Method for operating an autonomous driving function of a vehicle

Abstract

The invention relates to a method for operating an autonomous driving function of a vehicle (1), wherein the vehicle (1) has a computer unit (30) and a plurality of sensors (10) for sensing environmental data, and wherein the computer unit (30) is provided to determine a desired trajectory (2) from the sensed environmental data along which the vehicle (1) is to be driven. Furthermore, it is provided that a sensing of an actual trajectory (4) is carried out in a first step a), wherein the actual trajectory (4) depicts a path actually traveled by the vehicle (1), and a sensing of a distance to an object in the environment of the vehicle (1) is carried out. In a subsequent step b), a derivation of a quality of the autonomous driving function is carried out by comparing the actual trajectory (4) with the desired trajectory (2) and monitoring the sensed distance to the object in the environment. In a subsequent step c) of the method, an adjustment of the quality to a predefined target value is carried out by selecting sensors (10) from the plurality of sensors (10) to be used for the autonomous driving function and / or by changing a measurement rate at which a measurement from at least one sensor (10) of the plurality of sensors (10) is carried out. Further aspects of the invention relate to a computer program and a vehicle (1) which is provided to carry out the method.

Description

TECHNICAL FIELD

[0001] The present invention relates to a method for operating an autonomous driving function of a vehicle, wherein the vehicle has a computer unit and at least two sensors for sensing environmental data, and wherein the computer unit is configured to determine a desired trajectory from the sensed environmental data along which the vehicle is to be guided. Further aspects of the invention relate to a computer program and a vehicle which are configured to implement the method. BACKGROUND

[0002] Some vehicles with autonomous driving functions are known from the prior art with which the vehicle can be driven from a starting position to a target position along a trajectory without driver intervention. Here, the trajectory describes a track curve along which the vehicle moves. Such autonomous driving functions can relieve the driver of the vehicle in that the vehicle moves automatically in a defined situation or during the entire journey.

[0003] In order to provide such an autonomous driving function, precise environmental data are necessary which describe the environment of the vehicle. In order to sense these environmental data, a plurality of sensors are used. It is problematic here that, in order to process the environmental data, considerable computer resources are necessary.

[0004] DE 10 2012 108 543 A1 describes a method for adapting the environmental sensing of a vehicle on the basis of information from a digital map or traffic information. The traffic information can be, for example, a message about a construction site. In the method, it is known, for example, from the digital map whether the vehicle is located in an urban environment or on a motorway. Furthermore, data about the topology of the vehicle's environment can be known from the map. Then, the sensor system of the vehicle is regulated in accordance with these information in order to sense the surroundings. Here, sensors can be activated or deactivated and the active distance, the resolution, the sampling rate and / or the sensing area can be adjusted. By adapting the environmental sensing to the environmental conditions, computation time can be saved and the accuracy and efficiency of the environmental sensing can be improved.

[0005] DE 10 2017 114 049 A1 describes a system for determining and adapting a route for an autonomous vehicle. The vehicle comprises a sensor subsystem with different sensor systems such as cameras or radars. Furthermore, the system comprises submodules which carry out different subtasks. Here, the system comprises a vehicle perception submodule which implements a system function for perception. By means of the modules of the system, it is decided which sensors are to be used, how the sensors are to be used and how the sensor data is to be processed.

[0006] DE 10 2013 219 567 A1 describes a method for controlling a micro mirror scanner. In the method, it is provided to control the micro mirror scanner depending on a signal of a further sensor. If, for example, a particularly interesting object is found, it is finely scanned. The micro mirror scanner is also controlled depending on the weather situation, wherein, for example, the parameters of the scanner are optimized in the case of fog.

[0007] The methods known from the prior art for adjusting the operation of sensors for sensing environmental data only take into account certain partial aspects respectively, so that it is not ensured that in each case such environmental data is available for use which is necessary for the safe operation of an autonomous driving function. SUMMARY

[0008] A method for operating an autonomous driving function of a vehicle is proposed, wherein the vehicle has a computer unit and a plurality of sensors for sensing environmental data, and wherein the computer unit is provided for determining a desired trajectory from the sensed environmental data along which the vehicle is to be guided.

