Method for generating synthetic sensor data of a specific sensor generation

Abstract

The invention relates to a method (100) for generating synthetic sensor data (2) of a specific sensor generation (1), comprising the following steps: - providing (101) sensor data (3), wherein the sensor data (3) results from detection by at least one sensor (4) of a first sensor type, - compressing (102) the sensor data (3) using an encoder module (5) to generate a compressed representation (6) of the sensor data (3), - generating (103) the synthetic sensor data (2) based on at least one characteristic of a second sensor type (7), at least one characteristic of the specific sensor generation (8), and the compressed representation (6) of the sensor data (3) using a common decoder module (9) and a decoder module (11) specific to the specific sensor generation (1). The invention further relates to a computer program, a device, and a storage medium for this purpose.

Description

[0001] The invention relates to a method for generating synthetic sensor data of a specific sensor generation. Furthermore, the invention relates to a computer program, a device, and a storage medium for this purpose. State of the art

[0002] Generating synthetic sensor data through machine learning models is an advanced approach in artificial intelligence that allows for the creation of realistic, yet artificially generated data that simulates real-world sensor outputs. This technique is particularly useful in scenarios where access to real sensor data is limited or impossible, whether due to cost, privacy concerns, or practical constraints. Machine learning models, particularly those based on neural networks, can be trained to capture the characteristics and patterns of real sensor data and subsequently generate data similar to it. This makes it possible to develop more robust and efficient machine learning systems, as these models can be trained with a wide variety of data that reflect reality in different scenarios and conditions.

[0003] However, for specific sensor generations, it may be necessary to train a corresponding machine learning model each time, according to the methods known in the state of the art. Disclosure of the invention

[0004] The subject matter of the invention is a method having the features of claim 1, a computer program having the features of claim 10, a device having the features of claim 11, and a computer-readable storage medium having the features of claim 12. Further features and details of the invention emerge from the respective subclaims, the description, and the drawings. Features and details described in connection with the method according to the invention naturally also apply in connection with the computer program according to the invention, the device according to the invention, and the computer-readable storage medium according to the invention, and vice versa, so that with regard to the disclosure of the individual aspects of the invention, reference is or can always be made to each other.

[0005] The invention particularly relates to a method for generating synthetic sensor data of a specific sensor generation, comprising the following steps, wherein the steps can be performed repeatedly and / or sequentially. The specific sensor generation is, in particular, a specific model, such as a new product line, of a sensor, wherein the specific model also has respective specific physical characteristics or feature characteristics.

[0006] In a first step, sensor data is preferably provided, wherein the sensor data results from detection by at least one sensor of a first sensor type. The sensor data can be image data, for example. The at least one sensor can be, for example, a camera sensor or an infrared camera sensor, a radar sensor, a LiDAR sensor, or an ultrasonic sensor, so that the image data can also be embodied as infrared, radar, LiDAR, or ultrasonic image data. However, the above list is not exhaustive, so that other sensors are also conceivable in addition to those mentioned.

[0007] In a further step, the sensor data is preferably compressed using an encoder module to generate a compressed representation of the sensor data. In other words, the sensor data is, in particular, transferred into a latent space. A feature vector can be extracted based on the sensor data, which represents essential features of the sensor data. In other words, the encoder module is, in particular, a component designed to convert the sensor data into a more compact form. This conversion preferably aims to extract essential features and / or structures of the sensor data and, in doing so, reduce redundant or unimportant information.

