Method for guiding a vehicle, controller, vehicle, and computer program

Abstract

The invention relates to a method for guiding a vehicle (1). A manipulated variable is ascertained by means of a lateral controller (7) on the basis of a lateral deviation error which describes a lateral offset (4) of the vehicle (1) relative to a specified target trajectory (5), wherein the lateral position of the vehicle (1) is regulated to the target trajectory (5) on the basis of the manipulated variable. At least one calibration value is additionally applied to the manipulated variable, said calibration value being ascertained by at least one calibration function (12) at least partly on the basis of a temporal integration of the lateral deviation error.

Description

[0001] Description

[0002] Method for lane guidance of a vehicle, control unit, vehicle and computer program

[0003] The invention relates to a method for tracking a vehicle, wherein a control variable is determined by a lateral control system as a function of a lateral offset error describing a lateral displacement of the vehicle relative to a predetermined target trajectory, wherein a lateral position of the vehicle is adjusted to the target trajectory as a function of the control variable. Furthermore, the invention relates to a control unit, a vehicle, and a computer program.

[0004] Vehicles can be equipped with systems or functions for automatic lane guidance. These enable the lateral position of the vehicle to be automatically adjusted while driving according to a specified trajectory. The lateral position of the vehicle, i.e., the sideways position of the vehicle in a lane, can be changed or adjusted to a target position, for example, through automated steering movements.

[0005] Such lane keeping functions generally use environmental data, which is generated by sensors for detecting the vehicle’s surroundings, such as cameras, radar, or lidar systems, and / or by means of a

[0006] navigation system, for example a GPS, in order to guide vehicles along a planned trajectory. The trajectory is designed, for example, for a high level of driving comfort and therefore does not contain any strong or rapid changes in the vehicle's lateral position. The guidance of the vehicle along such a trajectory is generally carried out as a control system in which at least the lateral deviation of the vehicle from the planned trajectory is continuously determined. Depending on the deviation, control commands of a suitable size and direction are issued, for example to an electric steering assistance system of the vehicle. Criteria for assessing the control quality of such systems include, for example, the amplitude and dynamics of the lateral control deviation and the frequency of jerky movements on the vehicle's steering wheel.Ideally, the lateral control error is zero for all driving situations, so that the steering interventions by the lane guidance or the resulting steering wheel movements are determined exclusively by the course of the planned trajectory.

[0007] The performance of such a lane guidance system depends on the quality and performance of the sensors and actuators used in the vehicle. To a certain extent, the performance of an overall system can be increased by using multiple sensors or actuators and / or by implementing suitable software measures. This is particularly true for driver assistance systems, such as those that provide lane-keeping support functions.

[0008] Nevertheless, sensor errors—that is, any deviation of a sensor signal from its physical counterpart—affect the nature of the control system and thus also the control quality of the vehicle's lane guidance, particularly due to the resulting error propagation. Depending on the sensor information, sometimes different, sometimes overlapping, chains of effects come into play.

[0009] An imaging error in the camera capturing the lane markings can, for example, result in the determined curvature of the road deviating from the actual curvature, which represents a so-called ground truth error. Lane guidance typically comprises at least one controller and / or at least one feedforward control that adapts the lateral vehicle position depending on the determined road curvature. Curvature information describing the road curvature, often determined at a look-ahead point of the vehicle, can be used to calculate the steering angle required for the respective curvature using an inverted vehicle model and to switch this angle to a steering angle controller as a target steering angle component, for example as part of a feedforward control.If the curvature information has an offset compared to the actual curvature due to an error, the pre-controlled target steering angle also contains an offset component, which acts like a disturbance on the lateral control. This has the effect that the vehicle does not follow the trajectory as expected when lane keeping assistance is activated, but depending on the design of the lateral control or a trajectory following controller - i.e. use of controller types with or without steady-state accuracy - has a permanent lateral offset from the planned trajectory or at least causes a transient swerving before the vehicle then follows the planned trajectory after a compensation process has subsided. The permanent or transient lateral offset can impair driving comfort, which can have a negative impact on the driver acceptance of the driver assistance system.

[0010] The degree of camera aberration typically varies over the operating life of the driver assistance system, as camera calibration routines are typically continuously executed. This means that, depending on the status of the camera calibration, the magnitude of the curvature feedforward control offset and thus the degree of impairment of the vehicle's lateral guidance are time-varying, although the camera's transmission properties and thus also the curvature offset do not change abruptly, but rather in a band-limited or gradual manner.

[0011] The steering angle usually represents a significant auxiliary or controlled variable within a lane-guided assistance system, so routines for calibrating a steering angle offset are often used. Due to the finite accuracy of the calibration routines, the steering angle offset determined in this way always contains an uncompensated component. This residual offset error in a steering angle signal can have several causes. Typical calibration routines are based, for example, on the signal from a yaw rate sensor or a wheel speed sensor. Errors in these sensors, such as an offset and / or a linearity error, or in the case of wheel speed sensors, speed differences due to different tire pressures on an axle, then propagate into the determined steering angle offset.The effects of steering angle offsets are comparable to the previously described case of a curvature offset: When a lane keeping assistance system is activated, the vehicle does not follow the specified trajectory as expected, but, depending on the design of the lateral control or the trajectory following controller, exhibits a permanent or transient lateral offset from the planned trajectory with the adverse effects on the comfort and acceptance of the driver assistance system.

[0012] This applies equally to an uncompensated offset of a front axle steering angle and a rear axle steering angle.

[0013] The sensor signals for the yaw rate, the front and rear axle steering angle, and for lateral acceleration can also be used for other functions within the vehicle, which can also influence the lateral position or lateral guidance of the vehicle. For example, these sensor signals can also be used to estimate disturbance forces and load moments acting on the vehicle, as well as to estimate the vehicle's sideslip angle, for example, due to road inclination and / or crosswind. In such a case, the determined variables then also contribute to the vehicle's lateral guidance as a target steering angle component during disturbance variable compensation, with corresponding offsets of the sensor signals also propagating as errors to the target steering angle and thus the lateral position of the vehicle.The result here is also an undesirable transient or permanent lateral offset of the track-guided vehicle.

[0014] DE 10 2008 026 233 B4 discloses a method for offset compensation of a steering angle in a motor vehicle using a vehicle model. A yaw rate is calculated based on a detected steering angle, a vehicle speed, and an assumed steering wheel angle offset. Furthermore, the calculated yaw rate is compared with a measured yaw rate, the difference between which forms an error equation, from which a steering wheel angle offset is then determined. This offset is fed back into the vehicle model to iteratively determine and compensate for the steering angle offset. Such an offset calibration of a yaw rate or lateral acceleration has a finite accuracy and always lags behind the actual offset, which can result in additional, undesirable steering interventions even in the presence of the actual offset.

[0015] The invention is based on the object of improving the accuracy of a vehicle's lane guidance and thus the driving comfort and the acceptance of the assistance system by the driver even under the influence of offset errors.

[0016] To achieve this object, in a method of the type mentioned at the outset, the invention provides that the manipulated variable is additionally subjected to at least one calibration value, wherein the calibration value is determined by at least one calibration function at least partially as a function of a temporal integration of the transverse offset error.

[0017] The vehicle's lateral position is thus adjusted using the sum of the control variable and at least one calibration value. Depending on the design of the lateral control and / or lane guidance system, it is possible that additional components may be applied to the control variable in addition to the at least one calibration value.

[0018] The manipulated variable, which is subjected to the calibration value, influences the lateral position of the vehicle. The manipulated variable can be, for example, a target steering angle. This can be specified, for example, at a steering angle interface to an electric power steering (EPS). Alternatively, the manipulated variable can also be specified at a steering torque or torque interface of an EPS that generates a steering angle.

[0019] The calibration value is determined by the calibration function, particularly depending on the sign and amplitude of the integrated lateral deviation error, such that the calibration value, when applied to the control variable, always aims to reduce the current lateral deviation error. Applying the calibration value to the control variable thus corrects or compensates the target steering angle by the time-integrated lateral deviation error.