[0009] Furthermore, it is provided that in a first step a) an actual trajectory is sensed, wherein the actual trajectory depicts the path actually traveled by the vehicle and the distances to objects in the surroundings of the vehicle are sensed.

[0010] In a next step b), the quality of the autonomous driving function is ascertained by comparing the actual trajectory with the desired trajectory and the sensed distances to the objects in the surroundings.

[0011] In a next step c) of the method, the quality is adjusted to a predefined target value by selecting the sensors to be used for the autonomous driving function from the plurality of sensors and / or by changing the measurement rate at which at least one of the plurality of sensors measures.

[0012] Preferably, the steps a) to c) of the method are repeatedly performed during the operation of the autonomous driving function. The method can be repeated, for example, after the expiry of a time period in the range of 1 ms to 10 s, preferably in the range of 10 ms to 2 s, particularly preferably in the range of 100 ms to 1 s. In particular when the method is implemented regularly in the ms range, changes in the quality can be reacted to quickly, so that at any time when the autonomous driving function is implemented, driving safety is ensured.

[0013] The autonomous driving function is configured to guide the vehicle from a starting position to a target position without driver intervention. Preferably, the autonomous driving function is not limited here to a defined driving maneuver or situation and can also assume, inter alia, route planning to the target position. However, the autonomous driving function can also relate to vehicle guidance in defined situations, such as, for example, in a construction site or along a previously learned trajectory. During operation of the autonomous driving function, the vehicle is guided along a desired trajectory, which is determined by a calculator unit. The calculator unit requires environmental data provided by a plurality of sensors here.

[0014] The sensors used can relate, inter alia, to ultrasonic sensors, optical cameras, such as video cameras and infrared cameras, radar sensors and lidar sensors. Preferably, the plurality of sensors comprises different types of sensors here, wherein a plurality of samples can be used in one type of sensor. For example, the vehicle can comprise a plurality of ultrasonic sensors, a plurality of video cameras, one radar sensor and one lidar sensor.

[0015] In order to determine the actual trajectory, it is preferably provided that the position of the vehicle is determined continuously. The position determination of the vehicle takes place, for example, by using a satellite navigation system, by evaluating landmarks and / or by evaluating radio signals. It is particularly preferred that the various possibilities for position determination are combined with one another in order to increase the accuracy.

[0016] The quality of the autonomous driving function indicates how precisely the vehicle is guided. According to the application, the quality of the autonomous driving function is determined in such a way that the desired trajectory along which the vehicle is guided is compared with the actual trajectory actually traveled by the vehicle. The quality is thus an indication of "how precisely the vehicle has traveled the defined desired trajectory taking into account the currently requested environmental data". Furthermore, the quality of the autonomous driving function is derived by monitoring the distance to objects in the surroundings, such as further road users or fixed structures, such as trees, walls and pillars. If the distance to the objects in the surroundings is too small, this is an indication that the quality of the autonomous driving function is currently too low. In order to determine whether the distance is too small, a minimum distance value can be predefined, for example. Here, it can be provided that different minimum distance values are predefined for different categories or types of objects. For example, a greater minimum distance value can be predefined for movable objects, such as other road users, than for static objects, such as trees. When monitoring the distance, it is preferably additionally taken into account that the vehicle must at all times stand on the road or on a passable area. This condition must be fulfilled, even if the distance to the further object can be below the predefined minimum distance within a tolerance.

[0017] The target quality value is preferably predetermined such that it ensures the autonomous driving function is safely implemented within permissible tolerances at all times. This means that, given the target quality value, the vehicle follows the intended trajectory with predetermined accuracy and adheres to a predetermined minimum distance from objects in the vehicle's surrounding environment. Simultaneously, the target quality value is chosen to be smaller than the maximum achievable quality. This reduces the amount of environmental data processed by the computer unit compared to processing all available environmental data. Consequently, the load on the computer unit is reduced, and therefore, the energy requirements of the computer unit are significantly decreased.

