WO2018151404A1

WO2018151404A1 - Method for measuring uncertainty of gnss-based position estimation

Info

Publication number: WO2018151404A1
Application number: PCT/KR2017/014982
Authority: WO
Inventors: 정우진; 이우식
Original assignee: 고려대학교 산학협력단
Priority date: 2017-02-15
Filing date: 2017-12-19
Publication date: 2018-08-23
Also published as: KR101921483B1; KR20180094529A

Abstract

The present invention relates to a method for measuring the uncertainty of a GNSS-based position estimation, the method comprising the steps of: (A) measuring a sensor factor uncertainty value of a sensor factor uncertainty group under conditions where no environment factor uncertainty group and no model factor uncertainty group are generated; (B) measuring a model factor uncertainty value of the model factor uncertainty group under conditions where the environment factor uncertainty group is not generated while applying the sensor factor uncertainty value thereto; (C) measuring an environment factor uncertainty value of the environment factor uncertainty group while applying the sensor factor uncertainty value and the model factor uncertainty value thereto; and (D) measuring an uncertainty value of the GNSS-based position estimation on the basis of the sensor factor uncertainty value, the model factor uncertainty value, and the environment factor uncertainty value, wherein the sensor factor uncertainty group, the environment factor uncertainty group, and the model factor uncertainty group are obtained by grouping a plurality of uncertainty factors causing uncertainty of the GNSS-based position estimation according to causes that generate uncertainty. Accordingly, uncertainty generated during GNSS sensor-based position measurement, which is used for a self-driving vehicle that employs a GNSS sensor or for self-driving of an outdoor moving robot, for example, can be accurately measured and reflected for accurate position estimation of the self-driving vehicle.

Description

How to Measure Uncertainty of GPS-Based Position Estimation

The present invention relates to a method for measuring the uncertainty of GNSS-based position estimation, and more particularly, to be used for autonomous driving of an autonomous vehicle or an outdoor mobile robot (hereinafter referred to as an 'autonomous vehicle') using a GNSS sensor. The present invention relates to a method for measuring the uncertainty of GNSS-based position estimation, which can accurately measure the uncertainty generated in the position measurement of the GNSS sensor and reflect the accurate position estimation of the autonomous vehicle.

Recently, various studies for precise position estimation of autonomous vehicles have been conducted. Location estimation requires sensor data that measures the environment or location, and often converges various sensor data to increase the accuracy of location estimation.

In order to fuse the sensor data, the measurement of each sensor and the uncertainty of the measurement are known. This is called the sensor model. In addition, the more accurately the sensor model is designed, the more accurate the position estimation algorithm using the sensor model is.

GNSS (Global Navigation Satellite System) sensor has been used for the estimation of the position of autonomous vehicle for a long time in the form of fusion with other sensors. Therefore, it is important to accurately design GNSS sensor model for accurate position estimation of autonomous vehicle.

The most commonly used GNSS sensor model design method is to calculate the uncertainty of the measured value from the measured value of the GNSS sensor and the NMEA 0183 standard information provided by the GNSS sensor.

However, in an urban environment, the actual uncertainty of the GNSS sensor measurement value and the uncertainty calculated from the information provided by the GNSS sensor may be inconsistent due to the multipath effect and the foliage attenuation.

In addition to the above-mentioned uncertainty factors, various uncertainty factors cause the uncertainty of the GNSS sensor, which is described in Parkinson et al. In "Progress in Astronautics and Aeronautics (Global Positioning System: Theory and Applications. Vol. 2. Aiaa, 1996.)". Uncertainties that cause errors in position estimation include multipath effects, foliage attenuation, ephemeris error, satellite clock error, and ionospheric effect. ) Factors, Tropospheric effect factors, GNSS sensor hardware and software factors, and DOP (Delusion of Precision).

Here, the DOP is a factor of uncertainty caused by the geometry of the satellites, and the remaining factor of uncertainty (hereinafter, referred to as a range error) is an uncertainty caused by the satellite signal.

More specifically, the DOP represents an error in the position measurement due to the geometric relationship due to the relative arrangement of the satellites, and is a dimensionless number. Here, the DOP is calculated by the GNSS sensor in the process of location estimation, and obtained in real time as NMEA 0183 regulation information in an environment using the GNSS sensor, and the present invention is also used for this.

Range errors correspond to error sources that affect the movement of a signal while the satellite signal transmitted from the satellite reaches the GNSS sensor.

First, the ionosphere influencing factor is caused by the phenomenon that the satellite signal does not proceed at the ideal light speed by the free electrons of the ionosphere when the satellite signal passes through the ionosphere of the atmosphere.