[0020] The lateral offset describes the lateral or sideways offset of the vehicle, i.e., an offset in the vehicle's transverse direction, from the specified target trajectory. The lateral offset can therefore also be referred to as lateral offset or lateral deviation. In other words, the lateral offset expresses the current deviation of a vehicle's actual position from a desired target position described by the target trajectory. The determination of the vehicle's lateral offset relative to the target trajectory can be related to an agreed point along the vehicle's longitudinal axis. Thus, the lateral offset can advantageously be related to the geometric center of the rear axle or the front axle, or to any other forecast point along the vehicle's longitudinal axis.

[0021] The lateral deviation error can be continuously determined, for example, based on sensor data from at least one of the vehicle's environmental sensors. The environmental sensor can be, for example, a camera, a lidar sensor, a radar sensor, and / or an ultrasonic sensor. Additionally or alternatively, the lateral deviation error can also be determined based on position data from a navigation system. The lateral deviation error has a time-dependent value, which is integrated by the calibration function.

[0022] The lateral controller used for lateral control can, for example, be a trajectory tracking controller or include one. The calibration function, which can also be referred to as centering calibration or centering function, is present in addition to the trajectory tracking control and thus forms a secondary or higher-order control loop that also influences the manipulated variable and thus also the lateral position of the vehicle. In particular, the dynamics of the calibration function are significantly lower than those of the lateral control, so that the time-integrated lateral deviation error only influences the manipulated variable very slowly.

[0023] As a calibration function or as a basic calculation rule for determining the calibration value, for example, an integrator with a comparatively small integration coefficient can be used, which results in an integration of the cross-deviation error with low dynamics and thus also an influence of the calibration value with low dynamics.

[0024] According to the invention, the at least one calibration function can be implemented as an I-element or as a PT1 element. Compared to a pure I-element, a PT1 element also has an integrating property, but the advantage is that it does not require any of the measures necessary for limiting the integrator state that are necessary for I-elements. This advantage must be weighed against the disadvantage of the non-stationary exact compensation of the offset variables in the respective application by utilizing the parameterization degrees of freedom of the PT1 element.

[0025] The adaptation dynamics of the calibration value are specifically dimensioned to be on the order of magnitude of the change dynamics of the offset errors. One measure of this is, for example, the largest expected temperature-related offset drift of the sensors used for lateral control, converted to the corresponding steering angle level. This has the advantage of preventing overshoot in the lateral control, which, for example, with a higher dynamic range of the calibration function, would lead to the intervention of other compensation functions acting on the steering angle, such as compensation for road inclination and / or crosswind compensation.

[0026] By using the at least one calibration function or by applying the at least one calibration value to the control variable, it is advantageously achieved that offset errors that continuously affect the lane guidance are partially or completely eliminated over time, or that these offset errors are partially or completely compensated over time. Compensation functions that affect the steering angle are often not continuously active. For example, a lane keeping function usually only becomes effective after activation by the driver.

[0027] If the compensation functions detect a perceived control error due to offset errors at the time of their activation, they will output manipulated variables in an attempt to reduce the control error. Depending on the dynamics of the compensation functions, this can lead to jerky steering movements. By compensating for offset errors in advance, the occurrence of jerky steering movements to correct a built-up lateral offset can be advantageously avoided, thereby improving occupant comfort when lane guidance is activated. This also improves the acceptance of the lane guidance system or the driver assistance system that provides the lane guidance.

[0028] In summary, the higher-level calibration function can also be interpreted as a fixed-value control with a low dynamic range and a zero setpoint for the lateral offset error. Its effect is to reduce the sum of all direct and indirect offset influences of the sensors used in the system on the lateral offset of the vehicle during automatic lateral or lane guidance.

[0029] According to the invention, it can be provided that the time constant of the calibration function is greater than the dominant time constant of the cross-control, in particular by at least a factor of 2. Preferably, the time constant of the calibration function can also be greater than the dominant time constant of the cross-control by at least a factor of 5, at least a factor of 10, at least a factor of 100, or at least a factor of 1000. In principle, even greater differences between the time constants are also conceivable. A calibration function implemented, for example, as an I-element can, for example, have an integration coefficient between 0.01 degrees / (m*s) and 0.00005 degrees / (m*s).

[0030] In a preferred embodiment of the invention, it can be provided that an integration value of the calibration function is determined by the temporal integration of the lateral deviation error, wherein the integration value is retained when the track guidance is deactivated and is used as the starting value of the temporal integration of the lateral deviation error when the track guidance is reactivated.

[0031] The integration value describes the total value that has been generated by the temporal integration of the cross-offset error up to a point in time assigned to the integration value. The integration value therefore represents the result of the temporal integration of the cross-offset error up to this point in time for a specific point in time. If multiple calibration functions are used, an integration value assigned to the respective calibration function is generated separately for each of the calibration functions. The at least one integration value can be stored in a retrievable manner, for example, in a memory device of a computing device executing the method, such as a control unit.

[0032] The state of the calibration function, i.e., the state of an integrator or a comparable functional unit forming the calibration function, is therefore not reset during the current ignition cycle of the vehicle, for example, but is maintained even if the driver assistance system is repeatedly deactivated and activated or if the driver intervenes. However, resetting the integrator in the event of detected implausible activity, for example, a detected or communicated jump in the output of one or more offset calibrations of the vehicle sensors, or a request via a diagnostic interface of the vehicle, can also be provided. It is possible for the integration value to be retained even beyond a vehicle ignition cycle, i.e., across multiple journeys.

[0033] According to the invention, it can be provided that the lateral control comprises a feedforward control, wherein the feedforward control determines a feedforward control variable as a function of a measured and / or predicted curvature value of a roadway traveled by the vehicle, wherein the manipulated variable contains the feedforward control variable as a component. A measured curvature value and / or a predicted curvature value can be determined in particular from the sensor data of one or more of the vehicle's environmental sensors. The feedforward control can, in particular, use the measured curvature value and / or the predicted curvature value via an inverted vehicle model to calculate the steering angle required for the respective curvature and switch this as a feedforward control variable, for example, to the manipulated variable of a trajectory follower controller of the lateral control. The vehicle model changes its transmission behavior as a function of the vehicle speed.This shows that an error in the measured curvature value has an influence on the lateral guidance of the vehicle that depends on the vehicle speed.

[0034] According to the invention, it can be provided that the pre-control takes place as a function of a vehicle model describing a self-steering gradient of the vehicle, wherein the self-steering gradient is adapted during operation of the vehicle as a function of at least one vehicle parameter, in particular a vehicle mass and / or a tire stiffness.

[0035] When the vehicle corners, the lateral deviation error resulting from a feedforward control error depends not only on an offset error in the curvature, but also on the inaccuracy of the stored vehicle parameters, in particular on uncertainties in a self-steering gradient used in the vehicle model. An error or deviation in the self-steering gradient has the effect of disrupting the vehicle's lateral control and therefore proportionally also causes an additional lateral deviation error, which leads to an erroneous deflection of the calibration value generated by the calibration function. In such a case, the calibration value therefore reacts not only to sensor offset variables. However, as long as the errors or deviations in the self-steering gradient are limited, their impact is only minor due to the limited dynamic range of the calibration function.

[0036] In order to also consider the influence of larger errors and / or deviations in

[0037] To avoid self-steering gradients, the gradient can be adjusted, particularly continuously, depending on at least one vehicle parameter. The adjustment can be made, for example, depending on the current vehicle mass, the current tire stiffness of one or more of the vehicle's tires, and / or other vehicle parameters.

[0038] In a preferred embodiment of the invention, it can be provided that the temporal integration of the lateral deviation error is carried out with a weighting factor, in particular an integration coefficient, which depends on a speed of the vehicle and / or increases with increasing amount of the lateral deviation error.

[0039] While, for example, the use of a constant integration coefficient can represent an easily implemented compromise across the entire vehicle speed range, the use of a speed-dependent integration coefficient or a comparable integration weighting factor has the advantage of enabling better compensation for speed-dependent offsets. The influence of offset variables on the lateral deviation error can arise, for example, due to the commonly used setting of the control parameters for the lateral control system depending on the vehicle speed. This also changes the disturbance compensation properties of the lateral control system and thus also the effect of sensor offset variables on the lateral deviation error.To account for this, the temporal integration can also be made dependent on the driving speed by using a speed-dependent weighting factor, such as a speed-dependent integration coefficient. This advantageously allows for individually adapted calibration dynamics to be achieved for each driving speed range.