[0018] In principle, known adjustment methods can be used to adjust the quality. Therefore, for example, it is conceivable to adjust the quality to a pre-defined target value, pre-defined limit values ​​for the quality, and to decrease the number of selected sensors when the quality is above the limit value, and increase the number of selected sensors when the quality is below the limit value, and / or decrease the measurement rate of at least one sensor when the quality is above the limit value, and increase the measurement rate when the quality is below the limit value. Hysteresis can also be set.

[0019] Preferably, a statistical model is used when adjusting the quality to the target value, wherein the statistical model can consider other input values ​​besides quality during adjustment.

[0020] Preferably, at least one evaluation factor is determined taking into account at least one additional parameter, and this at least one evaluation factor is considered when selecting the sensor to be used and / or when changing the measurement rate, wherein the at least one additional parameter is selected from: the load of the calculator unit, the traffic conditions in which the vehicle is located, the scene in which the vehicle is located, and / or data about the weather at the vehicle location.

[0021] Preferably, efforts are made in the calculator unit to keep its load below a predetermined value, ensuring that sufficient resources are always available for processing environmental data. For example, the load can be predetermined to be kept below 80%, if possible. Furthermore, unnecessarily high loads on the calculator unit lead to unnecessarily high energy consumption and should therefore be avoided.

[0022] In traffic conditions, one might consider, for example, whether there is heavy traffic, i.e., whether a large number of moving objects must be reliably sensed and considered when creating the desired trajectory, or whether there is only light traffic, i.e., only a small number of moving objects in the surrounding environment need to be sensed. Accordingly, it may be necessary to select more sensors and / or operate these sensors at a higher measurement rate in traffic conditions with heavy traffic than in cases of light traffic, in order to achieve the target values ​​for quality. Furthermore, certain sensors are better suited to certain traffic conditions than others due to their type and / or due to their location on the vehicle, which is preferably considered when selecting sensors and / or selecting measurement rates.

[0023] The vehicle's location can be in environments such as urban areas, rural areas, or highways. Advantageously, different types of sensors contribute differently in terms of importance to the quality of autonomous driving capabilities, depending on the scenario, and / or the measurement rates of individual sensors can vary depending on the sensor type and its installation location on the vehicle. To determine the scenario, environmental data from the sensors can be analyzed and evaluated, and / or digital maps can be used.

[0024] Weather conditions also influence how strongly a particular type or type of sensor contributes to the quality of autonomous driving functionality. If a vehicle with autonomous driving capabilities is on the road, for example, in good weather, it is possible that data from the video camera positioned at the front of the vehicle is sufficient to achieve the target quality. The required trajectory, provided by the calculator unit, can then be determined solely using the video camera measurement data requested at a sufficient measurement rate to achieve the desired quality. Here, it is conceivable that no additional sensors are selected, and correspondingly, the calculator unit does not request measurement data from redundant sensors, such as radar sensors. Conversely, in poor weather conditions, it is conceivable that, for example, only measurement data from a lidar sensor is requested, and measurement data from the video camera is not requested. For example, weather data can be obtained from the vehicle's sensors and / or retrieved from a weather service, relating to the location.

[0025] Preferably, for each individual sensor among multiple sensors, individual evaluation factors are determined. These individual evaluation factors allow the characteristics of each sensor to be considered when adjusting quality. This ensures that the corresponding contribution of each sensor to the quality of autonomous driving function is evaluated during adjustment, and that those sensors currently making the greatest contribution to quality are preferably selected.

[0026] Preferably, the unselected sensors are either turned off or placed in standby mode. In the off state or in standby mode, the energy reception of the corresponding sensor decreases and environmental data is not acquired by that sensor.

[0027] In this method, additional or alternative possible measurement rate adaptations also advantageously affect energy reception. By selecting the measurement rate in this way, environmental data is obtained only at the rate precisely necessary to achieve the required quality within tolerances, thus reducing the amount of environmental data relative to the maximum possible amount. Here, the tolerances for quality include factors necessary for the safe operation of the autonomous vehicle at all times. Therefore, less environmental data must be processed by the computer unit, which in turn advantageously affects the load on the computer unit and thus its energy reception.