Atmospheric influencing factors are caused by satellite signals passing through the atmosphere and failing to propagate at the ideal speed of light by air molecules in the atmosphere.

The astronomical error error occurs when there is an error in the satellite arrangement information, and generally generates an error of 0.05m or less per axis.

The satellite time error factor occurs when there is an error in the time information of the satellite, and generally shows an error of 0.2m or less.

Multipath influencers occur when satellite signals are reflected from tall buildings, cliffs, etc. around the GNSS sensor and have longer paths than the line of sight. Although the error can be reduced by the filter and antenna design inside the receiver of the GNSS sensor, there may be an error of about 15m or more.

The signal obstruction factor is an uncertainty factor caused by an obstacle in the line of sight between the GNSS sensor and the satellite. In this case, as the strength of the satellite signal decreases, the position estimation uncertainty increases and an error of 50m or more may occur.

In addition, the hardware and software factors of the GNSS sensor are being reduced through the development of hardware and software, and the recent equipment has a precision of 1 s 0.3 m or less.

For the above uncertainty factor, Parkinson et al. Define the uncertainty value of GNSS based position estimation as shown in [Equation 1].

[Equation 1]

σ _g is the uncertainty value of GNSS-based position estimation, DOP is Delusion of Precision, and Range error is the uncertainty factor that constitutes the range error.

However, the above-mentioned uncertainty factors constituting the range error can be applied to Equation 1 when each uncertainty factor is individually measured or calculated, but the above-mentioned uncertainty factors cannot be individually measured or calculated, and [Equation 1] In case of, it has only a theoretical meaning and cannot be applied to measuring the uncertainty of the actual GNSS sensor.

To compensate for this, a method of selectively using measurements using the Mahalanobis distance between GNSS measurements is described in Bouvet, Denis, and Gaetan Garcia's article "Improving the accuracy of dynamic localization systems using RTK." GPS by identifying the GPS latency (Robotics and Automation, 2000. Proceedings. ICRA'00. IEEE International Conference on Vol. 3. IEEE, 2000.) ”. It is hard to say that the design of the accurate GNSS sensor model is not because the effect is not reduced.

Accordingly, the present invention has been made to solve the above problems, to accurately estimate the uncertainty generated in the position measurement of the GNSS sensor used in the autonomous driving of the autonomous vehicle using the GNSS sensor to accurately estimate the position of the autonomous vehicle The purpose is to provide a method for measuring the uncertainty of GNSS-based location estimation that can be reflected in the system.

According to the present invention, in the method of measuring the uncertainty of the position estimation based on GNSS, (a) the sensor factor uncertainty value of the sensor factor uncertainty group under the condition that the environmental factor uncertainty group and the model factor uncertainty group do not occur (B) measuring the environmental factor uncertainty value of the environmental factor uncertainty group under the condition that the model factor uncertainty group does not occur, wherein the sensor factor uncertainty value is applied and measured; and (c) the sensor Measuring an uncertainty value of the GNSS based location estimate based on a factor uncertainty value and the environmental factor uncertainty value; The sensor factor uncertainty group, the environmental factor uncertainty group, and the model factor uncertainty group include a plurality of uncertainty factors that cause the uncertainty of the GNSS-based location estimation grouped according to the occurrence of the uncertainty. This is achieved by measuring the uncertainty of the estimate.

On the other hand, the above object is, according to another embodiment of the present invention, in the method for measuring the uncertainty of the position estimation based on GNSS, (A) sensor factor uncertainty value of the sensor factor uncertainty group is environmental factor uncertainty group and model factor uncertainty group (B) the model factor uncertainty value of the model factor uncertainty group is measured under the condition that the environmental factor uncertainty group does not occur, but the sensor factor uncertainty value is applied and measured. (C) the environmental factor uncertainty value of the environmental factor uncertainty group is measured by applying the sensor factor uncertainty value and the model factor uncertainty value, and (D) the sensor factor uncertainty value, the model factor uncertainty value, and the Of GNSS-based Location Estimation Based on Environmental Factor Uncertainty Values A step which the authenticity and value measured; The sensor factor uncertainty group, the environmental factor uncertainty group, and the model factor uncertainty group include a plurality of uncertainty factors that cause the uncertainty of the GNSS-based location estimation grouped according to the occurrence of the uncertainty. It is also achieved by a method of measuring the uncertainty of the estimate.

Here, the sensor factor uncertainty group is grouped with uncertainty factors caused by a GNSS sensor itself, and the environmental factor uncertainty group is grouped with uncertainty factors caused by an obstacle between a satellite and the GNSS sensor, and the model factor. Uncertainty groups may be grouped with uncertainty factors whose uncertainty changes depending on the type of correction model applied to the GNSS-based position estimation.