[0008] In a further step, the synthetic sensor data are preferably generated based on at least one characteristic of a second sensor type, at least one characteristic of the specific sensor generation, and the compressed representation of the sensor data using a common decoder module and a decoder module specific to the specific sensor generation of the second sensor type. The specific sensor generation is, in particular, a specific sensor generation of the second sensor type. If the second sensor type is specific to a radar sensor technology, for example, the at least one characteristic can describe fundamental properties for sensor data from radar sensors. The specific sensor generation can then be, for example, a specific model or product line of a radar sensor.The at least one characteristic of the specific sensor generation can be, for example, a number of transmitting antennas of the radar sensor or an alignment of transmitting and receiving antennas of the radar sensor. Furthermore, a characteristic of the specific sensor generation can be how much data is available for this specific sensor generation. The machine learning model adaptation discussed in a subsequent section can also be influenced according to the amount of available data, i.e., the less data is available, the less the specific sensor generation is preferably taken into account in the machine learning model adaptation with regard to the common decoder module. One advantage here can be that during backpropagation for the training of the entire machine learning model (i.e.In a machine learning model (in particular a machine learning model comprising the encoder and decoder parts), weights and parameters for the sensor generation-specific decoder module can be adapted based on the training data for the specific sensor generation. However, the effect of backpropagation on weights and parameters of the common decoder module can be adapted in proportion to the amount of training data available for this specific sensor generation in comparison to, or ratio to, the data available for all sensor generations together, or training data. For example, for a new sensor generation that is still in advanced development, the sensor generation-specific decoder module can be adapted with the limited available data, but the common decoder module is preferably only adapted in proportion to the amount of training data for the new sensor generation compared to the older sensor generations.Another possibility would be to estimate the sensor generation- or subgeneration-specific decoder module based on the sensor generation or the sensor subgeneration characteristics for a preliminary estimation of the synthetic sensor data for this new specific decoder module. This estimate can be derived based on knowledge of the other sensor generation-specific decoder modules and their corresponding sensor generation-specific characteristics. This approach could also be used to initialize the weights and parameters for the new sensor generation- or subgeneration-specific decoder module under design, thus reducing training times. At least one further level can also be provided for a subgeneration of the sensor generation.Accordingly, the synthetic sensor data would further be generated based on at least one characteristic of the at least one subgeneration and further using a decoder module specific to the subgeneration of the second sensor type. Provision can be made for at least two decoder modules specific to the specific sensor generation of the second sensor type to be provided. A suitable common decoder module can then be determined or selected based on the at least one characteristic of the second sensor type. Furthermore, a suitable decoder module specific to the specific sensor generation of the second sensor type can be determined or selected by a selector module based on the at least one characteristic of the specific sensor generation of the second sensor type or by a provided input, for example from a user, which includes a description of the specific sensor generation.

[0009] The first sensor type and the second sensor type can be specific to a different sensor technology or to the same sensor technology. Possible sensor technologies include, for example, a camera sensor, an infrared camera sensor, a radar sensor, a LiDAR sensor, or an ultrasonic sensor. However, the above list is not exhaustive, so other analog sensor technologies are also applicable in addition to those mentioned. For example, the first sensor type could be a radar imaging device and the second sensor type a radar sensor, for example in a vehicle.

[0010] According to an advantageous development of the invention, the common decoder module can be a trained machine learning model trained to preprocess the compressed representation for the second sensor type, and in particular for a further sensor type. The preprocessing preferably comprises at least a partial reconstruction of the synthetic sensor data to be generated. The decoder module specific to the specific sensor generation of the second sensor type is preferably another trained machine learning model trained to generate the synthetic sensor data based on the preprocessed compressed representation.

[0011] Furthermore, within the scope of the invention, it is conceivable that the encoder module, the common decoder module, and the decoder module specific to the specific sensor generation of the second sensor type are each a machine learning model, in particular a neural network, wherein the encoder module is preferably a foundation machine learning model. A foundation machine learning model is, in particular, a comprehensive, pre-trained machine learning model based on a large amount of heterogeneous data. It serves, for example, as a fundamental architecture on which more specific and adapted models can be developed for a variety of applications. Basic machine learning models are characterized in particular by their ability to recognize and generalize complex patterns and relationships in large data sets. The aforementioned machine learning models are preferably each trained machine learning model.The same encoder module can be provided for training different common decoder modules and / or different decoder modules specific to the specific sensor generation of the second sensor type. The weights of this encoder module can be adjusted during training. Furthermore, the same common decoder module can be provided for training different decoder modules specific to the specific sensor generation of the second sensor type. The weights of this common decoder module and also of the encoder module can be adjusted during training.

[0012] Alternatively, it is also conceivable that the encoder module, the common decoder module and the decoder module specific to the specific sensor generation of the second sensor type are designed and trained as a single machine learning model, in particular a basic machine learning model.