[0028] Here, the measurement rate can be adapted in two directions: starting from the standard value, the measurement rate can be increased when the quality is too low, and the measurement rate can be decreased when the quality exceeds the required target value.

[0029] Preferably, the processing rate of the calculator unit for environmental data is equivalent to the measurement rate and adapts accordingly when the measurement rate changes. Alternatively, it is preferable to keep the processing rate of the environmental data constant and transmit a single measurement value multiple times when the measurement rate decreases. If the processing rate is, for example, 100 Hz and the sensor's measurement rate is adapted to 50 Hz, each measurement value is transmitted to the calculator unit twice consecutively for processing. Alternatively, it is also possible to transmit a prompt for "missing measurement value or corresponding sensor shutdown" instead of the measurement value, so that the calculator unit does not consider that measurement value as an actual measurement value during processing. Especially when completely shut down or when the sensor has been switched to standby, it is preferable to transmit such a shutdown value.

[0030] Preferably, the desired trajectory is determined by using a first artificial intelligence model, which has been obtained through machine learning. The first artificial intelligence model is specifically constructed and configured to provide autonomous driving capabilities.

[0031] Preferably, the selection of the sensor to be used and / or the change of the measurement rate are achieved through the use of a second artificial intelligence model obtained via machine learning. Preferably, the second model transmits environmental data from the sensor to the first model, wherein the output data of the second artificial intelligence model is preferably transmitted in an output data format standardized for the first artificial intelligence model.

[0032] Particularly preferably, not only should the determination of the trajectory be carried out, but also the selection of the sensors to be used and / or the change of the measurement rate should be carried out by using a common artificial intelligence model, which is obtained through machine learning.

[0033] One or more artificial intelligence models may involve, for example, models for machine learning, such as deep neutral network (DNN), Bayesian machine learning models, or the like. In principle, such methods and models for machine learning, as well as methods and models for training these models, are known to those skilled in the art.

[0034] Preferably, a second artificial intelligence model or a common artificial intelligence model for selecting the sensors to be used and / or changing the measurement rate is learned using training data. This is especially true when using neural networks. When learning with training data, data recorded during driving is preferably used. In particular, the determined vehicle position and environmental data obtained by these sensors at the corresponding sensors' maximum measurement rates are used as input data, wherein a target value for the quality of autonomous driving functions is input as a learning objective. Furthermore, data regarding the scene in which the vehicle is located, traffic density during the recorded driving period, and / or weather conditions during the recorded driving period can be used as input data.

[0035] If environmental data obtained at a reduced measurement rate is needed for training an artificial intelligence model, it is preferable to downsample the environmental data in the database based on existing environmental data recorded at the sensor's maximum measurement rate. This can be achieved in real time in the test environment using virtual sensors (e.g., Hardware in the Loop, HIL).

[0036] According to the invention, a computer program is further provided that, when implemented on a programmable computer device, performs one of the methods described in this regard. The computer program may, for example, relate to a module for implementing autonomous driving functions or subsystems thereof in a vehicle. The computer program may be stored on a machine-readable storage medium, such as a permanent or rewritable storage medium, or stored adjoining to the computer device, or stored on a removable CD-ROM, DVD, Blu-ray disc, or USB flash drive. Additionally or alternatively, the computer program may be made available for download on the computer device, for example, on a server, via a data network, such as the Internet, or a communication connection, such as a telephone line, or a wireless connection.

[0037] According to the present invention, a vehicle is further provided. The vehicle includes a calculator unit and at least two sensors for sensing environmental data, wherein the calculator unit is configured to provide autonomous driving functionality. Furthermore, the calculator unit is configured to implement the methods described above.

[0038] Because the vehicle is constructed for implementing one of the methods described in this regard, the features described in the scope of one of the methods are correspondingly applicable to the vehicle, and conversely, the features described in the scope of the vehicle are applicable to the method.