And the uncertainty factors grouped into the sensor factor uncertainty group include a hardware uncertainty factor and a software uncertainty factor of the GNSS sensor; The uncertainty factors grouped into the environmental factor uncertainty group include a multipath effect factor and a foliage attenuation factor; Uncertainty factors grouped into the model factor uncertainty group may include an Ephemeris error factor, a satellite clock error factor, an ionospheric effect factor, and a tropospheric effect factor.

Here, in the above steps (a) and (b), the condition that the uncertainty of the model factor uncertainty group does not occur is that a correction model having the model factor uncertainty value of '0' is applied to the GNSS-based position estimation. Can be satisfied.

Also, in the step (A), the condition in which the uncertainty of the model factor uncertainty group does not occur is satisfied by temporarily applying a correction model having the model factor uncertainty value '0' to the GNSS-based position estimation; In the step (B), the model factor uncertainty value may be measured by applying a correction model actually applied to the GNSS based position estimation.

And a condition in which the uncertainty of the environmental factor uncertainty group does not occur in steps (a) and (A) is satisfied by measuring at a position where no obstacle exists between the satellite and the GNSS sensor; The environmental factor uncertainty value may be measured by a deviation between the measured position and the actual position of the GNSS-based position estimate measured at the position where the obstacle does not exist.

In the steps (b) and (C), each driving position is determined by the deviation between the measured position and the actual position of the GNSS-based position estimation measured at each driving position through the driving process on the actual driving road. The environmental factor uncertainty value of can be measured.

In addition, the environmental factor uncertainty value may be mapped corresponding to a location on the map of the driving road and registered in the form of an environmental factor uncertainty map.

In addition, the uncertainty value of the GNSS-based position estimation is represented by an equation

(σ _G is the uncertainty value of the GNSS based position estimate, DOP is Delusion of Precision, AS is the model factor uncertainty value, RE is the sensor factor uncertainty value, and LC is the environmental factor uncertainty value) Can be expressed.

According to the present invention, according to the present invention, it is possible to accurately measure the uncertainty generated in the position measurement of the GNSS sensor used for the autonomous driving vehicle using the GNSS sensor or the autonomous driving of the outdoor mobile robot and accurately reflect the accurate position estimation of the autonomous driving vehicle. A method of measuring the uncertainty of a GNSS based position estimation is provided.

1 is a view for explaining a method of measuring the uncertainty of the position estimation based on the GNSS sensor according to the first embodiment of the present invention,

2 is a view for explaining a method of measuring the uncertainty of the position estimation based on the GNSS sensor according to the second embodiment of the present invention;

3 to 12 are views for explaining the effect of the method of measuring the uncertainty of the position estimation based on the GNSS sensor according to the present invention.

The present invention relates to a method for measuring uncertainty of position estimation based on GNSS, comprising the steps of: (A) measuring the sensor factor uncertainty value of the sensor factor uncertainty group under conditions in which the environmental factor uncertainty group and the model factor uncertainty group do not occur; (B) the model factor uncertainty value of the model factor uncertainty group is measured under a condition in which the environmental factor uncertainty group does not occur, and the sensor factor uncertainty value is applied and measured, and (C) the environmental factor uncertainty group An environmental factor uncertainty value is measured by applying the sensor factor uncertainty value and the model factor uncertainty value, and (D) based on the GNSS based on the sensor factor uncertainty value, the model factor uncertainty value and the environmental factor uncertainty value A step of measuring an uncertainty value of a position estimate of ; The sensor factor uncertainty group, the environmental factor uncertainty group, and the model factor uncertainty group are characterized in that a plurality of uncertainty factors causing uncertainty in the GNSS-based location estimation are grouped according to the cause of the uncertainty.

Hereinafter, with reference to the accompanying drawings will be described in detail an embodiment according to the present invention.

1 is a view for explaining a method of measuring the uncertainty of the position estimation based on the GNSS sensor according to the first embodiment of the present invention.

First, in the method of measuring the uncertainty of the position estimation based on the GNSS sensor according to the present invention, a plurality of uncertainty factors are grouped into a plurality of uncertainty groups (S10). In the present invention, the uncertainty factors include a multipath effect factor, a foliage attenuation factor, an ephemeris error factor, a satellite clock error factor, an ionospheric effect factor, and an atmospheric layer. Examples include Tropospheric effect factors, hardware and software factors of the GNSS sensor, and the DOP (Delusion of Precision).

