KR20200063538A - Method for self-diagnosing localization status and autonomous mobile robot carrying out the same - Google Patents

Abstract

The present invention relates to a method for self-diagnosing a position estimation state, (a) a distance error index, a direction angle error index, and a machine learning algorithm in which position estimation success for the distance error index and the direction angle error index is pre-registered. (B) measuring a plurality of distance measurements at the current position of the autonomous driving robot, and (c) estimating an estimated direction angle at the current position of the autonomous driving robot. Step, (d) calculating a distance error index at the current position based on the plurality of distance measurement values, and (e) calculating a direction angle error index at the current position based on the estimated direction angle. W, (f) The distance error index and the direction angle error index calculated in steps (d) and (e), respectively, are input to the machine learning algorithm to output whether the position estimation at the current position is successful. It includes; In the steps (a) and (d), the distance error index is calculated based on an error average value of the upper n% distance measurement values having high reliability among a plurality of distance measurement values measured at the current location, and the (a) ) And in the step (e), the direction angle error index is calculated based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position.

Description

METHOD FOR SELF-DIAGNOSING LOCALIZATION STATUS AND AUTONOMOUS MOBILE ROBOT CARRYING OUT THE SAME

The present invention relates to a method for diagnosing a position estimation state and an autonomous driving robot performing the same, and more specifically, a method for diagnosing a position estimation state capable of self-diagnosing whether the position estimation state of the autonomous driving robot is successful or failed when driving. It relates to an autonomous driving robot that performs this.

Reliable position estimation technology is essential for autonomous driving robots. The autonomous driving robot travels from the starting position to the target position based on the position estimation result. Therefore, reliable location estimation results must be provided for successful path creation and motion control.

So far, many studies have evaluated location estimation performance mainly based on location estimation error. However, if the position estimation error is maintained to a certain level or less, in many cases, the position estimation error does not significantly affect the performance of autonomous driving. For example, if the relative coordinate system of the autonomous driving robot can accurately detect an obstacle or a dangerous terrain, proper motion control is possible even if the position estimation error is somewhat inaccurate.

In actual use, it is more important to judge whether the currently estimated pose of the autonomous driving robot, that is, whether the position and the direction angle is reliable or not. In other words, it is important to judge the position estimation status. If the estimated pose is not reliable, the estimated pose should not be used to determine the driving behavior of the autonomous driving robot. Otherwise, the self-driving robot may fail the mission or behave abnormally.

On the other hand, position estimation technology using a distance sensor such as a lidar sensor has shown robust and reliable position estimation results in many environments. Nevertheless, location estimation failures sometimes occur when using location sensor based location estimation techniques. Positioning failures are caused by a variety of causes, such as extreme environmental changes, insufficient natural markings, and slippage of the wheels.

It is difficult to develop a location estimation technique that works robustly in all situations and environments. Therefore, there is a need to develop a technology capable of diagnosing the current position estimation state of the autonomous driving robot itself.

There have been various studies related to the diagnosis of the position estimation status. Moon et al. "Observation likelihood model design and failure recovery scheme toward reliable localization of mobile robots (Int. J. Adv. Robot. Syst., vol. 7, no. 4, pp. 113??122, 2010.)" Estimation used a matching error to detect the failure to estimate the position due to wheel slip. If the current matching error exceeds a certain range from the distribution of past matching errors using the sliding window technique, the position estimation state is classified as failed. This method takes into account the property that the matching error increases momentarily when the wheel is slipped.

Lee et al., “Discrete-status-based localization for indoor service robots (IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1737??1746, 2006.)” has an error of 10% or more. The distance measurement value was defined as an erroneous measurement value, and the reliability of the position estimation was defined based on the ratio of the erroneous distance measurement value. If the reliability is below the limit, the estimated pose is considered unreliable and is not reflected in the update process.

In Zhang et al.'s "Self-adaptive Monte Carlo Localization for Mobile Robots Using Range Finders (Robotica, vol. 30, no. 2, pp. 229-244, 2012.)" Detect. When the maximum probability of samples is below a threshold, a part of the local sample is converted into a global sample.