[0013] Within the scope of the invention, it can be provided that an encoder module specific to a sensor generation of the first sensor type is also provided, and the compression is performed using the encoder module specific to the sensor generation of the first sensor type and the encoder module. This allows the compression of the sensor data to be performed specifically for each sensor generation of the first sensor type and thus more precisely. In other words, according to this alternative, the encoder side is also designed in multiple stages. This allows, for example, a specific sensor generation to be taken into account by the corresponding encoder module specific to the sensor generation of the first sensor type.

[0014] A further advantage within the scope of the invention can be achieved if the method further comprises the following step: - Emulating a sensor of the second sensor type based on the generated synthetic sensor data.

[0015] For example, an application is conceivable in which radar data for a radar device is emulated and thus provided based on sensor data in the form of image data from a camera.

[0016] Furthermore, it may be advantageous within the scope of the invention for the method to further comprise the following steps: - Providing further sensor data, wherein the further sensor data are specific to the second sensor type and represent a same scene as the sensor data, wherein the further sensor data result from a detection of at least one further sensor, - Comparing the additional sensor data collected with the generated synthetic sensor data, - Detecting an error or impairment of the at least one further sensor based on a result of the comparison.

[0017] The scene can be specific to a movement in an environment or represent this. For example, within the scope of the above steps, the sensor data could be recorded simultaneously with a sensor of the first sensor type and the further sensor data with a sensor of the second sensor type in order to record the same scene. The synthetic sensor data can advantageously be generated with such precision that a check of a real sensor, i.e. a detection of errors or impairments of the sensor, is possible. An impairment could, for example, be a weather-related restriction of visibility of a camera sensor. The synthetic sensor data could then be generated on the basis of the sensor data of the sensor of the first sensor type. Alternatively, an error or impairment of the at least one sensor of the first sensor type could also be detected on the basis of the comparison.

[0018] It may optionally be possible for the method to further comprise the following step: - Compensating for the error or impairment of the at least one further sensor on the basis of the synthetic sensor data by modifying the further acquired sensor data on the basis of the synthetic sensor data.

[0019] For example, an impairment such as a weather-related restriction of visibility of a camera sensor could advantageously be supplemented by the generated synthetic sensor data, since a sensor of the first sensor type, such as a radar sensor, which records the sensor data, may not be restricted. If, based on the comparison, an error or impairment of the at least one sensor of the first sensor type was detected, an alternative can be provided for this error or impairment to also be compensated for by a corresponding modification. However, an opposite method may be necessary to generate corresponding synthetic sensor data. "Opposite" refers in particular to the fact that a corresponding encoder part according to the invention would then be necessary for the second sensor type and a corresponding decoder part according to the invention would then be necessary for the first sensor type.

[0020] According to a further possibility, the method may further comprise the following step: - Generating or verifying at least one label for training a machine learning model in a training data set, wherein the training data set is specific to sensor data of the second sensor type.

[0021] The marking can also be referred to and understood as a label. The training dataset can be a training dataset for a machine learning model that uses the markings, or labels, for training. The generated synthetic sensor data can advantageously provide increased variability in the training dataset. The generation of at least one marking can occur, for example, when an object that does not yet appear in the training dataset or a known object with additional details is simulated using the synthetic sensor data.

[0022] It is further conceivable that the sensor data and the synthetic sensor data are specific to road traffic and the method further comprises the following step: - Generating or adapting a road signature based on the generated synthetic sensor data.

[0023] A road signature refers in particular to a characteristic pattern or a series of features that characterise a particular road or road network. The road signature can comprise various elements, which can be of a physical or abstract nature. Physical features include, for example, the width and condition of a road, the type of development along the road, road markings, curbs, sidewalks and street lighting. Special features such as trees, green spaces or special architectural elements can also be physical features. Furthermore, a road signature can comprise a traffic pattern, i.e., the type and intensity of traffic flow, predominant modes of transport (e.g., cars, bicycles, pedestrians), and the presence of public transport.For example, the present invention advantageously allows the road signature to be updated from one sensor generation to a next sensor generation, which can provide further details or features in the road signature.