As described above, the DOP is calculated by the GNSS sensor in the process of position estimation, and is obtained in real time using NMEA 0183 regulation information in an environment using the GNSS sensor.

Uncertainty factors other than DOP are grouped according to the cause of the uncertainty. More specifically, the uncertainty factors caused by the GNSS sensor itself are grouped into sensor factor uncertainty groups. Uncertainty factors that make up the sensor factor uncertainty group may include hardware uncertainty factors and software uncertainty factors of the GNSS sensor.

Uncertainty factors caused by obstacles between GNSS sensors and satellites are grouped into environmental factor uncertainty groups. Here, the uncertainty factors forming the environmental factor uncertainty group may include a multipath effect factor and a foliage attenuation factor.

The remaining uncertainty factors, except the uncertainty factors belonging to the sensor factor uncertainty group and the environmental factor uncertainty group, are grouped into the model factor uncertainty group. It may include an error factor (Satellite clock error), an ionospheric effect factor, and a tropospheric effect factor. Model factor uncertainty is grouped with uncertainty factors whose uncertainty values vary depending on the type of correction model applied to the GNSS-based position estimation.

When the uncertainty factors are grouped into a sensor factor uncertainty group, an environmental factor uncertainty group, and a model factor uncertainty group, a calibration model for calibration of the GNSS sensor is registered (S11). Here, in the correction model according to the first embodiment of the present invention, a correction model having an uncertainty value of the model factor uncertainty group as '0' is registered, which will be described later.

With the grouping of the uncertainty factors and the correction model registered, the sensor factor uncertainty value for the sensor factor uncertainty group is measured (S12). Here, the sensor factor uncertainty value for the sensor factor uncertainty group is measured under the condition that the uncertainty of the remaining uncertainty group, that is, the environmental factor uncertainty group and the model factor uncertainty group, does not occur.

Model factors Uncertainty factors that make up the uncertainty group, Ephemeris error factor, Satellite clock error factor, Ionospheric effect factor and Tropospheric effect factor, are model factors according to the calibration model. Uncertainty values may change. In the present invention, an example in which a correction model having a model factor uncertainty value of '0' is applied to a GNSS sensor is satisfied under the condition that the uncertainty of the model factor uncertainty group does not occur, and the model factor uncertainty value is set to '0'. As an example, a real time kinematic correction model is applied to the model.

In addition, the condition in which the uncertainty factors of the environmental factor uncertainty group are not generated is measured at the position where there is no obstacle between the satellite and the GNSS sensor, so that the environmental factor uncertainty value of the environmental factor uncertainty group can be set to '0'.

In the present invention, the uncertainty value of the position estimation based on GNSS may be expressed as shown in [Equation 2].

[Equation 2]

In Equation 2, σ _G is an uncertainty value of the GNSS-based position estimation, DOP is a DOP (Delusion of Precision), AS is a model factor uncertainty value, RE is a sensor factor uncertainty value, and LC is an environmental factor Uncertainty value.

Therefore, under conditions where uncertainty belonging to the environmental factor uncertainty group and the model factor uncertainty group does not occur, i.e., using a real time kinematic calibration model, where no obstacles exist between the satellite and the GNSS sensor, For example, when measured on a rooftop of an open area or building without obstacles, Equation 2 may be expressed as Equation 3.

[Equation 3]

Using Equation 3, the sensor factor uncertainty value can be measured by the deviation between the measured position of the GNSS-based position estimation and the actual position using the data measured for a long time at the position where there is no obstacle.

As described above, when the sensor factor uncertainty value is measured, the environmental factor uncertainty value of the environmental factor uncertainty group is measured using the sensor factor uncertainty value. The environmental factor uncertainty value is measured under the condition that the uncertainty of the model factor uncertainty group does not occur. As described above, the model factor uncertainty value of the model factor uncertainty group is determined by the real time kinematic correction model. It can be estimated as '0' through use.

The environmental factor uncertainty group, as described above, depends on the driving position of the autonomous vehicle as the uncertainty that occurs when an obstacle, for example, a reflective object such as a building or glass, exists between the phase and the GNSS sensor.

Therefore, in measuring the environmental factor uncertainty value of the environmental factor uncertainty group, while the autonomous vehicle travels on the driving road (S13), the environmental factor uncertainty value is measured for each driving position (S14).

That is, the environmental factor uncertainty value at each driving position is measured by the measurement position of the GNSS-based position estimation measured at each driving position through the driving process on the actual driving road of the autonomous vehicle and the deviation between the actual position. . At this time, the sensor factor uncertainty value in [Equation 2] is measured through the above-described measurement process, and the model factor uncertainty value is estimated as '0' through the application of a real time kinematic correction model. do.

