KR101642286B1 - Heading Orientation Estimation Method Using Pedestrian Characteristics in Indoor Environment - Google Patents

Abstract

The present invention relates to a method for estimating a moving direction of a pedestrian in an indoor environment using a walking property. The method for estimating a moving direction of a pedestrian in an indoor environment using a walking property comprises: a waling data obtaining step (S100) obtaining walking data from a signal transmitted in accordance with a walking motion of a pedestrian from an inertial measurement unit (IMU) mounted on shoes of the pedestrian; a pedestrian position estimating step (S200) of estimating the position of the pedestrian using the obtained walking data; and a moving direction correcting step (S300) of correcting a moving direction of the pedestrian to reduce an estimation error of the pedestrian position estimated by the pedestrian position estimating step (S200). The moving direction correcting step (S300) comprises: a pedestrian exercise type determining step (S310) of determining an exercise type of the pedestrian; and a measurement value renewing step (S320) of renewing a measurement value in accordance with the determined exercise type. According to the composition of the present invention, the method for estimating a moving direction of a pedestrian in an indoor environment using a walking property enables a user to accurately grasp the position of a rescuer or a person who needs to be rescued, in an indoor environment in which it is difficult to receive a navigation signal.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pedestrian movement direction estimation method,

The present invention relates to a method for estimating a moving direction of a pedestrian in a room using gait characteristics, and more particularly to a method for estimating a moving direction of a pedestrian in a room, The present invention relates to a method for estimating a moving direction of a pedestrian using a walking characteristic that enables a user to appropriately respond to an emergency accident or the like.

In recent years, as the structure of a building has become complicated and large-scale, disaster such as a fire or terrorism has occurred in the inside of the building. In order to rescue the person inside the building or to improve the efficiency of the structure, Or the location of the rescue personnel.

For this reason, studies have been actively conducted to track or identify the position of pedestrians or rescue personnel in a room by using a positioning device of a wireless positioning system such as GPS, UWB, or WiFi. However, Since the tracking method has a weak signal receiving intensity in the indoor, the position tracking method using only the position guiding devices of these wireless positioning methods is limited in its application, and accordingly, it is necessary to develop additional means to supplement these methods , And a pedestrian position estimation method using an inertial measurement unit (IMU) disclosed in, for example, Japanese Laid-Open Patent Publication No. 10-2012-0001925.

In estimating the position of a pedestrian using the above-mentioned inertial measurement device (IMU), an inertial measurement device (IMU) is generally installed on the footwear of a pedestrian, and the stride property of the pedestrian is estimated by using the inertial measurement device , And pedestrian dead reckoning (PDR) that estimates the position of a pedestrian using the estimated stride information is combined with GPS to further improve the positioning accuracy even when the GPS signal is unstable indoors In this pedestrian position estimation algorithm (PDR), the pedestrian's stride is estimated by estimating the pedestrian's moving speed while the foot of the pedestrian touches the ground is zero (ZUPT, zero velocity update).

However, when estimating the stride of the pedestrian in the above-described position estimation algorithm (PDR), an inertial measurement device (IMU) composed of three axes of acceleration, gyro and geomagnetism as shown in Figs. 1 and 2, (Or shoes) are attached to the ground (stance phase), and the foot is separated from the ground (lifter, swing phase) when the output pattern is analyzed to detect walking and then using this information The pedestrian's current position is estimated by accumulating the travel distance after estimating the step width. At this time, the pedestrian's walking speed is zero, (Zero velocity correction, ZUPT) to correct the divergence of the error.

Correction of the divergence of the error as described above in the position estimation algorithm (PDR) can be achieved by estimating the position of the pedestrian using the inertial measurement device (IMU) mounted on the shoes, If the acceleration is doubly integrated, the result of the integration will diverge due to the accumulation of the error over time. Therefore, in order to solve such a problem, the moment when the sole reaches the ground (Zero speed correction) is assumed to be zero, so that the integration result is not diverted. In this way, when the foot of the pedestrian touches the ground, the actual It is possible to more precisely calculate the stride while the walker is walking by assuming that the moving speed of the walker is 0, Attribute an algorithm to estimate the position by applying a filter structure Inertial Navigation System (INS) to -Extended Kalman Filter (EKF) - referred to as Zero velocity UpdaTe (ZUPT).

However, if the position of the pedestrian depends only on the inertial sensor without the help of the outside, the gyro's bias generation and the difficulty in estimating the yaw value become different from each other over time, A large error in the position of a pedestrian can be caused. Accordingly, algorithms capable of reducing a moving direction error of a pedestrian by utilizing the condition of indoor to eliminate the moving direction estimation error have been developed. Heuristic Drift Elimination (HDE) algorithm.

