Computer-implemented method for determining a position of a first motor vehicle

Abstract

A computer-implemented method for determining a position of a first motor vehicle having a first receiver of a first satellite navigation system, the method comprising: driving a second motor vehicle including a second receiver of a second satellite navigation system and a third receiver of a differential satellite navigation system; acquiring second satellite navigation data with the second receiver and acquiring differential satellite navigation data with the differential receiver while the second motor vehicle is driving; storing the second satellite navigation data and the differential satellite navigation data as components of training data; training an artificial intelligence using the training data; acquiring first satellite navigation data with the first receiver while the first motor vehicle is driving; checking the first satellite navigation data for anomalies using the artificial intelligence during or after the first motor vehicle is driving; and correcting the first satellite navigation data using the executed and trained artificial intelligence while the first motor vehicle is driving, if any anomalies are present.

Description

Technical Field

[0001] The invention relates to a computer-implemented method for determining a position of a first motor vehicle. Background Art

[0002] It is known in the prior art to determine the position of a motor vehicle using satellite navigation systems. Examples of satellite navigation systems include GPS (Global Positioning System), Galileo, GLONASS, and BeiDou. Satellites transmit signals containing their position and time. The motor vehicle has a receiver for these signals, which can determine the vehicle's position based on them if a sufficient number of signals are received. As with all measurement methods, these methods also have sources of measurement errors.

[0003] US Pat. No. 10,776,948 B1 discloses a method for reducing errors in position determination using a satellite navigation system. Vector data are used for this purpose. Summary of the Invention

[0004] In contrast, the basic object of the present invention is to reduce errors in determining the position of a motor vehicle, in particular at high speeds.

[0005] This object is achieved by a method according to the following embodiments.

[0006] In this method, the position of a first motor vehicle, which has a first receiver of a first satellite navigation system, is to be determined. To this end, a second motor vehicle is first driven, which includes a second receiver of the second satellite navigation system and a third receiver of a differential satellite navigation system. In the differential satellite navigation system, correction signals are used to reduce inaccuracies caused by different signal velocities in the troposphere and ionosphere and by the fact that satellite orbits are not exactly known. Consequently, the position of the second motor vehicle can be determined more accurately with the third receiver than with the second receiver.

[0007] During travel of the second motor vehicle, second satellite navigation data is acquired using a second receiver, and differential satellite navigation data is acquired using a third receiver. The second and third satellite navigation data include the longitude and latitude of the respective receivers obtained from the received satellite data. The longitude and latitude define the acquisition location of the respective receivers.

[0008] The second satellite navigation data and the differential satellite navigation data are synchronized with each other. This can be done, for example, using time stamps as components of the second satellite navigation data and the differential satellite navigation data. The synchronized second satellite navigation data and the differential satellite navigation data are stored as components of the training data.

[0009] The artificial intelligence is trained using training data. For example, the differential satellite navigation data can be considered true values ​​from which the second satellite navigation data deviates to varying degrees. In particular, by exploiting the second satellite navigation data and the differential satellite navigation data, as well as the differences between them, the artificial intelligence can learn which measurement errors tend to occur in the second satellite navigation data under which circumstances. The trained artificial intelligence is then executed in the first motor vehicle. Alternatively or additionally, execution in the cloud or another backend located outside the vehicle is also proposed. Thus, erroneous satellite navigation data can be corrected after being recorded in the first motor vehicle.

[0010] During the travel of the first motor vehicle, the first receiver is used to acquire first satellite navigation data. During the travel of the first motor vehicle, the trained and executed artificial intelligence is used to check for anomalies in the first satellite navigation data. It may also be provided that the anomaly check is performed only after the travel. Anomalies may, for example, be a sudden jump in the position of the first motor vehicle as defined by the first satellite navigation data, or a sudden jump in the speed as defined by the change in position.

[0011] If an anomaly is present, the executed and trained artificial intelligence is used to correct the first satellite navigation data during or after the first motor vehicle is driven. For this purpose, the artificial intelligence can, for example, calculate correction values ​​by which the first satellite navigation data are corrected.

[0012] Based on the possibly corrected first satellite navigation data, the position of the first receiver and thus of the first motor vehicle can then be determined.