[0040] Additionally or alternatively, it is possible for the weighting factor to be made dependent on the magnitude of the cross-slip error, i.e. on the value of a cross-slip error to be integrated and / or a previously integrated cross-slip error, with greater weighting being applied for larger values ​​of the cross-slip error. For this purpose, for example, an assigned weighting factor or integration coefficient can be determined for the current cross-slip error depending on a non-linear assignment rule and / or one or more threshold values ​​for the cross-slip error and used in the temporal integration of the cross-slip error. The calibration speed can be effectively increased by progressive weighting of the magnitude of the cross-slip error, i.e. by increasing the weighting factor or integration coefficient with increasing magnitude of the cross-slip error.

[0041] For example, the integration coefficient and thus the adaptation rate of the calibration function can be switched between two or more values, with the integration coefficient being selected higher for larger absolute values ​​of the offset error. Advantageously, simply doubling the integration coefficient or the integration constant for a lateral offset error of more than 0.2 m leads to an effective, faster initial adaptation to the existing offset situation when the driver assistance function is started.

[0042] Even in the case that the vehicle has an additional sensor offset calibration function which corrects a detected offset drift of at least one of the vehicle's sensors after a confirmation time or debounce time has elapsed by means of a step-like correction of the respective offset, increasing the adaptation rate of the centering calibration as a function of the magnitude of the cross-slip error can ensure faster adjustment to the new offset situation. This becomes all the more important the longer the debounce time of the sensor offset calibration functions is, since during this time the calibration function increasingly compensates due to the temporal integration of the cross-slip error, which must then be integrated back after a signal update of the sensor offset calibration functions has taken place. This can be advantageously achieved by using a greater weighting orA larger integration value can increase the speed of the compensation process. Since step-like corrections to a sensor offset calibration function of the vehicle directly lead to a lateral jerk of the track-guided vehicle, which depends on the jump height, it can be provided for a vehicle comprising a control device designed to carry out a method according to the invention that all sensor offset calibration functions in the vehicle make a change to the respective offset not step-like, but only within a band-limited range.

[0043] According to the invention, it can be provided that the temporal integration of the cross-deviation error is stopped by the calibration function when:

[0044] - the integration value of the calibration function determined by the temporal integration of the cross-deviation error corresponds to a specified limit value,

[0045] - the measured and / or predicted curvature value of the roadway or a road surface travelled by the vehicle exceeds a predetermined threshold value,

[0046] - a curvature value of the target trajectory exceeds a specified threshold,

[0047] - the product of the square of the current vehicle speed and the curvature of the road exceeds a predetermined threshold, and / or

[0048] - the product of the square of the current vehicle speed and the curvature value of the target trajectory exceeds a specified threshold.

[0049] The contribution of the calibration value to the manipulated variable can be limited to a limit value by at least temporarily stopping the temporal integration of the lateral offset error in the calibration function, or by limiting the total value of the integration to the limit value. The limit value represents a maximum amplitude and can, for example, be based on the sum of the maximum influences of all offset variables on the steering angle setpoint and the actual steering angle. This approach advantageously eliminates the need to provide measures to prevent integrator windup effects in the calibration function.

[0050] As an alternative to limiting the integrator value to a threshold, the calibration function can also include a low-pass filter function, such as a PT-1 filter. This can also prevent windup effects. While a sensor offset cannot be completely compensated, the centering accuracy achieved by the calibration function can still be sufficient.

[0051] Additionally or alternatively, it is possible for the temporal integration of the lateral deviation error to be stopped by the calibration function if the or a measured and / or predicted curvature value of the or a roadway traveled by the vehicle exceeds a predetermined threshold value and / or if a curvature value of the target trajectory exceeds a predetermined threshold value.

[0052] When the vehicle corners, the lateral deviation error resulting from a possible feedforward control error depends not only on a curvature offset error, but also on the inaccuracy of the stored vehicle parameters, in particular on uncertainties in the vehicle's self-steering gradient. Incorrect vehicle curvature feedforward control can disrupt the vehicle's lateral control and thus proportionally cause an additional lateral deviation error. This additional lateral deviation error can lead to an erroneous intervention of the calibration function, which can be advantageously avoided or reduced by suspending the calibration function or at least by temporally integrating the lateral deviation error by the calibration function.

[0053] The effect of an error in the curvature precontrol on the ferry operation or the lateral position of the vehicle can depend in particular on the vehicle speed. To account for this speed dependence, it can therefore be provided that the temporal integration of the lateral deviation error is stopped by the calibration function when the product of the square of the current vehicle speed and the curvature of the road exceeds a predefined threshold and / or when the product of the square of the current vehicle speed and the curvature of the target trajectory exceeds a predefined threshold.

[0054] In a preferred embodiment, according to the invention, the temporal integration of the lateral deviation error is stopped and / or the application of the calibration value to the manipulated variable is suspended if a steering torque above a predetermined threshold value and / or a steering torque that increases the current lateral deviation error is generated by a steering intervention by a driver of the vehicle.

[0055] This advantageously prevents lateral offset errors caused by deliberate driver intervention from being included in the calibration value determined by the calibration function, since these lateral offset errors are not due to a sensor offset or the like that needs to be compensated. The influence of a lateral offset from the target trajectory deliberately induced by a driver, which does not originate from the lane guidance function, on the calibration value is thus advantageously avoided.

[0056] According to the invention, it can be provided that when a driver of the vehicle intervenes in the steering, which generates a steering torque that reduces the current lateral deviation error, the temporal integration of the lateral deviation error takes place with an increased weighting, in particular with an increased integration coefficient, at least for a predetermined period of time.

[0057] If the driver initiates a steering intervention that generates a steering torque acting in the same direction as the calibration value generated by the calibration function, this can be interpreted as confirmation of the calibration function's adaptation direction. Using an increased weighting or an increased integration value in this case advantageously enables improved calibration dynamics through the calibration function. From the driver's perspective, this translates into the ability to shift the vehicle's course laterally more toward the center of the lane, and the course will remain there in the future, even if the corrective steering torque is reduced.

[0058] In a preferred embodiment, it can be provided according to the invention that a plurality of integrating functions are used, wherein a different speed interval is assigned to each of the integrating functions, wherein the integrating functions each integrate the transverse offset error over time at a speed of the vehicle lying within the speed interval assigned to them, wherein the calibration value of the integrating function, whose speed interval includes the current speed of the vehicle, or a total calibration value determined as a function of this calibration value is applied to the manipulated variable.

[0059] As previously described, the effect of offset variables on the lateral deviation error is partially dependent on the vehicle speed. This applies, for example, to the offset errors contained in an output variable of a sensor disturbance compensation function or to the offset error that is a component of a curvature pre-control. To keep the lateral deviation error as low as possible, the calibration value connected to the manipulated variable can also be determined as a function of speed. Particularly due to the high integration time constants and the associated low dynamic response, adapting the calibration value of an individual calibration function to the changed vehicle speed can take a relatively long time, during which time the accuracy of the tracking can be reduced.

[0060] To account for this circumstance, it is advantageous to use a plurality of calibration functions for the entire vehicle speed range instead of a single calibration function. Each of the calibration functions can determine its own calibration value and, in particular, store it separately from the other calibration values. The ability to store the calibration values ​​separately, for example, by storing them in non-volatile memory, ensures that they can be continuously expanded through temporal integration during continued operation, and that all states of the plurality of calibration functions can be initialized separately when the vehicle's ignition is restarted.

[0061] The entire range of possible vehicle speeds can be subdivided into several sub-intervals, each of which covers a limited speed range and to which a dedicated calibration function is assigned. The manipulated variable is then assigned the calibration value of the calibration function that is assigned to the speed interval in which the current vehicle speed lies. This means that only one calibration function is active at a time, while the remaining calibration functions, in particular the temporal integration of the lateral deviation error by the other calibration functions, are stopped or an input signal of zero is switched to their inputs. The currently active calibration function, on the other hand, can be supplied with the current lateral deviation error. The lateral deviation error can be weighted with a constant weighting factor orThis can be done via an integration coefficient or a weighting factor or integration coefficient dependent on the vehicle speed. The calibration value of the integration function, whose speed interval includes the current vehicle speed, or a total calibration value determined based on this calibration value is then applied to the manipulated variable. The advantage of using multiple calibration functions or a bank of calibration functions over using a single calibration function is particularly evident during acceleration or deceleration, and thus when the vehicle's speed ranges change frequently.