As described above, the environmental factor uncertainty value measured at each driving position is mapped corresponding to the position on the map of the driving road, and is generated and registered in the form of an environmental factor uncertainty map (S15).

Through the above process, when the registration of the sensor factor uncertainty value and the environmental factor uncertainty value is completed, the environmental factor uncertainty value, the sensor factor uncertainty value, and the DOP calculated in real time in the driving process of the autonomous vehicle are calculated. The uncertainty value at the corresponding position is calculated using the value, and the calculated uncertainty value is applicable to the position estimation of the autonomous vehicle.

Hereinafter, a method of measuring the uncertainty of the GNSS based position estimation according to the second embodiment of the present invention will be described in detail with reference to FIG. 2.

First, a plurality of uncertainty factors are grouped into a plurality of uncertainty groups (S20). As in the first embodiment, the uncertainty factor is a multipath effect factor, a foliage attenuation factor, an ephemeris error factor, a satellite clock error factor, an ionospheric effect. Examples include effects factors, tropospheric effects factors, hardware and software factors of the GNSS sensor, and the inclusion of precision (DOP).

In addition, as in the first embodiment, the DOP is calculated in the process of position estimation by the GNSS sensor, and obtained in real time with NMEA 0183 regulation information in an environment using the GNSS sensor.

The uncertainty factor group is grouped into a sensor factor uncertainty group, an environmental factor uncertainty group, and a model factor uncertainty group, and the uncertainty factors constituting each uncertainty group are the same as in the first embodiment.

As such, when the uncertainty factors are grouped into a sensor factor uncertainty group, an environmental factor uncertainty group, and a model factor uncertainty group, a calibration model for calibration of the GNSS sensor is registered (S21). The correction model according to the second embodiment of the present invention is classified into a correction model applied to the calculation of the sensor factor uncertainty value, and a correction model that is actually applied to the GNSS-based position estimation during driving of the actual autonomous vehicle. Will be described later.

With the grouping of the uncertainty factors and the correction model registered, the sensor factor uncertainty value for the sensor factor uncertainty group is measured (S22). Here, the sensor factor uncertainty value for the sensor factor uncertainty group is measured under the condition that the uncertainty of the remaining uncertainty group, that is, the environmental factor uncertainty group and the model factor uncertainty group, does not occur.

Here, as in the first embodiment, an example in which a correction model having a model factor uncertainty value of '0' is applied to the GNSS sensor and is satisfied under the condition that the uncertainty of the model factor uncertainty group does not occur, the model factor uncertainty value As an example, a real time kinematic correction model is applied as a correction model with '0'.

In addition, the condition that the uncertainty factors of the environmental factor uncertainty group did not occur is measured at the position where there is no obstacle between the satellite and the GNSS sensor, for example, at the open ground or on the roof of the building, thereby determining the environmental factor uncertainty value of the environmental factor uncertainty group. Can be set to '0'.

Through this, the equation expressed as shown in [Equation 3] is used to determine the sensor factor uncertainty value by the deviation between the measured position of the GNSS-based position estimation and the actual position using the data measured for a long time at the position without obstacles. Measurement is possible.

As described above, when the sensor factor uncertainty value is measured, the model factor uncertainty value of the model factor uncertainty group is measured (S23). Here, the model factor uncertainty value is measured under conditions where an environmental factor uncertainty group does not occur, for example, as described above, on a roof of a open area or a building, and is measured by applying a sensor factor uncertainty value measured in step S22.

In more detail, when the autonomous vehicle is measured under the condition that the environmental factor uncertainty value is '0' with the correction model applied to the GNSS-based position estimation during actual driving, [Equation 2] is [Equation 2] 4].

[Equation 4]

That is, since the uncertainty factors belonging to the model factor uncertainty group vary according to the calibration model, the model factor uncertainty value is determined by using the deviation between the measured value and the actual position while the correction model applied to the autonomous vehicle is actually applied. Measurement is possible.

As described above, when the measurement of the sensor factor uncertainty value and the model factor uncertainty value is completed, the environmental factor uncertainty value of the environmental factor uncertainty group is measured (S25). As in the first embodiment, in measuring the environmental factor uncertainty value of the environmental factor uncertainty group, while the autonomous vehicle travels on the driving road (S24), the environmental factor uncertainty value is measured for each driving position (S25). .

That is, the environmental factor uncertainty value at each driving position is measured by the measurement position of the GNSS-based position estimation measured at each driving position through the driving process on the actual driving road of the autonomous vehicle and the deviation between the actual position. . At this time, since the sensor factor uncertainty value and the model factor uncertainty value are measured in [Equation 2], the environmental factor uncertainty value can be finally calculated.

