of the AI model used for estimating driving dynamics model parameters is based on the data determined by the vehicles of a fleet of vehicles. This makes it possible to gradually improve the prediction quality of the AI model and also to reliably enable the reliable estimation of driving dynamics model parameters in situations which rarely occur during the normal driving operation of an individually considered vehicle.Alternatively or additionally, it would also be possible for the computing unit to correct the driving dynamics model, in particular the tire model, for minimizing the deviation by means of a respective correction model. This correction model can likewise be based on physical modeling or on the use of machine learning, in particular recurrent neural networks and Gaussian processes. From a mathematical point of view, the result of the correction model is added to the result of the driving dynamics model, so that the deviation between the variables calculated by the driving dynamics model and the actual driving dynamics decreases.According to a further advantageous embodiment of the method according to the invention, a machine learning model used to reduce the deviation from the computing unit is trained with data collected by the vehicle if at least one of the following criteria is fulfilled:a fixed time period has elapsed;a set amount of data has been aggregated;the deviation is greater than a threshold value;an event has arrived;the vehicle travels a road of a specified road type;the vehicle is stopped at a fixed location.Possible triggers for training the machine learning model have already been discussed above. An event is present, for example, when a workshop stay is performed, a round on a race track has been completed, and the like.A further advantageous embodiment of the method according to the invention further provides that the computing unit stores at least one adapted model parameter in a digital road map depending on location. In the digital road map, driving dynamics model parameters that are best suited for the respective road segment can thus be stored and loaded for later use. The digital road map, comprising the respective driving dynamics model parameters, can be managed locally within the computing unit of the respective vehicle. Central management by a vehicle-external computing device, such as said backend of the vehicle manufacturer, is also possible. For example, a tire parameter such as wheel slip or grip coefficient may be determined for a particular road segment and stored in the digital road map for that road segment.The computing unit preferably reads a location-dependent model parameter from the digital road map and uses this for the driving dynamics model, in particular for the tire model, when the vehicle travels on the respective location. Thus, corresponding location-specific driving dynamics model parameters can be not only acquired and stored in the digital road map in a location-specific manner, but can also be taken into account for the application. This allows the quality of the driving dynamics model used to be improved even further. This can be advantageous in particular in the case of central management by a computing device external to the vehicle. Thus, the vehicles of a fleet of vehicles of a vehicle manufacturer can function as measurement vehicles and map the corresponding road network.Situation-specific different characteristics of one and the same driving dynamics model parameter are expected for one and the same road segment. There are various possibilities as to which precise value of the respective driving dynamics model parameter is then to be stored in the digital road map, managed by the central computing device. For example, it can be a statistical mean value. Boundary conditions present by artificial intelligence in the detection of the respective driving dynamics model parameter could also be detected and patterns or relationships could be recognized. These patterns or relationships can likewise be stored in the digital road map. When the host vehicle now travels along the road section, it is checked which boundary conditions are currently present, so that the optimum vehicle model parameter is selected for the road section for the boundary conditions currently present. It can thus be achieved that an optimum set of driving dynamics model parameters can be found for all possible road sections already traveled on at respective boundary conditions.Communication between vehicles and the backend is possible in a proven manner. Thus, proven vehicle-to-vehicle communication interfaces can be used. The backend may be reached via the Internet, wherein the vehicles themselves may be connected to the Internet by mobile radio or Wi-Fi. For this purpose, a respective vehicle can have a telecommunication unit providing a mobile radio connection or can communicate with a WLAN hotspot located within communication range.Further advantageous embodiments of the method according to the invention for operating the vehicle also result from the exemplary embodiments which are described in more detail below with reference to the figures.The following are shown: FIG. 1 is a schematic top view of a vehicle in a first traffic situation, executing a method according to the invention for operating the vehicle; FIG. 2 is a schematic top view of the vehicle in a second traffic situation; and FIG. 3 shows a time-ray illustrating the time sequence of the method according to the invention.FIG. 1 shows a plan view of a vehicle 1, travelling on a curved section of a roadway 6. For this purpose, a vehicle-internal computing unit executes a driving dynamics model, comprising a tire model, wherein the computing unit computes a safe state space within which the vehicle 1 can be controlled in a controlled manner for each point. Furthermore, the computing unit determines a movement trajectory 2 for the vehicle 1. Thus, the movement trajectory 2 can be obtained, for example, from the currently selected vehicle speed and the steering angle that is being applied. It can also be a setpoint trajectory predefined by another driver assistance system or a navigation system.According to the invention, the computing unit transfers the safe state space into a permissible driving corridor 3. In this case, the vehicle 1 determines the permissible driving corridor 3 at at least one current time increment t i. shown in FIG. 3. The vehicle 1 or the computing unit preferably determines the permissible driving corridor 3 at different consecutive time increments. The permissible driving corridor 3 describes all locations that the vehicle 1 can reach at the current time from the current position within a prediction time window 5 (see FIG. 3 ) situated ahead in time, while preserving safe controllability.To determine the permissible