In Fujii et al.'s paper "Detection of localization failure using logistic regression (Proc. IEEE Int. Conf. Intell. Robots Syst., 2015.)", location estimation failure was classified by logistic regression using the maximum probability and standard deviation of samples. . In addition, a hybrid position estimation technique using the probability of position estimation failure calculated by the logistic function is presented.

The location estimation status was also diagnosed for the purpose of updating the map. For example, in Sun et al.'s paper "Towards Effective Localization in Dynamic Environments (Proc. IEEE Int. Conf. Intell. Robots Syst., 2016.)" Matching score) was used to diagnose location estimation. The map is updated when the matching score exceeds the limit.

As described above, the existing studies have successfully diagnosed the target position estimation state using respective indicators. However, existing indicators can be highly dependent on the location estimation algorithm. For example, the matching error defined in a paper such as Moon and the matching score defined in a paper such as Sun are indicators that can be obtained as a result of each position estimation algorithm. In addition, the maximum probability or the size of the covariance of the sample may be less related to the position estimation result depending on the design strategy or environmental conditions of the sensor model.

Therefore, if the autonomous driving robot can self-diagnose whether its position estimation status is success or failure, regardless of the design strategy of the sensor model, environmental conditions, and the type of location estimation algorithm, it can guarantee a more accurate driving. Would be desirable.

Accordingly, the present invention was devised to solve the above problems, a method for diagnosing a position estimation state capable of self-diagnosing whether a position estimation state of a self-driving robot is a success or failure and an autonomous driving robot performing the same. The purpose is to provide.

In accordance with the present invention, in the self-diagnostic method of position estimation state, (a) the distance error indicator, the direction angle error indicator, and the success or failure of the position estimation for the distance error indicator and the direction angle error indicator are registered Registered and learned as learning data of the machine learning algorithm, (b) measuring a plurality of distance measurements at the current position of the autonomous driving robot, and (c) the estimated direction at the current position of the autonomous driving robot. The step of estimating the angle, and (d) calculating the distance error index at the current position based on the plurality of distance measurements, and (e) the direction angle error index at the current position based on the estimated direction angle. Is calculated, and (f) the distance error index and the direction angle error index calculated in steps (d) and (e), respectively, are input to the machine learning algorithm, and the position estimation at the current position is successful. It includes a step of outputting; In the steps (a) and (d), the distance error index is calculated based on the error average value of the high n% reliable distance measurement values among the plurality of distance measurement values measured at the current location, and the (a) The self-diagnosing method of the position estimation state of the autonomous driving robot, characterized in that the direction angle error index is calculated based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position in steps (e) and (e). Is achieved by.

In addition, the above object, in accordance with another embodiment of the present invention, in the self-driving robot performing a self-diagnosis position estimation state, and controls the driving based on the distance sensor and the distance measurement value measured by the distance sensor, The machine learning algorithm includes a pre-registered main processor; (a) registering and learning whether a distance error indicator, a direction angle error indicator, and a position error success for the distance error indicator and the direction angle error indicator are registered in the main processor as learning data of the machine learning algorithm; (b) measuring a plurality of distance measurements at the current position by the distance sensor, (c) estimating an estimated direction angle at the current position by the main processor, and (d) at the main processor. Calculating a distance error index at the current position based on a plurality of the distance measurement values, and (e) calculating a direction angle error index at the current position based on the estimated direction angle by the main processor. Wow, (f) the distance error index and the direction angle error index respectively calculated in the steps (d) and (e) by the main processor are input to the machine learning algorithm to successfully estimate the position at the current position. Whether or not a step of outputting is performed, the position estimation state is self-diagnosed; The main processor calculates the distance error index in steps (a) and (d) based on the error average value of the high n% reliable distance measurement values among the plurality of distance measurement values measured at the current location. In step (a) and (e), the position estimation state self-diagnosis is characterized by calculating the direction angle error index based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position. It can also be achieved by performing autonomous driving robots.

Here, the distance error index is an equation

는 상기 거리 오차 지표이고,

는 상기 오차 평균값으로 수학식

(여기서,

는 시간 t에서 측정된 거리 센서의 i번째 거리 측정값이고,

는 상기 자율주행로봇에 등록된 지도로부터 계산된

의 레퍼런스 거리이고,

는 시간 t에서 측정된 복수의 거리 측정값 중 n%에 대응하는 개수이고,

는 시간 t에서 측정된 복수의 거리 측정값 중 n%에 속하는 거리 측정값의 식별번호 i의 집합이다)에 의해 산출되고,

는 시간 t에서의 상기 오차 평균값에 대한 가중치이다)에 의해 산출될 수 있다.(here,

Is the distance error indicator,

Is the equation of the error average

(here,

Is the i-th distance measurement of the distance sensor measured at time t,

Is calculated from the map registered in the autonomous driving robot.