[0024] It is possible for the method according to the invention to be used in a vehicle. The vehicle can be designed, for example, as a motor vehicle and / or passenger vehicle and / or an autonomous vehicle. The vehicle can have a vehicle device, for example, for providing an autonomous driving function and / or a driver assistance system. The vehicle device can be designed to control and / or accelerate and / or decelerate and / or steer the vehicle at least partially automatically.

[0025] The invention also relates to a computer program, in particular a computer program product, comprising instructions that, when executed by a computer, cause the computer to carry out the method according to the invention. Thus, the computer program according to the invention provides the same advantages as those described in detail with reference to a method according to the invention.

[0026] The invention also relates to a data processing device configured to carry out the method according to the invention. The device can be, for example, a computer that executes the computer program according to the invention. The computer can have at least one processor for executing the computer program. A non-volatile data memory can also be provided, in which the computer program is stored and from which the computer program can be read by the processor for execution.

[0027] The invention may also provide a computer-readable storage medium that contains the computer program according to the invention and / or includes instructions that, when executed by a computer, cause the computer to carry out the method according to the invention. The storage medium is designed, for example, as a data storage device such as a hard disk and / or a non-volatile memory and / or a memory card. The storage medium can, for example, be integrated into the computer.

[0028] Furthermore, the method according to the invention can also be implemented as a computer-implemented method.

[0029] Further advantages, features, and details of the invention will become apparent from the following description, which describes embodiments of the invention in detail with reference to the drawings. The features mentioned in the claims and in the description may be essential to the invention individually or in any combination. They show: Fig. 1 a schematic visualization of a method, a sensor generation, a sensor, a device, a storage medium and a computer program according to embodiments of the invention, Fig. 2 a schematic representation of a method according to embodiments of the invention.

[0030] In Fig. 1, a method 100, a sensor generation 1, a sensor 4, a device 10, a storage medium 15 and a computer program 20 according to embodiments of the invention are schematically shown.

[0031] Fig. 1 shows in particular an embodiment of a method 100 for generating synthetic sensor data 2 of a specific sensor generation 1. In a first step 101, sensor data 3 is provided, wherein the sensor data 3 results from a detection by at least one sensor 4 of a first sensor type. In a second step 102, the sensor data 3 is compressed using an encoder module 5 to generate a compressed representation 6 of the sensor data 3. In a third step 103, the synthetic sensor data 2 is generated based on at least one characteristic of a second sensor type 7, at least one characteristic of the specific sensor generation 8, and the compressed representation 6 of the sensor data 3 using a common decoder module 9 and a decoder module 11 specific to the specific sensor generation 1.The first sensor type and the second sensor type are specific to a different sensor technology, but can also be specific to the same sensor technology. For example, the first sensor type could be a radar imaging device and the second sensor type a radar sensor, for example, in a vehicle.

[0032] The present invention, according to one exemplary embodiment, particularly utilizes a machine learning model, preferably a generative machine learning model, for generating synthetic sensor data 2, such as reflections, for a sensor generation 1 based on a scenario from sensor data 3, such as an image or video stream. In particular, a base machine learning model is used for the machine learning model. This should preferably be applicable across multiple sensor generations 1.

[0033] For this purpose, an input of sensor data 3 of a sensor technology such as a camera or LiDAR sensor 4 is preferably provided and, on the basis of this input, an output of synthetic sensor data 2 is generated which is specific for another sensor technology, such as reflections for radar, LiDAR or ultrasonic sensors.

[0034] A possible architecture of a machine learning model that can generate synthetic sensor data 2 for different sensor generations 1 is described below. The machine learning model can be improved in particular for all sensor generations 1, even if only training data from one of the sensor generations 1 is used for training. The machine learning model can, for example, be designed as a "plug-and-play" solution, in which the architecture depends on the input and the required output for the specific sensor generation 1. With the help of an encoder module 5 or a similar block, a compressed, i.e. in particular low-dimensional, representation 6 can be generated based on the input. Depending on the specific sensor generation 1 present, the subsequent decoder modules can be "plug and play" if the input remains the same.According to embodiments, a common decoder module 9 and at least one sensor generation-specific 1 or sensor subgeneration-specific decoder module 11 can be provided for a specific sensor type.