[0062] According to the invention, the total calibration value can be determined from two or more calibration values ​​generated by the integration functions in such a way that the total calibration value remains constant over time, even when switching between two or more speed intervals. This avoids abrupt changes in the calibration value applied to the manipulated variable, and thus abrupt interventions when changing the calibration functions due to a change in the vehicle's speed.In a preferred embodiment, an integration value of the calibration function assigned to the calibration function can be determined by the calibration functions by the temporal integration of the lateral deviation error, wherein the integration values ​​of the one or more integration functions in which the current speed lies outside the respective speed interval are continuously adapted, in particular at a constant or a vehicle speed-dependent rate, to the integration value of the calibration function in whose speed interval the current speed of the vehicle lies.

[0063] In order to avoid a situation where, when the automatic lane guidance function is used primarily in a narrow speed range, only a few or only a single calibration function are loaded with the cross-offset error, and therefore only these ultimately correctly represent the current offset situation, the inactive calibration functions not used at the current speed can also continue to integrate the current cross-offset error over time. The integration values, i.e. the integrator states of the calibration functions, which are each assigned to a range adjacent to the current speed range, are limited to a maximum value of the integration value of the calibration function assigned to the current speed range. In addition, the weighting factors orthe integration coefficients of the calibration functions assigned to non-current speed ranges are reduced compared to the active case, since if the integration coefficients remain unchanged, the advantage of subdividing into multiple integrator states would diminish. The degree of reduction of the adjacent integration coefficients depends on the expected maximum sensor drift values ​​and is, for example, in the order of 10% to 20%. This can prevent the states of the inactive calibration functions from becoming increasingly less up-to-date due to the typical drift of the offset values ​​for curvature, yaw rate, and lateral acceleration, which could temporarily lead to a larger lateral offset error and reduced driving comfort when entering an adjacent speed range until the adaptation to the current offset situation is complete.

[0064] The continuous adjustment of the integration values ​​of one or more integration functions where the current speed lies outside the respective speed interval, in particular at a constant or vehicle-speed-dependent rate, to the integration value of the calibration function in whose speed interval the current vehicle speed lies, can be considered the implementation of a forgetting factor. In this case, the states of the inactive calibration functions approach the state of the currently active calibration function, for example, with a uniform gradient or a predetermined gradient individually defined for the respective speed range.

[0065] For a control unit according to the invention for generating a manipulated variable for at least one lateral guidance actuator of a vehicle, it is provided that the control unit is configured to carry out a method according to one of the preceding claims. The control unit can also be configured to carry out lateral control, to determine the desired trajectory, and / or to carry out curvature pre-control. Alternatively, the control unit can implement only some of these functions. The control unit can be connected to at least one sensor of the vehicle, in particular at least one environmental sensor that at least partially detects the vehicle's surroundings and / or at least one speed sensor that detects the vehicle's speed, and / or to a navigation system of the vehicle.

[0066] For a vehicle according to the invention, it is provided that it comprises at least one lateral guidance actuator and a control unit according to the invention, wherein the control unit is configured to control the at least one lateral guidance actuator with the manipulated variable. The lateral guidance actuator can be, for example, a steering actuator, for example an electric motor servo drive of a front-axle steering system and / or a rear-axle steering system. For a computer program according to the invention, it is provided that it comprises instructions which cause a control device to execute a method according to the invention. The control device can, in particular, be a control unit connectable to at least one lateral guidance actuator of a vehicle.

[0067] All advantages and embodiments described above in relation to the method according to the invention also apply to the control unit according to the invention, the vehicle according to the invention and the computer program according to the invention and vice versa.

[0068] Further advantages and details of the invention will become apparent from the exemplary embodiments described below and from the drawings. These are schematic representations and show:

[0069] Fig. 1 shows an embodiment of a motor vehicle according to the invention,

[0070] Fig. 2 is a block diagram of an embodiment of a method according to the invention for tracking a vehicle, and

[0071] Fig. 3 is a block diagram of several calibration functions of the embodiment of the method according to the invention.

[0072] Figure 1 shows an embodiment of a vehicle 1. The vehicle 1 can, for example, be a motor vehicle such as a passenger car or a truck. Furthermore, the vehicle 1 can also be a vehicle combination, which, for example, has a towing vehicle and one or more trailers. Alternatively, the method can also be used with other types of vehicles 1. The vehicle 1 comprises a control device 2, which is designed to carry out a method for lane guidance of the vehicle 1. Furthermore, the vehicle 1 comprises at least one lateral guidance actuator 3, via which the lateral position of the vehicle 1 can be adjusted.

[0073] The lateral guidance actuator 3 is designed, for example, as a steering actuator, by means of which a front axle steering angle of the vehicle 1 can be automatically adjusted. The steering actuator can be designed, for example, as an electric servomotor. Additionally or alternatively, the or another lateral guidance actuator 3 can also be a steering actuator for automatically adjusting a rear axle steering angle.

[0074] The control device 2 is configured to determine a control variable by means of a lateral control as a function of a lateral offset error describing a lateral offset 4 of the vehicle 1 relative to a predetermined target trajectory 5 (shown in dashed lines), and to adjust a lateral position of the vehicle 1 to the target trajectory 5 as a function of the control variable. For this purpose, the control device 2 controls the lateral guidance actuator 3 with the control variable. The lateral offset 4 describes, for example, the deviation between the target trajectory 5 related to the vehicle center of the vehicle 1 and the actual direction of movement 6 of the vehicle center in the vehicle's transverse direction.

[0075] The deviation from the target trajectory can be determined, as shown in Fig. 1, at a defined distance from the front edge of vehicle 1 (look-ahead length), but also at other reference points. Other possible reference points are, for example, at the height of the rear axle or the front axle of vehicle 1. The reference points are preferably located on the vehicle's longitudinal axis, but other spatially fixed reference points of vehicle 1 during the observation period are also conceivable.

[0076] Fig. 2 shows a block diagram of the exemplary embodiment of the method for automated lane guidance of the vehicle 1, implemented by the control device 2. The automatic lane guidance comprises a lateral control 7, which controls the lateral position of the vehicle 1 to maintain the desired trajectory 4. The lateral control 7 generates a manipulated variable delta_soll for the lateral guidance actuator 3, which can be controlled, for example, directly or via an intermediate steering angle controller (not shown).

[0077] The lateral control 7 comprises a trajectory follower controller 8 for generating the manipulated variable delta_soll, a curvature feedforward control 9, one or more disturbance compensators 10, for example for compensating for crosswind, road gradient and / or other effects, as well as a steering angle offset calibration 11. A manipulated variable component delta_controller is determined by means of the trajectory follower controller 8.

[0078] The curvature precontrol 9 determines a precontrol variable delta_vorst as a function of a measured and / or predicted curvature value of a road surface traveled by the vehicle 1, wherein the manipulated variable delta_soll contains the precontrol variable delta_vorst as a manipulated variable component. Accordingly, the one or more disturbance compensators 10 generate one or more manipulated variable components delta_komp, and a manipulated variable component delta_offset is generated via the steering angle offset calibration 11.

[0079] Furthermore, at least one calibration function 12 is provided, which determines a calibration value delta_calib. The calibration value delta_calib is applied to the manipulated variable delta_soll generated by the lateral control system 7. A signal processor 13 communicates a yaw rate g and a lateral acceleration a_lateral of the vehicle 1 as input variables to the at least one disturbance compensator 10. For this purpose, the signal processor 13 can evaluate measured values ​​recorded by various sensors of the vehicle 1, which describe a yaw rate and / or a lateral acceleration.

[0080] A position detection and trajectory planning unit 14 supplies a lateral offset error delta_y to the calibration function 12 and the trajectory follower controller 8. The lateral offset error delta_y describes the lateral offset 4 shown in Fig. 1 between the desired trajectory 5 and the actual direction of movement 6 or the actual position of the vehicle 1.