Is the reference distance of

Is a number corresponding to n% of the plurality of distance measurements measured at time t,

Is a set of identification numbers i of distance measurement values belonging to n% among a plurality of distance measurement values measured at time t),

Is a weight for the error average value at time t).

In addition, the direction angle error index is an equation

는 상기 방향각 오차 지표이고,

는 수학식(here,

Is the direction angle error index,

Is a mathematical expression

는 시간 t에서의 상기 추정 방향각이고,

는 시간 t에서의 상기 허용 방향각 범위의 최대 방향각이고,

는 시간 t에서의 상기 허용 방향각 범위의 최소 방향각이다)에 산출되고,

는 시간 t에서의

에 대한 가중치이다)에 의해 산출될 수 있다.(here,

Is the estimated direction angle at time t,

Is the maximum direction angle of the permissible direction angle range at time t,

Is the minimum direction angle of the permissible direction angle range at time t),

At time t

It can be calculated by).

에 의해 산출되고; 상기 최소 방향각은 수학식

에 의해 산출되며;

는 시간 t-s에서 추정된 추정 방향각이고,

에 의해 산출되고,

및

는 각각 시간 t 및 시간 t-s에서 상기 자율주행로봇에 등록된 오도메트리 상의 오도메트리 방향각이고,

는 상기 자율주행로봇의 불확실성을 반영한 매개변수이다.Here, the maximum direction angle is an equation

Is calculated by; The minimum direction angle is an equation

Is calculated by;

Is the estimated direction angle estimated at time ts ,

Is calculated by,

And

Is an odometric direction angle on the odometry registered to the autonomous driving robot at time t and time ts , respectively.

Is a parameter reflecting the uncertainty of the autonomous driving robot.

In addition, the machine learning algorithm may include a binary classification algorithm to which supervised learning is applied.

According to the present invention according to the above configuration, there is provided a method for diagnosing a position estimation state capable of self-diagnosing whether a position estimation state of a self-driving robot is a success or failure and an autonomous driving robot performing the same.

1 is a view for explaining the configuration of an autonomous driving robot according to the present invention,
2 is a flowchart of a method for self-diagnosing a position estimation state of an autonomous driving robot according to the present invention,
3 is a flowchart for explaining the machine learning process of FIG. 2,
4 to 7 are views for explaining a method for self-diagnosing a position estimation state of an autonomous driving robot according to the present invention.

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

1 is a view for explaining the configuration of the autonomous driving robot 100 according to the present invention. Referring to FIG. 1, the autonomous driving robot 100 according to the present invention includes a wheel driving unit 130, a wheel sensing unit 140, a distance sensor 120, and a main processor 110. Here, the power supply unit 150 includes components for driving the autonomous driving robot 100, including a wheel driving unit 130, a wheel sensing unit 140, a distance sensor 120, and a main processor 110. To supply.

The wheel driving unit 130 rotates the wheel installed for driving of the autonomous driving robot 100. When the two-wheel differential method is applied to the autonomous driving robot 100 according to the present invention, it is possible to control the speed and the rotation angle depending on whether the number of revolutions or rotation of each wheel.

The wheel sensing unit 140 helps to detect the number of revolutions of each wheel, and the main processor 110 calculates a moving distance and a rotation angle using the number of revolutions of each wheel sensed by the wheel sensing unit 140 and , Through this, the position on the odometry of the autonomous driving robot 100 is estimated.

The distance sensor 120 measures the distance to the surrounding environment or obstacles. In the present invention, it is assumed that the laser distance sensor 120 is applied as the distance sensor 120, and it is possible to measure a distance measurement value at each angle while rotating in a certain angle range, for example, 1ㅀ unit from the current position. .