driving corridor 3, the driving dynamics parameters that can be reached at most by the vehicle 1 within the prediction time window 5 can be used as a basis. This means, for example, that the accelerator pedal or brake pedal of the vehicle 1 is depressed to the maximum extent, as a result of which a specific travel speed is established at the end of the prediction time window 5. The travel speed profile resulting from the maximum acceleration or deceleration for the prediction time window 5 is then taken into account by the driving dynamics model. Instead, at least one currently present driving dynamics parameter can also be calculated. Typically, when driving straight ahead, no coastdown loss threatens for the vehicle 1, so that the course of the section 7 of the boundary line 4 immediately ahead of the vehicle 1 in the direction of travel results from those locations which the vehicle 1 can reach at the end of the prediction time window 5 starting from the location at the time increment t i when driving at the corresponding travel speed. The course of the two lateral parts 8 of the boundary line 4 results from the maximum steering angles which can be adjusted, at which the vehicle 1 just still retains its control, taking into account the speed that is being established (or the current speed). If a larger steering angle were to be applied, at least one wheel of the vehicle 1 could slip, break out the front and / or rear axle of the vehicle 1 or even tip over the vehicle 1. Depending on the situation, the course of the lateral parts 8 (or of a subsection) of the boundary line 4 can also result from the maximum travel path that can be covered without loss of control if the loss of controllability of the vehicle is not to be expected at the considered travel speed for the respective steering angle.The computing unit of the vehicle 1 now balances the movement trajectory 2 with the permissible driving corridor 3. If the movement trajectory 2 runs within the permissible driving corridor 3, no further measures are required. If, on the other hand, the movement trajectory 2 leaves the permissible driving corridor 3, there is a risk of loss of control. This makes it possible to predictively actuate an actuator of the vehicle 1 in good time in order to change at least one current driving dynamics parameter in order to cause the movement trajectory 2 for the prediction time window 5 to finally run within the permissible driving corridor 3. For this purpose, the movement trajectory 2 can be displaced and / or the shape or size of the permissible driving corridor 3 can be adapted, in particular enlarged. A brake of the vehicle 1, a differential or locking differential, an actuator for setting the steering angle and / or the drive unit of the vehicle 1 itself are in particular suitable as the actuator. The vehicle 1 can also be embodied purely battery-electrically and thus have its own drive unit for each axle or even for each wheel.FIG. 1 shows in the sections following the arrow 9 how the permissible driving corridor 3 or the movement trajectory 2 could change. In the middle section, an enlargement or distortion of the permissible driving corridor 3 is shown. A displacement of the movement trajectory 2 is shown in the lower section. Both variants can also be combined with one another.The execution of the method according to the invention makes it possible to control the vehicle 1 particularly safely, even in manual operation. Because the vehicle 1 does not reactively adapt its behavior to the respective driving situation, but predicts a driving state in the future, the respective actuators for preventing covered driving states can be controlled at an early stage, which further reduces the probability of a loss of control.FIG. 2 shows the vehicle 1 in a further driving situation in which a commercial vehicle 10 blocks the lane 11 traveled by the vehicle 1. A movement trajectory 2 is thus determined, which is carried around the commercial vehicle 10. As FIG. 2 shows, the movement trajectory 2 also leaves the permissible driving corridor 3 here. Under certain circumstances, it may not be possible to have the movement trajectory 2 run within the permissible driving corridor 3. A deviation 12 between the movement trajectory 2 and the permissible driving corridor 3 can thus be determined, for example a distance between the movement trajectory 2 and the permissible driving corridor 3 or a time duration at which the vehicle 1 follows a section of the movement trajectory 2 lying outside the permissible driving corridor 3.In order to determine which actuator and to what extent the respective actuator is to be actuated, a cost function can be minimized. The aim is to minimize the deviation 12. FIG. 2 shows a case in which, although potentially the loss of control of the vehicle 1 is imminent, the extent of the loss of control can be limited or even entirely prevented. The cost function is particularly advantageously minimized on the basis of the execution of a model-based predictive control.FIG. 3 shows a time line 13 showing the course of time t. At a current time increment t i the method steps according to the invention are carried out. This means the determination and matching of the movement trajectory 2 and the permissible driving corridor 3 with one another for the prediction time window 5. In the exemplary embodiment shown in FIG. 3, there is a risk of loss of control via the vehicle 1, with the result that an actuator is actuated in order to manipulate the current driving dynamics parameters. In this case, FIG. 3 shows a setpoint profile 14 of the respective actuator, for example of the brake. The setpoint profile 14 corresponds, for example, to the requested brake pedal position or the requested braking torque.Particularly advantageously, the respective method steps are carried out again at a time increment t i+1 following the current time increment t i. The time interval between the time increments t i and t i+1 can be smaller, equal to or greater than the prediction time window 5. For example, the time interval between the two time increments t i and t is i+120 milliseconds long (not drawn to scale). At the time increment t i+1 the permissible driving corridor 3 is thus again determined accordingly at the present location of the vehicle 1 and is matched to the respective movement trajectory 2 now resulting. A new setpoint profile 14 for the underlying actuator can then be predefined.References included in the specificationThis list of documents cited by the applicant has been produced in an automated manner and is only included for the better information of the reader. The list is not part of the German patent application or utility model application. The DPMA does not take any adhesion for any faults or omissions.Patent Literature citedDE 10 2019 205 405 A1