[0081] The position detection and trajectory planning unit 14 further provides the curvature pre-control unit 9 with at least one measured and / or predicted piece of curvature information kappa, which describes the curvature of the desired trajectory 5 in a section of the desired trajectory 5 located in front of the vehicle 1 in the direction of travel. The position detection and trajectory planning unit 14 determines the desired trajectory 5 and, accordingly, also the curvature information kappa as a function of environmental data provided by an environmental detection unit 15 of the vehicle 1. The environmental detection unit 15 generates the environmental data as a function of environmental measurement data, which is generated in particular by at least one environmental sensor (not shown) that detects at least a partial area of ​​the environment of the vehicle 1.

[0082] The functionality of blocks 8 to 15 can be implemented entirely or partially in the control device 7. The output variables of blocks 8 to 11 all represent components of the manipulated variable delta_soll. Furthermore, the manipulated variable delta_soll is subjected to at least one calibration value delta_kalib of the calibration function 12.

[0083] The individual components of the control variable delta_soll and the at least one calibration value delta_kalib have different effects on the lane guidance of the vehicle 1 , which are described below.

[0084] The target steering angle component delta_vorst of the curvature pre-control 9 is calculated, for example, as

[0085] (1 ) delta_vorst = kappa * (I + EG * v A 2), where the parameter I corresponds to the wheelbase and EG to the self-steering gradient of vehicle 1. The self-steering gradient depends on

[0086] (2) EG = m*(ch*lh-cv*lv) / (ch*cv*(lh+lv)) depends on the distance of the centre of gravity from a front axle Iv or a rear axle Ih of the vehicle 1, on the vehicle mass m of the vehicle 1 and on the cornering stiffnesses of the front tyres cv and rear ch.

[0087] If the curvature information kappa has an offset kappa_offs compared to the actual curvature due to an error, the pre-controlled target steering angle also contains an offset component delta_vorst_kappaoff, which is determined by

[0088] (3) delta_vorst_kappaoff = kappa_offs * (I + EG * v A2) is given. The consequence of the pilot control value delta_vorst which is faulty with delta_vorst_kappaoff is that this acts like a disturbance on the lateral control 7 and the vehicle 1 does not follow the target trajectory 5 as expected when a lane keeping assistance system is activated, but depending on the design of the trajectory following controller 8 - i.e. use of controller types with or without steady-state accuracy - has a permanent lateral offset 4 to the planned target trajectory 5 or at least causes a transient swerving before the vehicle 1 then follows the planned target trajectory 5 after a compensation process has subsided.

[0089] The degree of aberration of a camera used as an environment sensor, which supplies environment measurement data to the position detection and trajectory planning unit 14, generally varies over the operating life of the driver assistance system, since camera calibration routines are typically continuously executed. This means that, depending on the calibration status, the offset of the curvature feedforward control 9 and thus the degree of impairment of the vehicle's lateral guidance is time-variant, although the transmission properties of the camera and thus also the curvature offset usually do not change abruptly, but rather within a band-limited range.

[0090] It is possible that the steering angle is used as an auxiliary or controlled variable within a lane guidance driver assistance system. This steering angle can be influenced or corrected by the steering angle calibration routine 11. Due to the finite accuracy of the steering angle calibration routine 11, the resulting steering angle offset (delta_offset) typically contains an uncompensated component.

[0091] A remaining offset error delta_offset_error in the steering angle signal can have several causes. Typically, a steering angle calibration routine 11 is based on the signal from a yaw rate sensor of the vehicle 1 and / or a wheel speed sensor of the vehicle 1. Errors in these sensors, such as offset and linearity errors, can thus propagate into the determined steering angle offset. The effects of steering angle offsets are comparable to the case of a curvature offset described above. Here, too, when the lane guidance driver assistance system is activated, the vehicle 1 does not follow the target trajectory 5 as expected, but, depending on the design of the trajectory follower controller 8, exhibits a permanent or transient lateral offset 4 from the planned target trajectory 5, which can have adverse effects on the comfort and acceptance of the driver assistance system.

[0092] For example, if a PD controller is used as the trajectory follower controller 8 with the transfer function

[0093] (4) Gr(s) = Kp(v) + s* Kd(v) / (1 + s*T), with and the speed-dependent controller factors Kp(v) and Kd(v), the result is with the transverse deviation error delta_y describing the lateral offset 4 as input variable and delta_controller as output variable of the trajectory follower controller 8 (5) delta_controller(s) = Gr(s) * delta_y(s).

[0094] Consequently, the trajectory follower controller 8 will output the uncompensated steering angle offset delta_offset_error in the steady state of equilibrium when driving straight ahead, i.e., at a physical steering angle of zero degrees, and neglecting lateral disturbance forces on the vehicle 1. The lateral offset 4 required for this is calculated from

[0095] (6) Delta_y(0) * Gr(0) = Delta_offset_error(0) to

[0096] (7) delta_y = delta_offset_error / Kp(v) in the time domain. The controller factor Kp cannot be chosen arbitrarily large during the design process, which means that an uncompensated steering angle offset delta_offset_error not equal to zero always results in a lateral offset 4 of vehicle 1 relative to the target trajectory 5. An uncompensated offset of a rear-axle steering angle with rear-axle steering of vehicle 1 has a comparable effect on the lateral offset 4 of vehicle 1 as an uncompensated offset of the front-axle steering angle.

[0097] The sensor signals for the yaw rate, the front and rear axle steering angle, and also the lateral acceleration are used, in particular, to estimate disturbance forces and load moments acting on vehicle 1, for example, due to a road incline and / or crosswind acting on vehicle 1. The determined forces and moments then contribute proportionally to the steering angle setpoint delta_soll during the disturbance variable compensation process delta_komp.

[0098] Furthermore, these sensors can also be used to estimate the float angle of the

[0099] Vehicle 1 can be used, which can also be used as an auxiliary variable in the load estimations. In a steady-state situation, offset errors in the yaw rate g, the steering angle, or the lateral acceleration (ajateral) often have a proportional impact on the estimated variables and thus also on the disturbance feedforward. Consequently, offset errors in the yaw rate g, the steering angle, and also the lateral acceleration (ajateral) are propagated to the disturbance-compensating component of the steering angle setpoint (delta_soll). The result here, too, is an undesirable transient or permanent lateral offset of the track-guided vehicle 1 from the planned target trajectory 5.Disturbance compensators 10 for the yaw rate g and the lateral acceleration a ateral also generally have a finite accuracy, analogous to the steering angle compensation 11, and can therefore lag behind the real offset, whereby the previously described adverse effects on the driving comfort of the vehicle 1 when using the lane guidance driver assistance system can also occur in the presence of these.

[0100] These effects are addressed and at least partially compensated for by the at least one calibration function 12, resulting in improved ride comfort of the vehicle 1. The calibration function 12 is superimposed on the lateral control 7 and generates the at least one calibration value delta_kalib applied to the manipulated variable delta_soll. The calibration value delta_kalib is determined by the at least one calibration function 12 at least partially as a function of a temporal integration of the lateral offset error delta_y.

[0101] Depending on the sign and amplitude of the lateral deviation error delta_kalib, the calibration function 12 determines a calibration value delta_kalib for the target steering angle delta_soll, which, particularly with regard to the effective direction, is always aimed at reducing the lateral deviation error delta_y. This calibration value delta_kalib is then finally added to the manipulated variable delta_soll from the manipulated variable components of the trajectory follower controller 8, the curvature feedforward control 9, the one or more disturbance compensators 10, and the output of the steering angle offset calibration 11 and applied as the calibrated manipulated variable delta_soll_kalib, for example, to a target steering angle interface of the lateral control actuator 3. Alternatively, the calibrated manipulated variable can also be applied to a torque interface of the lateral control actuator 3.

[0102] The calibration function 12 can, for example, be implemented as an integrator or as an I-element with a comparatively small integration coefficient ki, so that the calibration function 12 integrates the cross-deviation error with a low dynamic range. Alternatively, the calibration function can also be implemented as a PT1 element or as another type of function with at least partially integrating behavior.