[0035] An embodiment of a method according to such a “plug-and-play” solution is described in Fig.2. Sensor data 3 is first provided to an encoder module 5. The encoder module 5 compresses the sensor data 3 into a compressed representation 6 of the sensor data 3. A common decoder module 9 then pre-processes this compressed representation 6. The next stage in this exemplary embodiment is a selector 12, which receives as input at least one characteristic of a second sensor type 7 for which the synthetic sensor data 2 is to be generated. Subsequently, the characteristic of the second sensor type 7 and at least one characteristic 8 can be provided for a given specific sensor generation 1 in order to generate the respective synthetic sensor data 2 in a subsequent step by a decoder module 11 specific to the specific sensor generation of the second sensor type.

[0036] Due to the modular architecture, the encoder module 5 can also be reused for a different sensor type that has the same input, i.e., sensor data 3 specific to the same type of sensor 4, e.g., a camera sensor, according to embodiments. For example, an encoder module 5 from a machine learning model for input of image data and output of radar reflections could also be reused for machine learning models for input of image data and output of ultrasound or LiDAR sensor data 3.

[0037] One advantage of the method according to exemplary embodiments is, in particular, that even with training data for a specific sensor generation 1, additional components such as the encoder module 5 or even the entire machine learning model, i.e., which is designed in particular as a base machine learning model, can be trained and improved. The effects of the training on the sensor-generation-specific decoder module 11 can be greater. In the case of backpropagation for fine-tuning weights, for example, the sensor-generation-specific decoder module 11 is largely fine-tuned, while the influence on the common decoder module 9 is smaller. The proportion of the effects is based, in particular, on the sensor-generation-specific characteristics.

[0038] The at least one characteristic of the specific sensor generation 8 can also encompass multiple sensor generations 1, defining aspects such as a number of transmitting antennas, an orientation of transmitting and receiving antennas, or even a quantity of training data available for this specific sensor generation 1 in relation to a total quantity of training data for the entire training model for all generations. Thus, the entire machine learning model can also be improved by training for a specific sensor generation 1.

[0039] According to exemplary embodiments, the invention advantageously enables a performance estimate for a future sensor generation 1. Using the method described above, it is particularly possible to create an estimate, for example of reflections, for a future sensor generation 1 such as a sensor hardware front end for a specific scenario. In this case, for example, only a few important characteristics of the new sensor generation 1 may be known, such as a number of transmitting antennas of the new sensor generation 1 and / or an orientation of the antennas. If necessary, the sensor generation-specific decoder module 11 for this new sensor generation 1 could also be estimated using the knowledge from the previous sensor generation-specific decoder modules 11 and their corresponding sensor generation-specific characteristics 8 together with the new sensor generation-specific characteristics 8.Using this approach, for example, reflections for a specific scenario can be generated for the new sensor generation 1 to pre-validate the performance of sensor 4 for these specific scenarios. This can advantageously accelerate the development of front-end hardware for new sensor generation 1 through early feedback.

[0040] Furthermore, synthetic sensor data 2, such as reflections, can be generated using sensor data 3, such as an image or video stream at the highest resolution input, such as for a radar imager, which can then be simulated by a target simulator. The scene can then be used to simulate current sensor generations 1, depending on their processing capabilities, for example, what the specific sensor generation 1 has or has not been able to detect.

[0041] Using an emulator, synthetic sensor data 2 can be generated for any scenario. Subsequently, any sensor generation 1 under test can be used with its own specific characteristics. Depending on the hardware configuration of the sensor generation 1 under test, the same synthetic sensor data 2 can be used for all generations of sensor hardware in the loop simulations.

[0042] For example, the emulator cells emulate the reflections of an object, and based on the cells triggered by the electromagnetic waves received by the actual sensor under test (4), the corresponding emulation can be sent from the emulator cells back to the actual hardware under test. With such an approach, specific corner-case scenarios can be simulated using the target simulator to determine the first-generation sensor that offers the best performance for such scenarios. It would also be possible to accelerate the development of hardware frontends for sensors by validating their performance early on.