Hereinafter, the effects of the method of measuring the uncertainty of the GNSS based position estimation according to the present invention will be described with reference to FIGS. 3 to 12. FIG. 3 is a satellite photograph of an actual driving road for experimenting a method of measuring uncertainty of GNSS-based location estimation according to the present invention, and a section marked in white is a path to which the experiment is applied. The actual driving road has a characteristic of the urban environment in which the height of the surrounding buildings is high and interference of satellite signal reception is generated.

FIG. 4A illustrates an environmental factor uncertainty map by repeatedly measuring the environmental factor uncertainty value by repeatedly driving the driving road shown in FIG. 3 and mapping the measured environmental factor uncertainty value on a map. On the environmental factor uncertainty map, the red color, ie, the higher the gradation score, indicates that the uncertainty is greater.

FIG. 4B is a diagram showing measured values (GNSS data) and actual position (Ground truth) by the GNSS sensor. As shown in (b) of FIG. 4, it can be seen that the environmental factor uncertainty value is large as shown in FIG. 4 (a) in a region where the error of the measured value of the GNSS sensor is large.

5 to 7 are diagrams comparing the uncertainty value measured by the method of measuring the uncertainty of the GNSS-based position estimation according to the present invention, the uncertainty value calculated by the existing GNSS sensor, and the actual position error.

In FIGS. 5 to 7, two sigma of the uncertainty model is compared, and a red line (2σ of the proposed method) is a two sigma value by the method of measuring the uncertainty of the GNSS sensor-based position estimation according to the present invention. The green line (2σ receiver estimated) is the 2-sigma value output by the GNSS sensor, and the blue line (Position error) is the position error that actually occurred.

FIG. 5A is a graph of the entire section shown in FIG. 3, and FIG. 5B is an enlarged view of region A of FIG. 5A. Referring to FIG. 5 (b), it can be seen that there is no significant difference between the method according to the present invention and the existing model in a section where the position error is less than 0.3 m, but the uncertainty value according to the output of the existing GNSS sensor It can be seen that a large measurement (see the circle in FIG. 5B) occurs regardless of the magnitude of the actual position error.

6A and 6B are enlarged views of region B of FIG. 5. As described above with reference to FIG. 6A, where the position error is largely generated, it can be seen that the method according to the present invention wraps the position error and expresses the uncertainty well, and according to the output of the GNSS sensor. It can be seen that the uncertainty value is much larger than the actual position error in the multiple intervals.

In addition, referring to Figure 6 (b), when the position error is 10m or more, it can be seen that the uncertainty value according to the output of the GNSS sensor is provided with a value smaller than the actual error, on the other hand, the method according to the present invention The uncertainty value of is given as a little more than the actual position error, so it can be confirmed that it is close to the actual position error.

FIG. 7 is an enlarged view of region C of FIG. 5. Referring to FIG. 7, it can be seen that the C region or the same aspect as the A and B regions described above are shown, and the actual position error is well reflected as the uncertainty value of the method according to the present invention.

8 to 12 are diagrams showing experimental results of position estimation in which an odometry value and a correction model are fused using an uncertainty value provided through a method of measuring uncertainty of a position estimation based on GNSS according to the present invention.

8 is a diagram illustrating an entire section, and FIGS. 9, 10, 11, and 12 are enlarged views of regions A, B, C, and D of FIG. 8, respectively. 8 through 12, the line 'EKF (proposed)' is the result of the model obtained by applying the method according to the present invention, and the line 'EKF (receiver)' is the result of the model obtained by applying the output of the GNSS sensor. , The line 'GNSS data' is the measured value of the GNSS sensor, and the line 'Ground truth' is the actual position.

Referring to FIG. 9, as a result of the position estimation in the region A of FIG. 8, when the method according to the present invention is compared with the method using the output of the existing GNSS sensor, the GNSS sensor does not show a significant difference when it is precise. You can check it.

Since the width of the general driving lane is 3.2m, if the estimation error exceeds 1.6m, the position is estimated as being in a different lane. Therefore, it is considered as a critical point. All were 100%.

Referring to FIG. 10, when the position estimation error is large, the conventional method calculates a low uncertainty compared to the GNSS position estimation error and shows a very large error, but the method according to the present invention shows a large uncertainty value. The precise position estimation result was shown.

Referring to FIG. 11, although the position estimation error is small, as in the area A, there is no significant difference between the method according to the present invention and the existing method, but it can be confirmed that the method according to the present invention is more precise.