[0103] The determination of the calibration value delta_kalib by the calibration function 12 can therefore be carried out when the calibration function 12 is executed as an integrator with the integration coefficient ki, for example, as

[0104] (8) Delta_kalib = Delta_y * ki / s.

[0105] The time constant of the calibration function 12 is in particular greater than the dominant time constant of the lateral control 7 or the trajectory follower controller 8 by at least a factor of 2, at least a factor of 5, at least a factor of 10, at least a factor of 100 or at least a factor of 1000. The integration coefficient ki is for example in the order of magnitude of 0.001 degrees / (m*s), so that the gradient of the calibration value 12 is only 0.0002 degrees / s for a lateral deviation error delta_y of 0.2 m, assumed as an example for illustration.

[0106] The adaptation dynamics of the calibration value 12 are preferably dimensioned such that they are in the order of magnitude of the change dynamics of the offset errors described above. One measure of this is, for example, the largest expected temperature-related offset drift of the sensors used for lateral control 7, for example, converted to the corresponding steering angle level. If the dynamics were selected to be significantly larger, a typical overshoot would occur in the signal for delta_calib, since the calibration would then partially compete with the transient response of the compensation functions 10 for road inclination or crosswind. However, this could again limit the comfort gains achieved by the calibration function 12.

[0107] Calibration function 12 determines an integration value of calibration function 12 by integrating the lateral deviation error delta_y over time. This integration value represents the result of the temporal integration of the lateral deviation error delta_y at a specific point in time. The integration value is retained when lane guidance is deactivated and used as the starting value for the renewed temporal integration of the lateral deviation error delta_y when lane guidance is reactivated. The state of calibration function 12, implemented, for example, as an integrator, is therefore not reset during the current ignition cycle but is maintained even when the driver assistance system is repeatedly deactivated and activated, when the journey is ended, and / or when driver interventions override the automatic lane guidance.

[0108] The use of a constant integration coefficient ki in the calibration function 12 can represent a compromise for the entire speed range achievable by vehicle 1, since the influence of offset variables on the lateral deviation error delta_y can depend on the driving speed. This can be due, for example, to the commonly used controller parameter control of the trajectory follower controller 8 as a function of the driving speed. This also changes the disturbance compensation properties of the trajectory follower controller 8 and consequently also the effect of sensor offset variables on the lateral deviation error delta_y.

[0109] To account for this circumstance and expand the potential of the calibration function, the temporal integration of the lateral offset error delta_y can be performed with a weighting factor, in particular the integration coefficient ki, which depends on the vehicle speed. The integration coefficient ki can thus be made dependent on the current speed of vehicle 1 as ki(v). This advantageously results in an ideal, individually adaptable calibration dynamics of the calibration function 12 for each driving speed range.

[0110] Additionally or alternatively, the weighting factor or the integration coefficient ki can be provided to increase with increasing magnitude of the lateral deviation error delta_y. The calibration speed of the calibration function 12 can be effectively increased by additional nonlinear or progressive weighting of the magnitude of the lateral deviation error delta_y.

[0111] For example, the integration coefficient ki and thus the adaptation rate are switched between two or more values, with the integration coefficient ki being selected higher for larger absolute values ​​of the lateral offset error delta_y. For example, doubling the integration constant ki for a magnitude of the lateral offset error delta_y greater than 0.2 m can lead to a faster initial adaptation to an existing offset situation when starting the driver assistance function.

[0112] Even if the sensor compensators 10 and / or other sensor offset calibration functions in the vehicle 1 ultimately correct a detected offset drift of a sensor after a debounce time has elapsed by making a step-like correction of the respective offset, defining the adaptation rate of the calibration function 12 as a function of the magnitude of the lateral offset error delta_y can ensure faster adjustment to the new offset situation. This is particularly advantageous for comparatively long debounce times of the sensor compensators 10 or other sensor offset calibration functions, since during such a debounce time, the calibration function 12 increasingly already performs a compensation, which must be reintegrated after a successful signal update of the sensor compensators or sensor offset calibration functions.

[0113] Since step-like corrections of a sensor offset immediately lead to a lateral

[0114] Jerk of the vehicle 1, which depends on the jump height, it can be provided as an advantageous extension that all sensor offset calibration functions in the vehicle 1 make a change to the respective sensor offset not abruptly, but only in a band-limited manner. If, for example, the offset compensations for the vehicle sensors occur at most with a defined maximum gradient, which the output of the centering calibration delta_kalib can always follow with a small following error in the order of 0.03 degrees, then the non-linear or progressive weighting of the amount of the lateral deviation error delta_y in the calibration function 12 can also be dispensed with as a measure to increase the dynamics.

[0115] The contribution of delta_kalib to the total manipulated variable delta_soll_kalib is advantageously further limited in its maximum amplitude by limiting the integrator state in the calibration function 12. For this purpose, it can be provided that the temporal integration of the cross-alignment error delta_y is stopped by the calibration function 12 if the or an integration value of the calibration function 12 determined by the temporal integration of the cross-alignment error delta_y corresponds to a predetermined limit value.

[0116] The limit value or maximum amplitude of the integrator or calibration function 12 is based, for example, on the sum of the maximum influences of all offset variables on the steering angle setpoint delta_soll and the actual steering angle. For example, if delta_komp, delta_vorst, and delta_offset each exhibit a maximum offset error of 0.05 degrees, then delta_kalib must be limited to an interval of + / - 0.15 degrees. The angle specifications refer to the vehicle's steering angle, for example, a steering angle at the vehicle's front wheels. This approach also advantageously eliminates the need for the otherwise necessary measures to avoid integrator windup effects in calibration function 12.

[0117] As an alternative to limiting the integrator state, a low-pass filter function, such as a PT-1 filter, can be used. This can also prevent windup effects. While the offset situation cannot be completely eliminated, the achieved centner accuracy may still be sufficient.

[0118] When cornering, the lateral deviation error due to a feedforward control error depends not only on an offset error of the curvature, but also on the inaccuracy of the stored vehicle parameters, in particular on uncertainties of the self-steering gradient EG. The proportion of incorrect feedforward control due to errors of the self-steering gradient EG is obtained by applying the total differential to equation (1 ) and evaluating EG or its error EG_err as

[0119] (8) delta_vorst_EG_err = (d delta_vorst / d EG) * EG_err

[0120] = (d(kappa * (I + EG * v A 2)) / d EC ) * EC_err

[0121] = kappa * v A 2 * EG_err

[0122] Contributions from delta_vorst_EG_err have the effect of a disturbance on the lateral control of vehicle 1 and proportionally cause a lateral offset error delta_y, which leads to an erroneous deflection of delta_kalib. The delta_kalib signal therefore reacts not only to sensor offset variables. However, as long as EG_err is limited, the effect of EG_err on the vehicle plane is only minimal due to the large integration time constant of the centering calibration. The integration time constant is defined as the inverse of the integration coefficient ki.

[0123] This is also true because any incorrect deflection of delta_kalib in curves is corrected on the straight-ahead sections of the road. If EG_err varies significantly, the EG value in the vehicle can be continuously adjusted to the current conditions, such as vehicle mass m or tire stiffness, using an adaptation algorithm. This reduces EG_err and thus also the contribution of delta_vorst_EG_err.

[0124] Furthermore, with a larger variation of EG_err, the calibration function 12 can be used for high values ​​for the product kappa * v A 2. It can therefore be provided that the temporal integration of the lateral deviation error delta_y is stopped by the calibration function 12 when the product of the square of the current vehicle speed v and the curvature value kappa of the desired trajectory 5 and / or the curvature value of the roadway exceeds a predetermined threshold value.

[0125] For this purpose, the curvature kappa is advantageously first low-pass filtered, for example, using the PT-1 algorithm, and then the absolute value is subjected to a threshold comparison dependent on the driving speed v. If the filtered curvature is above this threshold, the calibration function 12 is temporarily stopped.

[0126] Furthermore, it can be provided that the temporal integration of the lateral deviation error delta_y is stopped by the calibration function 12 if the or a measured and / or predicted curvature value kappa of the or a roadway traveled by the vehicle 1 exceeds a predetermined threshold value and / or if a curvature value of the desired trajectory 5 exceeds a predetermined threshold value.