The main processor 110 controls the wheel driving unit 130 so that the autonomous driving robot 100 moves along the path based on the measured values of the wheel sensing unit 140 and the distance sensor 120. In the present invention, the machine learning algorithm applied to the position estimation state self-diagnosis is registered in the main processor 110, which will be described later in detail.

Hereinafter, a method for self-diagnosing a position estimation state performed by the main processor 110 of the autonomous driving robot 100 according to the present invention will be described in detail with reference to FIG. 2.

Referring to FIG. 2, in the self-diagnostic method of the position estimation state of the autonomous driving robot 100 according to an embodiment of the present invention, the distance error index, the direction angle error index, and the success or failure of the position estimation of the main processor 110 Machine learning is registered as machine learning algorithm training data (S10).

Here, the distance error index is calculated based on the error average value of the distance measurement value of the upper n% having high reliability among the plurality of distance measurement values measured by the distance sensor 120 at the current position, and the direction angle error index is the current position It is calculated based on the estimated direction angle estimated at and the allowable direction angle range at the current position.

Hereinafter, the machine learning process through the machine learning algorithm in the self-diagnostic method of the position estimation state of the autonomous driving robot 100 according to the present invention will be described in detail.

First, in the driving environment of the autonomous driving robot 100, for example, in the parking lot shown in FIG. 4, whether the distance measurement value at each time, the estimated direction angle, and the position estimation success in the process of driving the autonomous driving robot 100 It is obtained (S20). 4 is a view showing a real example of a parking lot as a driving environment of the autonomous driving robot 100, and whether the location estimation is successful can be determined based on an artificial marker installed at each location of the parking lot. That is, the position and the estimated direction angle according to the distance measurement value obtained in the course of the driving of the autonomous driving robot 100 can confirm the success or failure of the position estimation based on the artificial marker installed in the parking lot.

Referring again to FIG. 3, when distance measurement values at the current position are obtained during the driving process of the autonomous driving robot 100, reliability of the distance measurement value is calculated (S21 ). Referring to FIG. 5, the portion RD indicated by a circle in FIG. 5A shows relatively reliable distance measurement values. The length indicated by the dotted line by a dynamic obstacle 210 such as a person or an unknown static obstacle 220 such as a box becomes a distance error for the corresponding distance measurement. In (a) of FIG. 5, although a large distance error occurs for the entire distance measurement values, the distance error for the relatively reliable distance measurement values is small. That is, since the relatively reliable distance measurement values are accurate, the position estimation can be successfully performed.

On the other hand, as shown in (b) of FIG. 5, the distance error may be large even for relatively reliable distance measurement values. As a result, the position estimation result becomes unreliable.

Accordingly, in the present invention, a distance error index is calculated using an error average value for distance measurement values having a relatively high reliability among all distance measurement values, and a distance measurement value of a top n% having high reliability among the distance measurement values For example, extracting them (S22) and calculating an error average value using the same. Here, if the error average value is small, it can be assumed that the position estimation has been successfully performed, and in the opposite case, it can be assumed that the position estimation has failed. In the present invention, the error average value can be expressed as [Equation 1].

[Equation 1]

는 오차 평균값이고,

는 시간 t에서 측정된 거리 센서(120)의 i번째 거리 측정값이고,

는 자율주행로봇(100)에 등록된 지도로부터 계산된

의 레퍼런스 거리이고,

는 시간 t에서 측정된 복수의 거리 측정값 중 n%에 대응하는 개수이고,

는 시간 t에서 측정된 복수의 거리 측정값 중 n%에 속하는 거리 측정값의 식별번호 i의 집합이다.In [Equation 1]

Is the mean value of the error,

Is the i-th distance measurement of the distance sensor 120 measured at time t,

Is calculated from the map registered in the autonomous driving robot (100)

Is the reference distance of

Is a number corresponding to n% of the plurality of distance measurements measured at time t,

Is a set of identification numbers i of distance measurements belonging to n% of the plurality of distance measurements measured at time t.

Here, if the value of n is large, errors for a number of distance measurements are calculated. As a result, the possibility that it is determined that the position estimation error increases in a situation in which the position estimation error does not actually increase increases. Conversely, when the value of n decreases, it becomes insensitive to the situation where the position estimation error actually increases. In the present invention, it is assumed that the value is set to 20% empirically.