[0004] DE 10 20196 933 A1

[0005] DE 10 2016 011 244 A1

[0006]

Claims

Method for operating a vehicle (1), wherein an in-vehicle computing unit calculates a safe state space using a driving dynamics model comprising a tyre model, wherein the vehicle (1) can be controlled in a controlled manner for each point in the safe state space, the computing unit determines a movement trajectory (2) for the vehicle (1) and actuates at least one actuator of the vehicle (1) in order to control the vehicle (1) within the safe state space, characterized in that a) the computing unit calculates an admissible driving corridor (3) for the vehicle (1), wherein the computing unit determines, for determining a boundary line (4) of the admissible driving corridor (3), which locations the vehicle (1) can still just reach at the current time increment (t i) under a control for a prediction time window (5) according to the driving dynamics parameters limiting the safe state space; b) the arithmetic unit matches the movement trajectory (2) with the permissible driving corridor (3) and determines whether the movement trajectory (2) leaves the permissible driving corridor (3) for the current time increment (t i) and, if so: c), the arithmetic unit controls the at least one actuator for manipulating the current driving dynamics parameters, so that the movement trajectory (2) for the prediction time window (5) runs within the permissible driving corridor (3).Method according to Claim 1, characterized in that, in step c), the arithmetic unit actuates the at least one actuator for reducing the current driving dynamics parameters in order to increase the area of the permissible driving corridor (3).Method according to Claim 1 or 2, characterized in that the arithmetic unit actuates the at least one actuator in step c) in order to place the profile of the movement trajectory (2) within the permissible driving corridor (3).Method according to one of Claims 1 to 3, characterized in that the arithmetic unit repeats steps a) to c) at a future time increment (t i+1) following the current time increment (t i) in particular while maintaining a fixed repetition frequency.Method according to one of Claims 1 to 4, characterized in that - if the vehicle (1) is controlled manually, the arithmetic unit estimates an estimation trajectory which is dependent on the control behavior of a person carrying the vehicle and uses the estimation trajectory in step b) for the movement trajectory (2); or - if the vehicle is controlled at least in a semi-automated manner, the arithmetic unit uses a setpoint trajectory provided for guiding the vehicle (1) in step b) for the movement trajectory (2).Method according to one of Claims 1 to 5, characterized in that the arithmetic unit for determining the movement trajectory (2) determines a route profile of the roadway (6) on which the vehicle (1) is travelling and a position of the vehicle (1) relative to the roadway (6), and plans the movement trajectory (2) such that a fixed relation of the vehicle position to the route profile is observed.Method according to Claim 6, characterized in that the arithmetic unit excludes the route profile from a digital road map.Method according to Claim 6 or 7, characterized in that the arithmetic unit determines the route profile from sensor data which are generated by at least one of the following sensors: - at least one environment sensor; - at least one vehicle state sensor.Method according to one of Claims 1 to 8, characterized in that the arithmetic unit formulates for step c) the departure from the movement trajectory (2) of the permissible driving corridor (3) as a cost function, wherein the arithmetic unit finds the at least one actuator to be controlled and the respective control extent by minimizing the cost function by means of a model-based predictive control.Method according to one of Claims 1 to 9, characterized in that the arithmetic unit causes an indication message to be output in the vehicle (1) when step c) is carried out.Method according to one of Claims 1 to 10, characterized in that the control for the at least one actuator determined by the arithmetic unit in step c) is overwritten by a manual control input carried out by a person driving the vehicle.Method according to one of Claims 1 to 11, characterized in that the arithmetic unit executes steps a) to c) only if a sensor data analysis carried out by the arithmetic unit reveals that a hazardous situation is present.Method according to one of Claims 1 to 12, characterized in that - the arithmetic unit determines a deviation between the result of the driving dynamics model and the actual driving dynamics actually arising; and - the arithmetic unit adapts at least one model parameter of the driving dynamics model while minimizing the deviation.Method according to one of Claims 1 to 13, characterized in that - the arithmetic unit determines a deviation between the result of the driving dynamics model and the actual driving dynamics actually arising; and - the arithmetic unit corrects the driving dynamics model, in particular the tire model, by minimizing the deviation by means of a respective correction model.Method according to Claim 13 or 14, characterized in that a machine learning model used to reduce the deviation by the arithmetic unit is trained with data collected by the vehicle (1) if at least one of the following criteria is fulfilled: - a defined time duration has elapsed; - a defined data amount has been aggregated; - the deviation is greater than a threshold value; - an event has arrived; - the vehicle (1) travels on a road of a defined road type; - the vehicle (1) stops at a defined location.Method according to one of Claims 13 to 15, characterized in that the arithmetic unit stores at least one adapted model parameter in a digital road map as a function of location.Method according to Claim 16, characterized in that the arithmetic unit reads a location-dependent model parameter from the digital road map and uses it for the driving dynamics model, in particular the tire model, when the vehicle (1) travels on the respective location.

Images