[0013] Because the artificial intelligence is trained using the second satellite navigation data and the differential satellite navigation data, the artificial intelligence can detect and correct anomalies in the first satellite navigation data during driving of the first motor vehicle. Detection and correction are particularly accurate because the particularly accurate differential satellite navigation data is used as the true value for training the artificial intelligence, from which the second satellite navigation data may deviate as an error value.

[0014] In particular, at higher speeds of the first and second motor vehicles, the accuracy of the first or second satellite navigation data is lost. Therefore, it is particularly advantageous if the second motor vehicle is traveling at a relatively high speed, or if the travel at least includes relatively high speeds. At such high speeds, the differential satellite navigation data loses little or no accuracy. If the first motor vehicle then also moves at a relatively high speed, the artificial intelligence can reliably detect and correct the subsequent, more frequent anomalies.

[0015] According to one embodiment of the present invention, while the second motor vehicle is traveling, additional parameters of the second motor vehicle can be acquired, synchronized with the second satellite navigation data and the differential satellite navigation data, and stored as part of the training data. For synchronization, the system time of the second motor vehicle can be used, for example. In this way, the additional parameters can also be taken into account when training the artificial intelligence with the training data. This can improve the accuracy of the correction of the first satellite navigation data by the artificial intelligence.

[0016] According to one embodiment of the present invention, the additional parameters may include speed, braking force, lateral acceleration, longitudinal acceleration, accelerator pedal position, torque applied by the drive of the second motor vehicle, yaw rate, and / or steering angle. These parameters can potentially improve the accuracy of subsequent corrections of the first satellite navigation data using artificial intelligence. Furthermore, it is possible for the first and / or second derivatives of the lateral and / or longitudinal acceleration to also be components of the additional parameters.

[0017] Within the context of this specification, lateral acceleration is understood to mean, in particular, the acceleration in the transverse direction of the respective motor vehicle. Within the context of this specification, longitudinal acceleration is understood to mean, in particular, the acceleration in the longitudinal direction of the respective motor vehicle. The position of the accelerator pedal can, for example, be given as a percentage of the position at a stop. Within the context of this specification, the drive of the respective motor vehicle is understood to mean, in particular, the drive that causes the motor vehicle to accelerate when the user actuates the accelerator pedal.

[0018] According to one embodiment of the present invention, the artificial intelligence can use the first satellite navigation data and further parameters of the first motor vehicle when checking for anomalies in the first satellite navigation data. The further parameters of the first motor vehicle can in particular correspond to further parameters of the second motor vehicle. Thus, the further parameters of the first motor vehicle can include speed, braking force, lateral acceleration, longitudinal acceleration, accelerator pedal position, torque applied by the drive of the first motor vehicle, yaw rate, and / or steering angle.

[0019] Using the additional parameters of the first motor vehicle when checking the first satellite navigation data for anomalies is particularly advantageous if the additional parameters of the second motor vehicle are part of the training data. In this case, the artificial intelligence is trained to correct the satellite navigation data based on the additional parameters. This allows for particularly reliable identification and particularly accurate correction of anomalies. This is particularly applicable when driving at higher speeds, which are characterized in particular by typical values ​​for the additional parameters.

[0020] According to one embodiment of the present invention, when checking for anomalies in the first satellite navigation data, the artificial intelligence may use additional parameters of the first motor vehicle to calculate predicted satellite navigation data. The predicted satellite navigation data is compared with the first satellite navigation data. If the deviation between the predicted satellite navigation data and the first satellite navigation data is greater than a threshold, an anomaly is detected. If this condition is met, the first satellite navigation data is therefore deemed erroneous and corrected.

[0021] It is particularly advantageous to also take into account past first satellite navigation data when calculating the predicted satellite navigation data. This allows for particularly accurate predicted satellite navigation data to be calculated. For example, the current position can be calculated based on the position, speed, and acceleration of the first motor vehicle according to the first satellite navigation data from the previous second.

[0022] According to one embodiment of the present invention, the training data can be cleaned before training the artificial intelligence. This can mean, in particular, that values ​​acquired or measured when the second motor vehicle is below a threshold speed of, for example, 30 km / h are not taken into account and are therefore excluded from the training data. This can be advantageous in that the training of the artificial intelligence is particularly limited to relatively high speeds, at which corrections are particularly necessary. Furthermore, zero values ​​of the further parameters can be deleted during the calculation.