The reliability calculation method according to the present invention is a method for evaluating the reliability of the estimated distance type with respect to the measured distance of the "laser distance sensor 120" disclosed in Korean Patent Registration No. 10-1888295 filed and registered by the applicant of the present application, and It is assumed that the calculation is performed using the "position estimation method of the mobile robot using this", and detailed description thereof is omitted.

가 큰 값을 유지하는 동안, 위치 추정 유차는 계속 증가한다. 게다가, 오차 평균값

이 갑자기 감소하더라도, 위치 추정 오차는 바로 감소하지 않는다. 즉, 위치 추정 오차는 오차 평균값

의 과거 값에도 영향을 받게 되는 바, 본 발명에서는 거리 오차 지표를 [수학식 2]와 같이 산출하는 것을 예로 한다(S23).Here, since the position estimation error accumulates and increases, the error average value

While it maintains a large value, the position estimation error continues to increase. Moreover, the error average

Even if this sudden decrease, the position estimation error does not decrease immediately. That is, the position estimation error is the average error value.

Since it is also influenced by the past value of, it is assumed that the present invention calculates the distance error index as [Equation 2] (S23).

[Equation 2]

는 거리 오차 지표이고,

는 시간 t에서의 오차 평균값에 대한 가중치가 된다. 즉, 본 발명에서는 거리 오차 지표로 거리 추정 오차의 지수 가중 평균(Exponentially weighted moving average)을 사용하는 것을 예로 한다.In Equation 2, at time t

Is the distance error indicator,

Is the weight for the average error value at time t. That is, in the present invention, as an example of using an exponentially weighted moving average of the distance estimation error as a distance error index.

Hereinafter, a method of calculating the direction angle error index in the self-diagnostic method of the position estimation state of the autonomous driving robot 100 according to the present invention will be described (S23). In the present invention, the direction angle error index is calculated based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position.

In more detail, an allowable direction angle range that is an allowable direction angle of the autonomous driving robot 100 may be calculated in consideration of motion uncertainty and angular displacement of the autonomous driving robot 100. If the estimated direction angle is outside the allowable range, it is highly likely that an abnormal position estimation has occurred. In the present invention, it is assumed that the direction angle error index is calculated through [Equation 3].

[Equation 3]

는 방향각 오차 지표이고,

는 [수학식 4]를 통해 산출되고,

는 시간 t에서의

에 대한 가중치이다.In [Equation 3]

Is the direction angle error indicator,

Is calculated through [Equation 4],

At time t

Is the weight for.

[Equation 4]

는 시간 t에서의 추정 방향각이고,

는 시간 t에서의 허용 방향각 범위의 최대 방향각이고,

는 시간 t에서의 허용 방향각 범위의 최소 방향각이다.here,

Is the estimated direction angle at time t,

Is the maximum direction angle of the permissible direction angle range at time t,

Is the minimum direction angle of the permissible direction angle range at time t.

Then, the maximum direction angle and the minimum direction angle are calculated by [Equation 5] and [Equation 6], respectively.

[Equation 5]

[Equation 6]

는 시간 t-s에서 추정된 추정 방향각이고, 수학식

에 의해 산출되고,

및

는 각각 시간 t 및 시간 t-s에서 자율주행로봇(100)에 등록된 오도메트리 상의 오도메트리 방향각이고,

는 자율주행로봇(100)의 불확실성을 반영한 매개변수이다. [수학식 4] 내지 [수학식 6]은 도 6와 같이 도식화할 수 있다.In [Equation 5] and [Equation 6]

Is the estimated direction angle estimated at time ts , the equation

Is calculated by,

And

Is an odometric direction angle on the odometry registered in the autonomous driving robot 100 at time t and time ts , respectively.

Is a parameter reflecting the uncertainty of the autonomous driving robot 100. [Equation 4] to [Equation 6] can be diagrammed as shown in FIG.

Referring again to FIG. 3, when the distance error index and the direction angle error index at the current position are calculated as described above (S23), the calculated distance error index, the direction angle error index, and the position estimation at the current position Success is generated as a learning data set of a machine learning algorithm (S24) and registered in the machine learning algorithm (S26). Here, the distance error index and the direction angle error index constituting the learning data set are registered as input data of the machine learning algorithm, and whether the position estimation is successful is registered as the output data.