Referring to FIG. 12, it can be seen that the method according to the present invention exhibits better performance as in the region B with a large GNSS position measurement error.

FIG. 13 is a diagram illustrating an average square root error of a location estimation model obtained by applying an output of a conventional GNSS sensor and a method of measuring uncertainty of a location estimation based on GNSS according to the present invention. Was applied. 'EKF proposed' is the mean square root error of the method of measuring the uncertainty according to the present invention, 'EKF conventional' is the mean square root error of the position estimation model obtained by applying the output of the conventional GNSS sensor. As shown in FIG. 13, it can be seen that in all circumstances, the method according to the present invention reduces the mean square root error.

14 to 21 show the numbers shown on the horizontal axis of FIG. 14 to 21 (a) is a path in which a section marked in white as a satellite image of an actual driving road is applied to an experiment, and the number indicated is the experiment number of FIG. 13. 14 to 21 (c) show an environmental factor uncertainty map by mapping environmental factor uncertainty values measured by repeatedly driving the driving roads of FIGS. 14 to 21 (a) on a map. The closer to blue, i.e., the smaller the gradient value, the smaller the uncertainty, and the closer to yellow, that is, the larger the gradient value, the greater the uncertainty size. 14 (B) is a diagram of a part of an actual driving environment.

In FIGS. 14 to 21 (d), the line 'EKF propoesed' is the result of the model obtained by applying the method according to the present invention, and the line 'EKF conventional' is the result of the model to which the uncertainty value provided by the GNSS sensor is applied as it is. Value, the line 'GNSS' is the measured value of the GNSS sensor, and the line 'Ground truth' is the actual position.

It can be seen that the position estimation model to which the method of measuring uncertainty according to the present invention is applied is closer to the ground truth, which is the actual position, than the method of using the uncertainty provided by the GNSS sensor, and the reliability of the GNSS is lowered, resulting in good position estimation. You can check.

Referring to (b) of FIG. 18, in the experiment environment No. 5 of FIG. 13, an elevated road exists around the experiment environment. Referring to FIG. 20 (b), the experiment environment of FIG. A viaduct exists. Both environments cover the GNSS sensor and thus affect the reception of the GNSS sensor. As shown in FIG. 3, it can be seen that the representative environment deteriorates the accuracy of the GNSS sensor in actual driving together with an environment in which a high building exists around.

Although the preferred embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

The present invention can be used for autonomous driving of an autonomous vehicle or an outdoor mobile robot using a GNSS sensor as a method for measuring the uncertainty of GNSS-based position estimation.

Claims

In the method of measuring the uncertainty of the position estimation based on GNSS,

(a) the sensor factor uncertainty value of the sensor factor uncertainty group is measured under conditions in which no environmental factor uncertainty group and model factor uncertainty group occur;

(b) measuring the environmental factor uncertainty value of the environmental factor uncertainty group under the condition that the model factor uncertainty group does not occur, wherein the sensor factor uncertainty value is applied and measured;

(c) measuring an uncertainty value of the GNSS based position estimate based on the sensor factor uncertainty value and the environmental factor uncertainty value;

The sensor factor uncertainty group, the environmental factor uncertainty group, and the model factor uncertainty group include a plurality of uncertainty factors that cause the uncertainty of the GNSS-based location estimation grouped according to the occurrence of the uncertainty. How to measure uncertainty in estimation.
The method of claim 1,

The sensor factor uncertainty group is grouped with the uncertainty factors caused by the GNSS sensor itself,

The environmental factor uncertainty group is grouped with uncertainty factors caused by an obstacle between a satellite and the GNSS sensor;

The uncertainty of the GNSS-based position estimation is grouped into the model factor uncertainty group, the uncertainty factors that change the uncertainty according to the type of the correction model applied to the GNSS-based position estimation.
The method of claim 2,

The uncertainty factors grouped into the sensor factor uncertainty group include a hardware uncertainty factor and a software uncertainty factor of the GNSS sensor;

The uncertainty factors grouped into the environmental factor uncertainty group include a multipath effect factor and a foliage attenuation factor;

The uncertainty factors grouped into the model factor uncertainty group include an epoch error factor, a satellite clock error factor, an ionospheric effect factor, and a tropospheric effect factor. How to measure the uncertainty of GNSS-based location estimation.
The method of claim 3,

The condition in which the uncertainty of the model factor uncertainty group does not occur in steps (a) and (b) is satisfied when a correction model having the model factor uncertainty value of '0' is applied to the GNSS-based position estimation. Method for measuring the uncertainty of the position estimation based on GNSS characterized in that.
The method of claim 3,