In the HDE algorithm, assuming that the walls and corridors of a building are mostly straight lines and parallel or perpendicular to each other, the typical direction of these walls and corridors is defined as the dominant direction, and there is a kitchen scent in the indoor environment Assuming that the calculated azimuth angle is matched to the nearest main direction after calculating the azimuth angle by setting the main directions in the four directions or eight directions, Can converge to almost zero, but can produce excellent results. However, in the exceptional case where the pedestrian does not walk in the main direction or walks in a large circular shape, a large error may be generated. Therefore, Development of a method that can be used is required.

KR 10-2012-0001925 A KR 10-2013-0012714 A KR 10-2013-0116151 A KR 10-2009-0082711 A

Disclosure of Invention Technical Problem [8] Accordingly, the present invention has been made to solve the above-mentioned problem of conventional pedestrian movement direction estimation in indoor, and it is an object of the present invention to accurately estimate a moving direction of a pedestrian in a propagation dark area, The present invention provides a pedestrian movement direction estimating method that can more accurately grasp a current position of a pedestrian located in a room.

According to another aspect of the present invention, there is provided a method for estimating a moving direction of a pedestrian, comprising the steps of: obtaining walking data from a signal transmitted in accordance with a walking operation of a pedestrian from an inertial measurement device mounted on a shoe of a pedestrian; A pedestrian position estimating step of estimating a position of the pedestrian using the obtained walking data; And a moving direction correcting step of correcting the moving direction of the pedestrian so as to reduce the position estimation error of the pedestrian estimated by the pedestrian position estimating step, and the moving direction correcting step is a step of determining a pedestrian movement type determining step And a measurement value updating step of updating the measurement value according to the determined movement type.

)보다 큰지 여부에 따라 판단하는 것을 특징으로 한다.In the present invention, the walking direction of the pedestrian in the exercise type determining step is determined by the difference between the current walking position and the previous walking position of the pedestrian according to Equation (6), and whether the walker straightly walks forward is expressed by Equation And the difference between the present walking direction and the average value of the previous walking direction is found by the following equation

) Is greater than a predetermined threshold value.

&Quot; (6) "

&Quot; (7) "

Further, the present invention is characterized in that the inertial navigation system error is estimated by INS-EKF-ZUPT when it is determined that the pedestrian is walking in a non-straight line by the exercise type determination step.

In addition, when the pedestrian is determined to walk in a straight line by the exercise type determining step, Equation (11) is used as a measurement value of the error model, and Equation (12) is used as a measurement matrix of the filter.

&Quot; (11) "

&Quot; (12) "

Further, according to the present invention, when the pedestrian is determined to walk in a straight line along the main direction by the exercise type determining step, equation (13) is used as a measurement value of the error model, and .

&Quot; (13) "

&Quot; (14) "

In the indoor environment where GPS is difficult to use, the present invention can accurately estimate the position of the pedestrian with only the inertial navigation sensor mounted on the shoe without using any other system, so that it is unnecessary to construct an infrastructure such as WiFi.

Further, since the present invention corrects the walking direction according to the three types of walking characteristics of the pedestrian in the room, i.e., nonlinear motion, linear motion, and motion in the main direction, ultimately, the position of the pedestrian can be more accurately estimated have.

Fig. 1 is a diagram showing a pedestrian's walking motion in a divided manner,
FIG. 2 is a view showing the body coordinate system of the ARS mounted on the shoe,
FIG. 3 is a flowchart sequentially illustrating a pedestrian movement direction estimation method in a room using a walking characteristic according to the present invention;
FIG. 4 is a flowchart illustrating a process of classifying a pedestrian movement type according to the present invention;
FIG. 5 is a block diagram illustrating a method for estimating a moving direction of a pedestrian in a room using a walking characteristic according to the present invention.
6 is a graph showing a result of a comparative experiment in a case where a pedestrian walks along the main direction in the room,
FIG. 7 is a graph showing a result of a comparative experiment in a case where the user walks in various movement types in an open space.

The present invention provides a pedestrian movement direction estimation method capable of accurately estimating a moving direction of a pedestrian in a propagation shaded area which is difficult to receive a navigation signal such as a room, And then estimates the position of the pedestrian through the INS-EKF-ZUPT structure based on the pedestrian data. At this time, the movement direction error of the pedestrian is reduced The present invention will be described in more detail with reference to the accompanying drawings which illustrate preferred embodiments of the present invention.

The moving direction estimating method of the present invention comprises a walking data acquiring step (S100), a pedestrian position estimating step (S200) and a moving direction correcting step (S300) as shown in FIG. 3, and the moving direction correcting step (S300) The pedestrian movement type determination step S310 and the measurement value update step S320.