The process of generating the learning data set as described above is continuously generated and registered in the driving process of the autonomous driving robot 100, and repeatedly generating the driving process for each time zone so that changes in the driving environment over time can be reflected. Yes. 7 shows an example of an environmental change over time in the parking lot environment shown in FIG. 4. In the case of a parking lot environment, the vehicle parked in the parking area is different for each hour or day, so it is possible to reflect environmental changes well.

When the learning data set is generated through the above process and registered in the machine learning algorithm, the machine learning process proceeds (S10). In the present invention, an example is applied to a binary classification algorithm to which supervised learning is applied as a machine learning algorithm.

Supervised learning is a machine learning algorithm that uses labeled data. That is, in supervised learning, each learning data is a pair of an input value and a desired output. The purpose of supervised learning is to find a function that maps input values to output values.

Since the present invention classifies the position estimation status, that is, whether the position estimation is successful or not, by using the position error indicator and the direction angle error indicator, as a classification problem among regression problems and classification problems in supervised learning Approach.

In order to categorize the position error indicator and the direction angle error indicator according to the actual position estimation state, a rule for classifying the actual position estimation state is required, and in the present invention, the rules for classifying the position estimation state are defined as follows.

-If the position error index and the direction angle error index exceed the prescribed limit, the position estimation is classified as'failure'.

-All non-failure processes are classified as success.

That is, when the position estimation error is below a certain level, in many cases, the position estimation error does not significantly affect autonomous driving performance. Therefore, only when the position estimation error exceeds a predetermined threshold, the position estimation state is classified as'failed'. The limit value of the positioning error can be determined according to the driving purpose or task.

In the present invention, the location estimation status is classified into success and failure. Accordingly, the position error index and the direction angle error index collected in the position estimation success state and the position estimation failure state generated and registered in the method as shown in FIG. 3 are used as learning data of the machine learning algorithm.

As described above, in the machine learning algorithm according to the present invention, a binary classification algorithm using supervised learning is applied, and Altman's book "An Introduction to Kernel and Nearest-Neighbor Nonparametric Regression (The American Statistician, vol. 46, no. 3 , pp. 175-185, 1992.), the logistic regression algorithm disclosed in Pampel's book "Logistic regression (a primer, Sage Publication, 2000.)", a paper by Cristianini et al. "An introduction to support vector machines and other kernel-based learning methods (Cambridge University Press, 2000.).

Referring again to FIG. 2, in the state where the machine learning is completed through the above-described process (S10), the autonomous driving robot 100 is in the current position by the distance sensor 120 in the process of driving the driving environment. The distance measurement values of are measured (S11). Then, the estimated direction angle at the current position is estimated (S12).

Then, the reliability of the distance measurement values is calculated (S13), and the upper n% distance measurement value is extracted based on the calculated reliability (S14). Here, the reliability calculation and extraction of the upper n% distance measurement value correspond to the above-described machine learning process, and detailed description thereof will be omitted.

Then, using the extracted distance measurement value and the estimated direction angle, as described above, the distance error index and the direction angle error index are calculated (S15), and the calculated distance error index and the direction angle error index are machine learning algorithms. By inputting to either one of success and failure is output (S16), it is determined whether the estimation of the current autonomous driving robot 100, that is, the estimation of the position and direction angle is success or failure.

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.

100: autonomous driving robot 110: main processor
120: distance sensor 130: wheel drive
140: wheel detection unit 150: power supply
210: dynamic obstacle 220: static obstacle

Claims (10)

In the position estimation state self-diagnostic method,
(a) the distance error indicator, the direction angle error indicator, and the distance error indicator and the position estimation success with respect to the direction angle error indicator are registered and learned as learning data of a pre-registered machine learning algorithm,
(b) measuring a plurality of distance measurement values at the current position of the autonomous driving robot,
(c) estimating the estimated direction angle at the current position of the autonomous driving robot,
(d) calculating a distance error index at the current position based on the plurality of distance measurements;
(e) calculating a direction angle error index at the current position based on the estimated direction angle;
(f) the distance error index and the direction angle error index calculated in steps (d) and (e) are input to the machine learning algorithm, and outputting whether or not the position estimation is successful at the current position is output. And;
In the steps (a) and (d), the distance error index is calculated based on an error average value of the upper n% distance measurement values having high reliability among a plurality of distance measurement values measured at the current location,
In the step (a) and the step (e), the direction angle error indicator is calculated based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position. According to claim 1,
The distance error index is an equation

(here,

Is the distance error indicator,

Is the equation of the error average

(here,

Is the i-th distance measurement of the distance sensor measured at time t,

Is calculated from the map registered in the autonomous driving robot.