[0127] If the curvature itself temporarily has a larger offset error until the camera calibration has compensated for this error, this can temporarily reduce the calibration activity of calibration function 12 for higher driving speeds. However, the adaptation range only shifts by the amount of the curvature offset, so that the calibration activity does not completely cease. Furthermore, when the driving speed v is reduced and the product kappa * v is thus reduced, A 2 the calibration is carried out again with increased intensity.

[0128] Calibration disturbances can also be caused, in particular, by the intervening driver, who can directly influence the lateral deviation error delta_y, for example, by steering the steering wheel of vehicle 1. To remedy this, the calibration function 12, i.e. the integration of the lateral deviation error delta_y, can be stopped when the driver intervenes at the steering wheel. In particular, it is possible for the temporal integration of the lateral deviation error delta_y to be stopped and / or for the application of the calibration value delta_calib to the manipulated variable delta_soll to be suspended if a steering intervention by a driver of vehicle 1 generates a steering torque M above a predetermined threshold value and / or a steering torque M that increases the current lateral deviation error delta_y.

[0129] However, higher availability and thus faster convergence of the calibration function 12 can be achieved by stopping the calibration function 12 only when the driver steers in the direction of increasing the lateral deviation error delta_y. This can be done by comparing the signs of the driver steering torque M and the lateral deviation error delta_y. For example, if the counting arrow of the driver steering torque M is defined such that steering to the left in the direction of travel results in a positive torque and a positive lateral deviation error means the positioning of the vehicle to the right of the planned target trajectory 5 in the direction of travel, then the signs of the driver steering torque M and the lateral deviation error must be different as a stop condition for the calibration function.

[0130] The integration coefficient ki is initially set as a basic design for the execution of the lane assistance for driver steering interventions below a defined intensity, evaluated, for example, based on the level of the measured driver steering torque M. Additionally or alternatively, it can be provided that when a driver of the vehicle 1 intervenes in the steering and generates a steering torque M that reduces the current lateral deviation error delta_y, the temporal integration of the lateral deviation error delta_y is carried out with an increased weighting, in particular with an increased integration coefficient, at least for a predetermined period of time. As a result, the integration coefficient can be specifically increased compared to the basic design if a driver intervention is detected and the driver simultaneously steers in the direction of reducing the lateral deviation error delta_y.The driver's steering can be regarded as a confirmation of the calibration process of the calibration function 12, which justifies an increase in the calibration dynamics by increasing the integration coefficient ki.

[0131] As described, the effect of offset variables on the lateral deviation error delta_y is partially dependent on the driving speed v. This applies, for example, to the offset errors contained in the output of the disturbance compensators delta_komp or the offset error that is part of the curvature feedforward control delta_vorst. When the driving speed changes, delta_kalib must therefore always change as well in order to keep the lateral deviation error low, even though the offset variables causing the lateral deviation error, namely yaw rate g, lateral acceleration a ateral and feedforward curvature kappa, are assumed to be constant in the analysis. Due to the high integration time constant of the calibration function 12 and thus the low dynamics of the calibration function 12, the adaptation to the changed driving speed v takes a relatively long time, during which time the accuracy of the tracking is reduced.

[0132] To take this circumstance into account, it is advantageous to use an arrangement of several integrators or memories for the entire range of vehicle speed instead of a single integrator as memory for delta_kalib, as shown, for example, in Fig. 3.

[0133] Fig. 3 illustrates the use of several integration functions 12_1 to 12_N. Furthermore, an interpolation unit 16, an initialization unit 17, an integration coefficient determination unit 18, and a sequence control unit 19 are shown as part of the control device 2.

[0134] The integrating functions 12_1 to 12_N are each assigned a different speed interval, with the integrating functions 12_1 to 12_N each integrating the lateral offset error delta_y over time at a speed of vehicle 1 within the speed interval assigned to them. The entire vehicle speed range is thus subdivided into N > 1 subintervals, each of which covers only a limited speed range and to which a calibration function 12 or a dedicated integrator is assigned. Thus, only a single calibration function 12 is active at any one time, while the remaining calibration functions 12 are paused or have an input signal from zero applied to their inputs.The respectively active calibration function (calibration function 12_1 to 12_N) is supplied with the weighted lateral deviation error delta_y, and each integrator generates its own delta_kalib_v(k), k = [1...N]. The calibration value delta_kalib_v(1) to delta_kalib_v(N) of the respective calibration function 12_1 to 12_N, whose speed interval includes the current speed of vehicle 1, can be applied directly to the control variable delta_soll as the resulting delta_kalib. The weighting of the lateral deviation error delta_y can be carried out using a constant integration coefficient ki or via a factor ki(v) dependent on the driving speed.

[0135] The integration coefficient ki or ki(v) can be determined for the individual calibration functions 12_1 to 12_N by the integration coefficient determination unit 18. For this purpose, the integration coefficient determination device 18 can receive, for example, the vehicle speed v, the steering torque M of a driver of the vehicle 1, the current lateral deviation error delta_y, and the measured or predicted curvature kappa as input variables.

[0136] The centers of the N driving speed intervals essentially represent the support points for the output values ​​delta_kalib_v(k). In order to obtain a resulting delta_kalib from the individual contributions delta_kalib_v(k), an overall calibration value delta_kalib_ges determined by the interpolation unit 16 from the calibration values ​​delta_kalib_v(1) to delta_kalib_v(N) can additionally or alternatively be applied to the manipulated variable delta_soll. The overall calibration value delta_kalib_ges is determined from two or more of the calibration values ​​delta_kalib_v(1) to delta_kalib_v(N) generated by the integration functions 12, in particular in such a way that the overall calibration value delta_kalib_ges is temporally continuous even when switching between two or more of the speed intervals. This can be implemented by the interpolation unit 16, for example, via linear interpolation, wherein the interpolation unit receives the vehicle speed v as input variable.

[0137] For this purpose, for example, depending on the two distances of the current driving speed to the two centers of neighboring speed ranges, delta_calib_ges is linearly combined from delta_calib_v(k) and delta_calib_v(k+1).

[0138] (9) delta_kalib_ges = delta_kalib_v(k) * a + delta_kalib_v(k+1 ) * (1-a) with a = (vv(k)) / (v(k+1 ) - v(k)), v > v(k), v < v(k+1 )

[0139] An interpolation between the individual integrator states delta_kalib_v(k) is also advantageous because there is no jump in delta_kalib_ges when transitioning from one to another speed range.

[0140] Jumps in the steering angle setpoint delta_soll_kalib lead to a lateral jerk of vehicle 1 and should be avoided for reasons of driving comfort. The advantage of using a bank of calibration functions 12_1 to 12_N or a bank of integrators over using a single integrator is particularly evident during acceleration or deceleration, and thus the frequent changes in speed ranges.

[0141] If the driver assistance function is operated predominantly within a narrow speed range, it may occur that only one of the calibration functions 12 is loaded according to the lateral offset error delta_y, so that only this calibration function 12 correctly maps the current offset situation. Due to the typical drift of the offset variables for curvature kappa, yaw rate gun, and lateral acceleration a aterial, the states of the remaining integrators become increasingly less up-to-date. Upon entering a neighboring speed range, a larger lateral offset error delta_y and, at the same time, reduced driving comfort could temporarily occur until adaptation to the current offset situation is complete. As a remedy, a forgetting factor can advantageously be implemented.For this purpose, it can be provided that the calibration functions 12_1 to 12_N each determine an integration value of the calibration function 12_1 to 12_N assigned to the calibration function 12_1 to 12_N by the temporal integration of the lateral deviation error, wherein the integration values ​​of the one or more integration functions in which the current speed lies outside the respective speed interval are continuously adapted, in particular at a constant or a vehicle speed-dependent rate, to the integration value of that calibration function 12_1 to 12_N in whose speed interval the current speed of the vehicle 1 lies.

[0142] For example, all integrator states or integration values ​​of the inactive calibration functions 12_1 to 12_N strive for the state or integration value of the currently active calibration function 12_1 to 12_N with a uniform gradient or a gradient individually defined for the respective speed range. This gradient can, for example, be based on the change dynamics of the offset variables. In the application, it can advantageously be dimensioned such that it corresponds to 25% of the resulting gradient for delta_kalib_v(k) for a lateral offset error of 0.2 m. If the adjusted gradient is selected too high, the advantage of using a bank of calibration functions 12_1 to 12_N compared to using a single calibration function 12 diminishes.