[0039] Preferably, the sensor configuration includes ultrasonic sensors, optical cameras such as video cameras and infrared cameras, radar sensors, and lidar sensors, wherein the vehicle includes preferably different types of sensors. Multiple samples of one type may be used. For example, a vehicle may include multiple ultrasonic sensors, multiple video cameras, radar sensors, and lidar sensors.

[0040] Preferably, the connection between the sensor and the calculator unit is bidirectional, allowing not only the transmission of environmental data from the sensor to the calculator unit, but also the transmission of instructions from the calculator unit to the sensor. These instructions can be specifically configured to shut down one or more sensors, place them in standby mode, or turn one or more sensors on. Furthermore, the instructions can be configured to change the configuration of one or more sensors, particularly to change the measurement rate of one or more sensors.

[0041] The sensor can be connected to the calculator unit, for example, via a bus system.

[0042] The computational load of the calculator unit for providing autonomous driving capabilities is dynamically adjusted by selecting sensors and / or adjusting the measurement rate of at least one sensor, as configured according to the invention. In this way, the calculator unit can be designed to be smaller and cheaper, since the analysis and evaluation are based solely on environmental data that is actually necessary for the safe operation of the autonomous vehicle.

[0043] Advantageously, according to the present invention, an objective metric is provided by determining the quality of the autonomous driving function, which can be used to assess the accuracy and safety of the autonomous driving function at any time.

[0044] If, for example, it is confirmed that the quality is below a pre-defined target value, the performance of the autonomous vehicle in the current situation can be improved by selecting additional sensors and / or by increasing the sensor measurement rate. This improves safety during autonomous vehicle operation.

[0045] Furthermore, the energy requirements for providing autonomous driving capabilities can be reduced by shutting down unselected sensors or placing them in standby mode and decreasing the computational load on the calculator unit. This, in particular, can increase the effective range in electrically powered vehicles. Attached Figure Description

[0046] Embodiments of the invention will be explained in more detail with reference to the drawings and the following description.

[0047] The only accompanying drawing schematically illustrates a vehicle with autonomous driving capabilities. Detailed Implementation

[0048] The accompanying drawings are merely schematic depictions of the subject matter of the invention.

[0049] Figure 1 Vehicle 1 is shown, which has a large number of sensors 10 and a calculator unit 30 to provide autonomous driving capabilities. Multiple sensors 10 are connected to the calculator unit 30. Furthermore, the calculator unit 30 is connected to other systems of vehicle 1 to perform not only longitudinal guidance (i.e., acceleration and braking) but also lateral guidance (i.e., steering) of vehicle 1. Figure 1 In the middle, this is roughly drawn through the connection with the steering wheel 7 and the pedal 8.

[0050] exist Figure 1 The vehicle 1 shown in the schematic diagram has different types of sensors 10. In this embodiment, two ultrasonic sensors 11 and a video camera 12 are arranged at the front of the vehicle 1. In addition, a lidar sensor 13 is arranged on the top of the vehicle 1. Needless to say, more or fewer sensors 10 can be provided, or other sensors can also be provided.

[0051] In addition, to determine the current position of vehicle 1, a receiver 20 for a satellite navigation system is provided, which is connected to the calculator unit 30. Preferably, another data source is also utilized to assist in position determination. For example, images from video camera 12 can be analyzed and evaluated to identify landmarks in the surrounding environment of vehicle 1 and to more accurately determine the vehicle's position.

[0052] In the autonomous driving function, vehicle 1 is guided from a starting position to a target position without driver intervention. To provide this autonomous driving function, the calculator unit 30 determines the expected trajectory 2. In determining the expected trajectory 2, environmental data from sensor 10 is used in particular, and the vehicle position determined by receiver 20 is utilized in particular.

[0053] Vehicle 1 is guided along a predetermined trajectory 2, wherein the current vehicle position relative to trajectory 2 is related to a reference point 6, which, in the example shown, is located at the center of the rear axle of vehicle 1. Furthermore, in Figure 1 The actual trajectory 4 is plotted in the diagram, indicating which path vehicle 1 actually traveled with respect to reference point 6. Similarly, the actual trajectory 4 is determined by a satellite navigation system and / or by the identification of landmarks in environmental sensor data and / or by radio signals and / or a combination thereof.