(1) Gait data acquisition step (S100)

In this step, an inertial measurement device (IMU) is mounted on the footwear of a pedestrian as shown in FIG. 2, and walking data such as an acceleration signal transmitted in accordance with the walking operation of the pedestrian from the inertial measurement device is received through the wireless receiver The received signal is divided into a stance phase section and a swing phase section according to a walking operation of a pedestrian. Herein, the step is a section in which the sole is in contact with the ground, The foot section refers to a section where the sole moves away from the ground and moves in space.

(2) Pedestrian position estimation step S200

This step estimates the position of the pedestrian using the walking data after obtaining the walking data of the pedestrian by the step of obtaining the walking data (S100) as described above. In the present invention, in order to estimate the position of the pedestrian, (Inertial Navigation System) -EKF (Extended Kalman Filter) -ZUPT (Zero velocity UPdaTe) proposed by E. Foxlin, and the INS-EKF-ZUPT (Inertial Navigation System Means an Inertial Navigation System (INS) in which the velocity of the shoe attached to an inertial measuring unit (IMU) is 0 while the shoe is attached to an extended Kalman filter (EKF).

)와 함께 동작하는데, 이러한 상태벡터는 아래의 수학식 1에서와 같이 가속도 센서와 자이로스코프의 바이어스(

)와, 자세, 위치 및 속도에 대한 에러(

)의 5개의 에러 성분을 포함하고, 5개의 에러 성분은 각각 3개의 축 방향을 가짐으로써 전체적으로 15개의 에러 상태벡터를 구성하게 된다.
The extended Kalman filter (EKF) used in the INS-EKF-ZUPT consists of 15 error state vectors

), Which corresponds to the bias of the acceleration sensor and the gyroscope (as shown in Equation 1 below)

), Errors in position, position, and velocity (

), And five error components each have three axial directions, thereby constituting 15 error state vectors as a whole.

는 k번 째 시간에서의 상태변수,

는 자세 오차,

는 자이로 바이어스,

는 위치 오차,

는 속도 오차,

는 가속도 바이어스를 뜻한다.
here,

Is the state variable at time k,

The posture error,

Gyro bias,

The position error,

Speed error,

Is the acceleration bias.

At this time, the inertial measurement unit (IMU) is generally a micro electro mechanical systems (MEMS) class microsensor. The bias for such a microsensor can be simplified as shown in the following Equation 2, and each state in Equation Can be estimated by time propagation of Equation (3) and updating of the measurement value of Equation (4). At this time, the zero velocity correction (ZUPT) as in Equation (5) Assuming that the velocity is zero, each error can be estimated, and therefore, by removing the estimation error in each state, it is possible to estimate a more accurate attitude and position, and a series of processes and procedures are well known A more detailed description thereof will be omitted.

는 필터의 시스템 모델,

는 측정치간의 시간,

은 바디(body) 축에서 항법 축으로의 자세변환행렬, I는 [3*3]의 단위행렬,

는 각각 x, y, z축의 가속도를 나타낸다.
here,

The system model of the filter,

Is the time between measurements,

Is the posture conversion matrix from the body axis to the navigation axis, I is the unit matrix of [3 * 3]

Represents the acceleration of the x, y, and z axes, respectively.

은 공분산 행렬이고, W는 시스템 모델 오차(process noise), Q는 측정치 오차(measurement noise)를 나타낸다.
here

Is a covariance matrix, W is a system model error, and Q is a measurement noise.

는 측정치 함수를 뜻하고, 그 구성은 수학식 5와 같다. 그리고

은 측정치,

은 노이즈이다.

는 칼만이득(Kalman gain),

은 시스템 모델 오차이다.
here

Denotes a measurement value function, and its configuration is shown in Equation (5). And

Is a measurement,

Is the noise.

Is the Kalman gain,

Is the system model error.

(3) Moving direction correction step (S300)

As described above, using the INS-EKF-ZUPT can reduce the error in walking navigation, but it is not easy to accurately estimate the yaw value and bias of the gyroscope related to the vertical axis. Therefore, In the present invention, an AHDE (Advanced Heuristic Drift Elimination) is applied in order to reduce the position error of the pedestrian by the estimation step S200. At this time, A pedestrian movement type determination step S310 for determining the type of the pedestrian movement, and a measurement value update step S320 for determining the movement type and then updating the measured value according to the movement type.

In the AHDE of the present invention, since the building is generally constructed in a rectangular shape, it is assumed that there are four kitchen sculptures perpendicular to each other in the interior as in the conventional HDE algorithm.

1) Determination of pedestrian movement type (S310)

This step determines the type of movement of the pedestrian. It is based on the previous pedestrian position data and the current position data, and the pedestrian is selected from one of three types of motion, namely, nonlinear motion, linear motion, (Judgment).