[0043] Furthermore, according to embodiments, the invention can perform early fusion for performance improvement and ego-vehicle localization in sensor fusion units. In vehicles, in particular, a zonal architecture is increasingly being provided, in which vehicle computers are used that process the data from multiple sensors in order to fuse the data for the perception of the environment. The method according to embodiments for generating the synthetic sensor data 2, such as reflections, can be embedded in the central computer, which generates the synthetic sensor data 2 from the sensor data 3, e.g., a camera or video stream, and compares it with the additional sensor data supplied by the individual sensor generations, such as the radar sensors.This can provide early fusion feedback for sensor data from multiple sensors, such as cameras and radar, and can also improve the overall performance of the fusion system. It could also be helpful to alert in case of discrepancies between different sensors, such as cameras / radars, etc.

[0044] During the early development phases of the fusion units and the A / B sensor testing phases, direct feedback can be provided on the performance of a sensor system, or rather a specific sensor generation 1. For example, a fusion unit can receive sensor data 3 from the radar sensors under test. The fusion unit can also perform the method 100 proposed above for generating synthetic sensor data 3, in particular reflections, based on the sensor data 3, in particular camera images. Based on the generated synthetic sensor data 3, in particular reflections, the positions of objects can be determined. These determined positions can be compared with the positions actually provided by the radar sensors. If there is a discrepancy between the two, a snapshot can be created with the camera image and the positions of the objects provided by the radar sensors, if possible with a brief description.The snapshot data can be analyzed later to improve performance.

[0045] The machine learning model proposed above can be deployed, particularly in the central fusion unit, and, based on faulty or degraded sensor data 3 such as a camera input, further generate synthetic sensor data 2 such as reflections for the specific sensor generation 1 for which it was trained. An actual sensor 4 mounted on the vehicle preferably also processes the sensor data 3, particularly reflections it detects. The missing or degraded sensor data 3 can be analyzed based on the synthetic sensor data 2 and used as input for the camera system, which was unable to correctly detect the reflections due to lighting conditions, a dirty lens, etc.

[0046] Furthermore, a correction of a sensor 4, for example, a camera perception, can be provided in the case of problems such as those caused by lighting conditions or a dirty lens. The method according to embodiments could also be trained to generate synthetic sensor data 2, such as a camera image, based on sensor data 3, such as reflections from a sensor 4, such as radar, LiDAR, etc., at the input. The generated synthetic sensor data 2, in particular a generated camera image, can then be compared with the actual sensor data 3, such as a camera image, in order to correct the actual sensor data 3, or the actual camera image.

[0047] Various sensor system models known in the prior art, which are used to estimate the sensor performance for a specific scenario, are partially unrealistic and, in particular, have lower performance, such as the ray tracing model used for radar systems. With the method proposed above according to embodiments, which can be trained to a desired performance level, the sensor system models required for "X-in-the-loop simulations" would no longer be needed. The at least one described machine learning model according to embodiments can be trained with sufficient training data so that it closely approximates reality until the sensor system models are sufficiently mature. The sensor system models could also be improved with the machine learning models proposed above in order to increase accuracy.

[0048] The method according to embodiments can, for example, also be used to generate reflections from an image. The reflections are then processed on computers, in particular without actual sensor hardware, using the sensor algorithms to obtain objects and their position relative to the scenario at the output. This object information can then be used to label the data for the camera image, i.e., with labels, or to validate the labels for the camera image. In this way, the quality of the labels in camera images can be improved. The proposed idea can be used to provide labels for any sensor, such as cameras, radar, LiDAR, etc.