[0023] According to one embodiment of the present invention, the training data may be standardized before training the artificial intelligence. This may facilitate better transfer and comparison of the training data to other motor vehicles, thereby enabling the trained artificial intelligence to be used in as many motor vehicles as possible.

[0024] According to one embodiment of the present invention, the second satellite navigation data and the differential satellite navigation data can be converted into metric values ​​before training the artificial intelligence. This is particularly advantageous if additional parameters of the second motor vehicle are also considered for training the artificial intelligence, which are usually acquired as metric values. The metric values ​​of the second satellite navigation data and the differential satellite navigation data can be converted, in particular, into distances from the equator, the South Pole or the North Pole, and / or the prime meridian. Conversion to distances from any other arbitrary longitude or latitude is also possible.

[0025] According to one embodiment of the present invention, the artificial intelligence can be designed as a recurrent neural network. In particular, the artificial intelligence can be designed as a long short-term memory (LSTM). When correcting anomalies, the artificial intelligence can use only a single model to calculate two correction values ​​used when correcting the first satellite navigation data. One of the correction values ​​can be, for example, a correction value for longitude, while the other correction value can be, for example, a correction value for latitude. Recurrent neural networks, and in particular long short-term memory (LSTM), have been confirmed in experiments to be particularly suitable for correcting the first satellite navigation data.

[0026] According to one embodiment of the present invention, corrections to the first satellite navigation data can be statistically evaluated. This evaluation can, for example, relate to the quality of the corrections. For example, the mean square error, the root mean square error, and / or the coefficient of determination can be used for the evaluation. Based on this statistical evaluation, in particular, inferences about the quality of the trained artificial intelligence can be drawn. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Other features and advantages of the present invention will become clear from the following description of preferred embodiments with reference to the accompanying drawings. In the drawings:

[0028] Figure 1 A schematic flow chart shows a method for determining the position of a first motor vehicle according to one embodiment of the invention. DETAILED DESCRIPTION

[0029] In step 1, second satellite navigation data and further parameters are acquired while a second motor vehicle is traveling. The further parameters include speed, braking force, lateral acceleration, longitudinal acceleration, accelerator pedal position, torque applied by the drive of the second motor vehicle, yaw rate, and / or steering angle. Furthermore, in step 2, data from a differential satellite navigation system are simultaneously acquired, hereinafter referred to as differential satellite navigation data. A third receiver of the differential satellite navigation system is also situated in or on the second motor vehicle, so that the position of the third receiver corresponds to the position of the second motor vehicle.

[0030] In step 3, the second satellite navigation data and differential satellite navigation data acquired in steps 1 and 2, as well as further parameters, are synchronized. In the case of the second satellite navigation data and differential satellite navigation data, this can be done via the timestamps of the second satellite navigation data and differential satellite navigation data. The further parameters can be synchronized with the second satellite navigation data and differential satellite navigation data via the system time of the second motor vehicle. The synchronization causes the simultaneously measured further parameters and the second satellite navigation data and differential satellite navigation data to be correlated with one another, so that when the further parameters and the second satellite navigation data and differential satellite navigation data are subsequently used as training data for artificial intelligence, it is clearly specified which further parameters were acquired simultaneously with which second satellite navigation data and differential satellite navigation data.

[0031] In step 4, some of the additional parameters are selected that are considered particularly relevant for the current driving situation of the second motor vehicle. In step 5, the additional parameters, the second satellite navigation data, and the differential satellite navigation data are cleaned. Zero values ​​for the additional parameters can be deleted. Furthermore, the additional parameters, the second satellite navigation data, and the differential satellite navigation data can be checked for plausibility. This allows, for example, measurement errors to be eliminated. Furthermore, all satellite navigation data and additional parameters acquired when the motor vehicle was below a threshold speed of, for example, 30 km / h are deleted. Because correcting the satellite navigation data with the aid of artificial intelligence is particularly important at higher speeds, such values ​​can be ignored.

[0032] In step 6, further parameters are generated which are used as training data for training the artificial intelligence. These may be, for example, the first and second derivatives of the acceleration of the second motor vehicle.

[0033] The further parameters are normalized in step 7. This is advantageous so that even if an artificial intelligence trained with the further parameters subsequently corrects satellite navigation data of another motor vehicle, the further parameters can still be used as training data.

[0034] In step 8, the second satellite navigation data is converted into metric units. In step 9, the error of the second satellite navigation data is calculated. The error is the difference between the second navigation data and the differential satellite navigation data, because the differential satellite navigation data is regarded as the true value.