The condition in which the uncertainty of the environmental factor uncertainty group does not occur in step (a) is satisfied by measuring at a position where no obstacle exists between the satellite and the GNSS sensor;

And the uncertainty value of the GNSS-based position estimate is measured by the deviation between the measured position and the actual position of the GNSS-based position estimate measured at the position where the obstacle is not present.
The method of claim 3,

In step (b)

GNSS characterized in that the environmental factor uncertainty value at each driving position is measured by the deviation between the measured position and the actual position of the GNSS-based position estimation measured at each driving position through the driving process on the actual driving road. A method of measuring the uncertainty of based location estimation.
The method of claim 6,

The environmental factor uncertainty value is mapped to a location on the map of the driving road and registered in the form of an environmental factor uncertainty map, the method of measuring the uncertainty of the location estimation based on GNSS.
The method of claim 3,

The uncertainty value of the GNSS based position estimation is represented by an equation

(σ G is the uncertainty value of the GNSS based position estimate, DOP is Delusion of Precision, AS is the model factor uncertainty value, RE is the sensor factor uncertainty value, and LC is the environmental factor uncertainty value)
In the method of measuring the uncertainty of the position estimation based on GNSS,

(A) the sensor factor uncertainty value of the sensor factor uncertainty group is measured under conditions where no environmental factor uncertainty group and model factor uncertainty group occur;

(B) measuring the model factor uncertainty value of the model factor uncertainty group under the condition that the environmental factor uncertainty group does not occur, wherein the sensor factor uncertainty value is applied and measured;

(C) measuring the environmental factor uncertainty value of the environmental factor uncertainty group by applying the sensor factor uncertainty value and the model factor uncertainty value;

(D) measuring an uncertainty value of the GNSS based position estimate based on the sensor factor uncertainty value, the model factor uncertainty value, and the environmental factor uncertainty value;

The sensor factor uncertainty group, the environmental factor uncertainty group, and the model factor uncertainty group include a plurality of uncertainty factors that cause the uncertainty of the GNSS-based location estimation grouped according to the occurrence of the uncertainty. How to measure uncertainty in estimation.
The method of claim 9,

The sensor factor uncertainty group is grouped with the uncertainty factors caused by the GNSS sensor itself,

The environmental factor uncertainty group is grouped with uncertainty factors caused by an obstacle between a satellite and the GNSS sensor;

The uncertainty of the GNSS-based position estimation is grouped into the model factor uncertainty group, the uncertainty factors that change the uncertainty according to the type of the correction model applied to the GNSS-based position estimation.
The method of claim 10,

The uncertainty factors grouped into the sensor factor uncertainty group include a hardware uncertainty factor and a software uncertainty factor of the GNSS sensor;

The uncertainty factors grouped into the environmental factor uncertainty group include a multipath effect factor and a foliage attenuation factor;

The uncertainty factors grouped into the model factor uncertainty group include an epoch error factor, a satellite clock error factor, an ionospheric effect factor, and a tropospheric effect factor. How to measure the uncertainty of GNSS-based location estimation.
The method of claim 11,

A condition in which the uncertainty of the model factor uncertainty group does not occur in step (A) is satisfied by temporarily applying a correction model having the model factor uncertainty value '0' to the GNSS-based position estimation;

In the step (B), the uncertainty value of the model factor is measured by applying a correction model actually applied to the GNSS-based position estimation.
The method of claim 11,

The condition in which the uncertainty of the environmental factor uncertainty group does not occur in step (A) is satisfied by measuring at a position where there is no obstacle between the satellite and the GNSS sensor;

And the uncertainty value of the GNSS-based position estimate is measured by the deviation between the measured position and the actual position of the GNSS-based position estimate measured at the position where the obstacle is not present.
The method of claim 11,

In the step (C)

GNSS characterized in that the environmental factor uncertainty value at each driving position is measured by the deviation between the measured position and the actual position of the GNSS-based position estimation measured at each driving position through the driving process on the actual driving road. A method of measuring the uncertainty of based location estimation.
The method of claim 14,

The environmental factor uncertainty value is mapped to a location on the map of the driving road and registered in the form of an environmental factor uncertainty map, the method of measuring the uncertainty of the location estimation based on GNSS.
The method of claim 11,

The uncertainty value of the GNSS based position estimation is represented by an equation

(σ G is the uncertainty value of the GNSS based position estimate, DOP is Delusion of Precision, AS is the model factor uncertainty value, RE is the sensor factor uncertainty value, and LC is the environmental factor uncertainty value)

Method for measuring the uncertainty of the position estimation based on GNSS, characterized in that represented by.