[0049] As sensor generation 1 advances, road signatures from an earlier sensor generation 1 may become invalid in the future because newer sensor generations 1 can detect more objects or with improved performance or accuracy. For example, radar systems can currently generate locations with fewer reflections from objects, while future sensor generation 1 of radar imagers may be able to generate point clouds of objects and thus detect more objects. Using the method according to embodiments, it may be possible to use road signatures from earlier sensor generations 1 for newer ones as well. For this purpose, for example, reflections from a camera image can be generated for a newer sensor generation 1 if characteristics of the newer sensor generation 1 are known.The reflections generated by the newer sensor generation 1 can then be processed with the characteristics of the newer generation to output the detected objects. Static objects such as guardrails along the road or road signs from the output can be mapped to the existing sensor road signature objects, and the delta between the two can be added to the road signature for the newer generation. Thus, when a vehicle equipped with the newer sensor generation 1 drives along the road, it has access to the generation-specific road signature from a cloud, for example. The method according to embodiments thus offers, in particular, the possibility of an incremental road sensor signature for updating sensor generations 1.

[0050] The above explanation of the embodiments describes the present invention exclusively by way of examples. Of course, individual features of the embodiments can be freely combined with one another, provided they are technically feasible, without departing from the scope of the present invention.

Claims

[1] Method (100) for generating synthetic sensor data (2) of a specific sensor generation (1), comprising the following steps: - Providing (101) sensor data (3), wherein the sensor data (3) result from a detection of at least one sensor (4) of a first sensor type, - compressing (102) the sensor data (3) using an encoder module (5) to generate a compressed representation (6) of the sensor data (3), - generating (103) the synthetic sensor data (2) on the basis of at least one characteristic of a second sensor type (7), at least one characteristic of the specific sensor generation (8) and the compressed representation (6) of the sensor data (3) using a common decoder module (9) and a decoder module (11) specific to the specific sensor generation (1). [2] Method (100) according to claim 1, characterized by , that the common decoder module (9) is a trained machine learning model which is trained to preprocess the compressed representation (6) for the second sensor type, and in particular for a further sensor type, wherein the decoder module (11) specific to the specific sensor generation (1) of the second sensor type is a further trained machine learning model which is trained to generate the synthetic sensor data (2) on the basis of the preprocessed compressed representation (6). [3] Method (100) according to one of the preceding claims, characterized by that the encoder module (5), the common decoder module (9) and the decoder module (11) specific for the specific sensor generation (1) of the second sensor type are each a machine learning model, in particular a neural network, wherein the encoder module (5) is preferably a basic machine learning model. [4] Method (100) according to one of the preceding claims, characterized by that an encoder module specific for a sensor generation of the first sensor type is further provided and the compression (102) is carried out using the encoder module specific for the sensor generation of the first sensor type and the encoder module (5). [5] Method (100) according to one of the preceding claims, characterized by that the method (100) further comprises the following step: - Emulating a sensor of the second sensor type based on the generated synthetic sensor data (2). [6] Method (100) according to one of the preceding claims, characterized by that the method (100) further comprises the following steps: - providing further sensor data (3), wherein the further sensor data (3) are specific to the second sensor type and represent a same scene as the sensor data (3), wherein the further sensor data (3) result from a detection of at least one further sensor (4), - comparing the additional sensor data recorded (3) with the generated synthetic sensor data (2), - Detecting an error or impairment of the at least one further sensor (4) based on a result of the comparison. [7] Method (100) according to claim 6, characterized by that the method (100) further comprises the following step: - compensating for the error or impairment of the at least one further sensor (4) on the basis of the synthetic sensor data (2) by modifying the further acquired sensor data (3) on the basis of the synthetic sensor data (2). [8] Method (100) according to one of the preceding claims, characterized by that the method (100) further comprises the following step: - generating or checking at least one label for training a machine learning model in a training data set, wherein the training data set is specific for sensor data (3) of the second sensor type. [9] Method (100) according to one of the preceding claims, characterized by that the sensor data (3) and the synthetic sensor data (2) are specific to road traffic and the method (100) further comprises the following step: - Generating or adapting a road signature based on the generated synthetic sensor data (2). [10] Computer program (20) comprising instructions which, when the computer program (20) is executed by a computer (10), cause the computer (10) to carry out the method (100) according to one of the preceding claims. [11] Device (10) for data processing, which is arranged to carry out the method (100) according to one of claims 1 to 9. [12] A computer-readable storage medium (15) comprising instructions which, when executed by a computer (10), cause the computer (10) to carry out the steps of the method (100) according to any one of claims 1 to 9.

Images