[0035] In step 10, a data frame is generated based on the additional parameter, the second satellite navigation data and the differential satellite navigation data, and the data frame is stored as training data. In step 11, the training data is used to train the artificial intelligence.

[0036] In step 12, artificial intelligence is executed in a first motor vehicle. The first motor vehicle includes a first receiver for first satellite navigation data. Furthermore, in step 12, the artificial intelligence calculates future predicted satellite navigation data based on the first satellite navigation data of the first motor vehicle and additional parameters. Subsequently acquired first satellite navigation data is checked for anomalies by comparison with the predicted satellite navigation data. In this way, first satellite navigation data that does not match previously acquired first satellite navigation data of the first motor vehicle and the additional parameters can be detected as an anomaly.

[0037] The further parameters include speed, braking force, lateral acceleration, longitudinal acceleration, position of the accelerator pedal, torque applied by the drive of the first motor vehicle, yaw rate and / or steering angle.

[0038] In step 13, the predicted satellite navigation data are interpolated to obtain a smooth output curve. In step 14, this smooth output curve is used as the corrected first satellite navigation data to determine the position of the first motor vehicle. In addition, in step 14, the improvement in the determination of the position of the first motor vehicle is statistically evaluated in terms of quality.

Claims

1. A computer-implemented method for determining a position of a first motor vehicle, the first motor vehicle having a first receiver of a first satellite navigation system, wherein the first receiver is designed to acquire first satellite navigation data, wherein the method comprises the following steps: - travelling with a second motor vehicle, wherein the second motor vehicle comprises a second receiver of a second satellite navigation system and a third receiver of a differential satellite navigation system; - acquiring (1) second satellite navigation data with the second receiver during travel of the second motor vehicle; - acquiring (2) differential satellite navigation data with the differential receiver during travel of the second motor vehicle; - synchronizing the second satellite navigation data with the differential satellite navigation data (3); - storing the second satellite navigation data and the differential satellite navigation data (10) synchronized with each other as components of the training data; - using said training data to train (11) artificial intelligence; - executing (12) the trained artificial intelligence in the first motor vehicle or in the cloud; - acquiring first satellite navigation data with the first receiver during travel of the first motor vehicle; - using the artificial intelligence to detect (12) anomalies in the first satellite navigation data during or after the first motor vehicle is driven; as well as - If an anomaly exists, correcting (14) said first satellite navigation data using the executed and trained artificial intelligence during travel of said first motor vehicle.

2. The method according to claim 1, characterized in that During the driving of the second motor vehicle, further parameters of the second motor vehicle are acquired (1), synchronized with the second satellite navigation data and the differential satellite navigation data, and stored as a component of the training data.

3. The method according to the preceding claim, characterized in that The further parameters include speed, braking force, lateral acceleration, longitudinal acceleration, position of the accelerator pedal, torque applied by the drive of the second motor vehicle, yaw rate and / or steering angle.

4. The method according to one of the two preceding claims, characterized in that The first satellite navigation data and further parameters of the first motor vehicle are used by the artificial intelligence when checking the first satellite navigation data for anomalies.

5. Method according to the preceding claim, characterized in that When checking the first satellite navigation data for anomalies, the artificial intelligence uses further parameters of the first motor vehicle to calculate predicted satellite navigation data, wherein the predicted satellite navigation data are compared with the first satellite navigation data, and wherein the presence of an anomaly is detected if the predicted satellite navigation data deviates from the first satellite navigation data by more than a threshold value.

6. The method according to any one of claims 1 to 3, characterized in that Before training the artificial intelligence, the training data is cleaned (5).

7. The method according to any one of claims 1 to 3, characterized in that Before training the artificial intelligence, the training data is normalized (7).

8. The method according to any one of claims 1 to 3, characterized in that Prior to training the artificial intelligence, the second satellite navigation data and the differential satellite navigation data are converted (8) into metric values.

9. The method according to any one of claims 1 to 3, characterized in that The artificial intelligence is designed as a recurrent neural network and, during correction of the anomaly, the artificial intelligence uses only a single model to calculate two correction values ​​used in correcting the first satellite navigation data.

10. The method according to any one of claims 1 to 3, characterized in that Corrections to the first satellite navigation data are statistically evaluated (14).

Images