Is the reference distance of

Is a number corresponding to n% of the plurality of distance measurements measured at time t,

Is a set of identification numbers i of distance measurement values belonging to n% among a plurality of distance measurement values measured at time t),

Is a weight for the error average value at time t). According to claim 1,
The direction angle error index is an equation

(here,

Is the direction angle error index,

Is a mathematical expression

(here,

Is the estimated direction angle at time t,

Is the maximum direction angle of the permissible direction angle range at time t,

Is the minimum direction angle of the permissible direction angle range at time t),

At time t

It is calculated by the weight) for position estimation status self-diagnostic method. According to claim 3,
The maximum direction angle is equation

Is calculated by;
The minimum direction angle is an equation

Is calculated by;

Is the estimated direction angle estimated at time ts ,

Is calculated by,

And

Is an odometric direction angle on the odometry registered to the autonomous driving robot at time t and time ts , respectively.

Is a self-diagnostic method for position estimation, characterized in that it is a parameter reflecting the uncertainty of the autonomous driving robot. According to claim 1,
The machine learning algorithm includes a binary classification algorithm to which supervised learning is applied. In the self-driving robot performing a self-diagnosis position estimation state,
Distance sensor,
Controlling a driving based on the distance measurement value measured by the distance sensor, the machine learning algorithm includes a pre-registered main processor;
(a) registering and learning whether a distance error indicator, a direction angle error indicator, and a position error success for the distance error indicator and the direction angle error indicator are registered in the main processor as learning data of the machine learning algorithm;
(b) measuring a plurality of distance measurements at the current position by the distance sensor;
(c) estimating an estimated direction angle at the current position by the main processor,
(d) calculating, by the main processor, a distance error index at the current location based on the plurality of distance measurements;
(e) calculating, by the main processor, a direction angle error index at the current position based on the estimated direction angle;
(f) The distance error index and the direction angle error index respectively calculated in steps (d) and (e) by the main processor are input to the machine learning algorithm to determine whether the position estimation at the current position is successful. The outputting step is performed, and the position estimation status is self-diagnosed;
The main processor
In the steps (a) and (d), the distance error index is calculated based on an error average value of the upper n% distance measurement values having high reliability among a plurality of distance measurement values measured at the current position,
Position estimation state self-diagnosis is performed in steps (a) and (e), wherein the direction angle error index is calculated based on the estimated direction angle estimated at the current position and the allowable direction angle range at the current position. Self-driving robot. The method of claim 6,
The distance error index is an equation

(here,

Is the distance error indicator,

Is the equation of the error average

(here,

Is the i-th distance measurement of the distance sensor measured at time t,

Is calculated from the map registered in the autonomous driving robot.

Is the reference distance of

Is a number corresponding to n% of the plurality of distance measurements measured at time t,

Is a set of identification numbers i of distance measurement values belonging to n% among a plurality of distance measurement values measured at time t),

Is a weight for the error average value at time t). The method of claim 6,
The direction angle error index is an equation

(here,

Is the direction angle error index,

Is a mathematical expression

(here,

Is the estimated direction angle at time t,

Is the maximum direction angle of the permissible direction angle range at time t,

Is the minimum direction angle of the permissible direction angle range at time t),

At time t

Autonomous driving robot that performs self-diagnosis of the position estimation state, characterized in that it is calculated by the weight. The method of claim 8,
The maximum direction angle is equation

Is calculated by;
The minimum direction angle is an equation

Is calculated by;

Is the estimated direction angle estimated at time ts ,

Is calculated by,

And

Is an odometric direction angle on the odometry registered to the autonomous driving robot at time t and time ts , respectively.

Is a parameter reflecting the uncertainty of the autonomous driving robot. The method of claim 6,
The machine learning algorithm includes a binary classification algorithm to which supervised learning is applied.

Images