[0054] The method according to the invention further includes comparing the actual driving trajectory 4 with the intended trajectory 2 and determining the quality of the autonomous driving function from the deviation. The smaller the deviation, the higher the quality. Furthermore, preferably, the spacing of objects in the surrounding environment is monitored and considered together when determining the quality.

[0055] After obtaining the quality, for example, it is set to be compared with a pre-defined target value for the quality, including allowable tolerances, and then a selection is made: which sensors 10 should be used to provide autonomous driving functionality. If the quality is higher than the target value, some sensors in sensor 10 can be turned off to save energy and reduce the load on the calculator unit 30. If the quality is lower than the target value, some other sensors in sensor 10 are turned on, and the environmental data from these other sensors is processed by the calculator unit 30.

[0056] In another embodiment of the invention, the measurement rate of one or more sensors 10 may be adapted, additionally or alternatively, according to the quality. If the quality exceeds the target value (including tolerances), the measurement rate can be reduced, thereby reducing the computational load on the calculator unit 30.

[0057] The invention is not limited to the embodiments described herein and the aspects highlighted therein. Rather, numerous modifications that are well-known to those skilled in the art can be made within the scope defined by the claims.

Claims

1. A method for operating the autonomous driving function of a vehicle (1), wherein, The vehicle (1) has a calculator unit (30) and a plurality of sensors (10) for sensing environmental data, wherein the calculator unit (30) is configured to determine a proper trajectory (2) based on the sensed environmental data, and the vehicle (1) is guided along the proper trajectory, characterized in that the method comprises the following steps: a) Sensing the actual trajectory (4), which depicts the path actually traveled by the vehicle (1), and sensing the distances to objects in the surrounding environment of the vehicle (1). b) The quality of the autonomous driving function is determined by comparing the actual trajectory (4) with the intended trajectory (2) and monitoring the sensed distance to objects in the surrounding environment. c) Adjusting the quality to a predetermined target value by selecting a sensor (10) from the plurality of sensors (10) to be used for the autonomous driving function and / or by changing the measurement rate of at least one of the plurality of sensors (10) used to make the measurement.

2. The method according to claim 1, wherein, At least one evaluation factor is determined with regard to at least one additional parameter, and this at least one evaluation factor is taken into account when selecting the sensor (10) to be used and / or when changing the measurement rate, wherein the at least one additional parameter is selected from: the load of the calculator unit, the traffic conditions in which the vehicle (1) is located, the scene in which the vehicle (1) is located and / or the weather data regarding the location of the vehicle (1).

3. The method according to claim 2, characterized in that, The evaluation factors are determined separately for each individual sensor (10) among the plurality of sensors (10).

4. The method according to any one of claims 1 to 3, characterized in that, Turn off or put the unselected sensor (10) into standby mode.

5. The method according to any one of claims 1 to 3, characterized in that, The calculator unit (30) processes the environmental data at a rate equivalent to and adapted to the measurement rate, or the environmental data processing rate is constant and multiple transmissions of a single measurement value are performed when the measurement rate decreases.

6. The method according to any one of claims 1 to 3, characterized in that, The expected trajectory (2) is determined by using a first artificial intelligence model, which is obtained through machine learning.

7. The method according to any one of claims 1 to 3, characterized in that, The selection of the sensor (10) to be used and / or the change of the measurement rate are performed by using a second artificial intelligence model, which is obtained through machine learning.

8. The method according to any one of claims 1 to 3, characterized in that, The determination of the intended trajectory (2) and the selection of the sensor (10) to be used and / or the change of the measurement rate are carried out by using a common artificial intelligence model obtained through machine learning.

9. A computer program product comprising a computer program that, when run on a computer, implements the method according to any one of claims 1 to 8.

10. A vehicle (1) comprising a calculator unit (30) and a plurality of sensors (10) for sensing environmental data, wherein, The calculator unit (30) is configured to provide autonomous driving functionality, characterized in that the calculator unit (30) is further configured to implement the method according to any one of claims 1 to 8.

Images