[0143] As an alternative to the gradient method, the calibration functions of the currently irrelevant speed ranges can be continued with integration coefficients reduced to, for example, 10% to 20%. With this approach, exceeding the integrator states beyond the integrator state of the currently active speed range is avoided by appropriate limitations. Depending on the dynamics of the offset influences, it may be advantageous to measure the state of a single calibration function 12 or all states of a plurality of

[0144] Calibration functions 12_1 to 12_N in a non-volatile memory, for example a memory of the control device 3, in order to

[0145] Ignition restart of vehicle 1 to be able to set the already calibrated states of one or more calibration functions 12 or 12_1 to 12_N.

[0146] This is particularly advantageous when the majority of offset errors have no or only a small temporal drift.

[0147] The storage of the states or the integration values ​​of the one or more calibration functions 12_1 to 12_N can be carried out by the initialization unit 17. The initialization unit 17 can also carry out the initialization of the

[0148] Calibration functions 12_1 to 12_N and performing a status comparison between the calibration functions 12_1 to 12_N or their current

[0149] Perform integration values.

[0150] The sequence control unit 19 can, for example, track an activation and / or a deactivation of the automatic lane guidance, the occurrence of a condition for stopping a temporal integration and / or for applying the at least one calibration value delta_kalib to the manipulated variable delta_soll, the change between the speed intervals and / or other events and accordingly cause an operation of the at least one calibration function 12.

[0151] In its basic function, the calibration function 12 can also be viewed as an integral component in the trajectory follower controller 8, but preferably has the essential properties or the essential differences that the integration coefficient or another weighting factor used for the temporal integration of the lateral deviation error delta_y is very low, contrary to the design criteria typical for a lateral control 7, and / or that the calibration function 12 or the temporal integration of the lateral deviation error delta_y is not initialized with zero when the assistance system is restarted, but rather starts from the last achieved state or adopts it. The advantage of this is that the vehicle 1 has virtually no lateral transient response when the driver assistance function is activated, but can travel directly along the planned target trajectory 5.The previously described extensions of the at least one calibration function 12 also represent further differences to a classic integral component of a lateral control controller.

[0152] The at least one calibration function 12 can be implemented as a separate function or, alternatively, in addition to an existing integral component in the trajectory follower controller 8. The at least one calibration function 12 will, in particular, only perform an adaptation during the activation phases of the driver assistance system until the integral component of the trajectory follower controller 8 has reduced the lateral deviation error delta_y to zero, or whenever the integral component of the trajectory follower controller 8 is reset and needs to be rebuilt. This can occur, for example, for a Level 2 autonomous system when the driver turns the steering wheel.

[0153] In general, the calibration dynamics of the calibration function 12, i.e., the time required until an adjustment of the lateral offset 4 of the vehicle 1 to the current offset situation has taken place, can be greater than with a driver assistance system without an integral component in the trajectory follower controller 8. This circumstance can be at least partially counteracted by increasing the integration coefficient(s) when using the one or more calibration functions 12 with a trajectory follower controller s without an integral component. Furthermore, when using an integral component in the trajectory follower controller 8, one or more of the disturbance compensations 10 for crosswind and / or road inclination can be omitted.This has the consequence that a significant source of offset errors is also eliminated and thus fewer offset effects have to be taken into account by the calibration function 12, which advantageously results in lower requirements on the calibration dynamics of the at least one calibration function 12.

Claims

Patent claims 1. Method for tracking a vehicle (1), wherein a manipulated variable is determined by a lateral control (7) as a function of a lateral offset (4) of the vehicle (1) relative to a predetermined desired trajectory (5), wherein a lateral position of the vehicle (1) is adjusted to the desired trajectory (5) as a function of the manipulated variable, characterized in that the manipulated variable is additionally subjected to at least one calibration value, wherein the calibration value is determined by at least one calibration function (12) at least partially as a function of a temporal integration of the lateral offset error.

2. Method according to claim 1, characterized in that the time constant of the calibration function (12) is greater, in particular at least by a factor of 2, than the dominant time constant of the transverse control (7).

3. Method according to claim 1 or 2, characterized in that an integration value of the calibration function (12) is determined by the temporal integration of the lateral deviation error, wherein the integration value is retained when the track guidance is deactivated and is used as the starting value of the temporal integration of the lateral deviation error when the track guidance is reactivated.

4. Method according to one of the preceding claims, characterized in that the lateral control (7) comprises a pilot control (9), wherein the pilot control (9) is dependent on a measured and / or predicted curvature value of a road travelled by the vehicle (1). A pilot control variable is determined from the road surface, wherein the manipulated variable contains the pilot control variable as a component. Method according to claim 4, characterized in that the pilot control is carried out as a function of a vehicle model describing a self-steering gradient of the vehicle, wherein the self-steering gradient is adapted during operation of the vehicle as a function of at least one vehicle parameter, in particular a vehicle mass and / or a tire stiffness. Method according to one of the preceding claims, characterized in that the temporal integration of the lateral deviation error is carried out with a weighting factor, in particular an integration coefficient, which depends on a speed of the vehicle (1) and / or increases with an increasing amount of the lateral deviation error. Method according to one of the preceding claims, characterized in that the temporal integration of the lateral deviation error is stopped by the calibration function (12) when: - the or an integration value of the calibration function (12) determined by the temporal integration of the cross-deviation error corresponds to a predetermined limit value, - the or a measured and / or predicted curvature value of the or a roadway travelled by the vehicle (1) exceeds a predetermined threshold value, - a curvature value of the target trajectory (5) exceeds a predetermined threshold value, - the product of the square of the current vehicle speed and the curvature of the road exceeds a predetermined threshold, and / or the product of the square of the current vehicle speed and the curvature value of the target trajectory (5) exceeds a predetermined threshold value.

8. Method according to one of the preceding claims, characterized in that the temporal integration of the lateral deviation error is stopped and / or the application of the calibration value to the manipulated variable is suspended if a steering torque above a predetermined threshold value and / or a steering torque which increases the current lateral deviation error is generated by a steering intervention by a driver of the vehicle (1).

9. Method according to one of the preceding claims, characterized in that when a driver of the vehicle (1) intervenes in the steering, which generates a steering torque that reduces the current lateral deviation error, the temporal integration of the lateral deviation error is carried out at least for a predetermined period of time with an increased weighting, in particular with an increased integration coefficient.

10. Method according to one of the preceding claims, characterized in that a plurality of integrating functions (12) are used, wherein the integrating functions (12) are each assigned a different speed interval, wherein the integrating functions (12) each integrate the transverse offset error over time at a speed of the vehicle (1) lying within the speed interval assigned to them, wherein the calibration value of the integrating function (12), whose speed interval includes the current speed of the vehicle (1), or a total calibration value determined as a function of this calibration value is applied to the manipulated variable. Method according to claim 10, characterized in that the total calibration value is determined from two or more of the calibration values ​​generated by the integrating functions (12) in such a way that the total calibration value is temporally constant even when changing between two or more of the speed intervals.Method according to claim 10 or 11, characterized in that an integration value of the calibration function (12) assigned to the calibration function (12) is determined by the calibration functions (12) in each case by the temporal integration of the lateral deviation error, wherein the integration values ​​of the one or more integration functions for which the current speed lies outside the respective speed interval are continuously adapted, in particular at a constant or a vehicle speed-dependent rate, to the integration value of the calibration function (12) in whose speed interval the current speed of the vehicle (1) lies. Method according to one of the preceding claims, characterized in that the at least one calibration function (12) is in each case designed as an I element or as a PT1 element.Control unit for generating a control variable for at least one transverse guidance actuator (3) of a vehicle (1), wherein the control unit (2) is configured to carry out a method according to one of the preceding claims. The vehicle comprises at least one transverse guidance actuator (3) and a control unit (2) according to claim 14, wherein the control unit (2) is configured to. Control of the at least one transverse guide actuator (3) with the manipulated variable. Computer program comprising instructions which cause a control device to carry out a method according to one of the Claims 1 to 14.

Images