걸음과 현재의 한 걸음이 사용되며, 본 발명에서는 결정의 정확성을 높이는 동시에 시간을 단축할 수 있도록 이전의 5걸음과 현재의 1걸음에 기초하여 직선 운동인지 여부를 결정한다.In order to do this, the AHDE must first determine whether the pedestrian walks straight ahead or not. In order to determine whether the pedestrian's walk is a linear motion,

A step and a current step are used. In the present invention, it is determined whether the linear motion is based on the previous five steps and the present one step so as to increase the accuracy of the determination and shorten the time.

)와 현재의 위치 (

)사이의 차이를 이용하여 걸음 방향을 구한 다음, 보행자가 전방을 향해 직선으로 걷는지 여부는 수학식 7에서와 같이

걸음의 방향의 분산과 현재의 걸음 방향과 이전의

걸음

방향의 평균값과의 차이를 구하여 이 값이 임계값(

)보다 큰지 여부에 따라 판단한다.
More specifically, in order to determine whether or not the pedestrian is walking in a straight line, the following equation (6)

) And the current location (

). Then, whether or not the pedestrian walks straight ahead in the forward direction can be calculated as shown in Equation (7)

The variance of the direction of the step and the direction of the current step

step

And the difference between the average value and the average value of the direction

Is greater than a predetermined value.

는 각 걸음(

)에서의 이동방향 값이며,

는

번째 걸음에서의 위치를 뜻한다.
here,

Each step (

), ≪ / RTI >

The

Means the position at the second step.

는 걸음수,

,

는 각각 한 걸음 이전,

걸음 이전의 걸음 수이다.
here,

The number of steps,

,

Respectively,

It is the number of steps before the step.

)과 보행자의 평균 걸음 방향(

)의 차이(

)를 구한 다음, 수학식 9에서와 같이 이 차이가 임계값(

)보다 작은 경우에는 보행자가 주방향을 따라 걷는 것으로 판단하며, 이때 임계값(

)은 수학식 10에서와 같이 필터의 yaw 에러의 분산에 따라 바뀌며, 이상과 같은 보행자의 운동 유형을 구분하는 일련의 과정을 흐름도로 나타내면 도 4와 같이 나타낼 수 있다.
Next, if it is determined that the pedestrian is walking in the straight direction as a result of performing the above-described process, it is determined whether the pedestrian is walking along the main direction. To this end, in the present invention,

) And the average walking direction of the pedestrian

) Difference

), And then, as in Equation (9), this difference becomes a threshold value

), It is determined that the pedestrian is walking along the main direction. At this time, the threshold value

) Is changed according to the yaw error distribution of the filter as shown in Equation (10), and a series of processes for classifying the motion type of the pedestrian as described above can be represented as a flow chart as shown in FIG.

는 가장 가까운 주방향 성분의 이동방향 값을 나타낸다.
here,

Represents the moving direction value of the nearest principal direction component.

는 yaw의 오차 공분산 값,

는 임계값 상수이다.
here

Is the error covariance value of yaw,

Is a threshold constant.

(2) the measurement value updating step (S320)

This step is a step of updating the measured value according to the movement type of the pedestrian after the pedestrian movement type is determined by which of the three types according to the pedestrian movement type classification step S310 as described above, The procedure for updating the measurement value in the following is as follows.

1) In case of nonlinear motion

Since the AHDE algorithm does not affect the filter if the pedestrian does not walk in a straight line, the INS error is estimated by INS-EKF-ZUPT because it is not easy to accurately estimate the yaw value and bias associated with the vertical axis of the gyroscope , In which case the error variance of the yaw axis increases.

2) Linear motion

)은 0이라고 가정할 수 있다.

걸음 동안의 걸음 방향의 변화율을 현재 추정되고 있는 방향의 변화율이라고 할 수 있고, 이 값과 yaw율(

)의 차이가 오차모델의 측정치로 사용되는데 yaw율(

)이 0이기 때문에 아래의 수학식 11과 같이 전개되며, 이때 필터의 측정치 행렬은 수학식 12와 같다.
If the pedestrian does not walk along the main direction but makes a walk in the forward direction, the correction is made to the bias of the gyroscope. If the pedestrian walks in a straight line, the yaw rate

) Can be assumed to be zero.

The change rate of the walking direction during the step can be regarded as the rate of change in the currently estimated direction, and this value and the yaw rate

) Is used as a measure of the error model. The yaw rate

) Is 0, the following equation (11) is developed, and the measurement matrix of the filter is expressed by Equation (12).

3) linear motion in the main direction

)과 가장 가까운 주방향(

)과의 차이(

)가 수학식 13에서와 같이 오차 모델의 측정치로 사용되고, 이때 측정치 행렬은 수학식 14와 같다.
When a pedestrian is walking in a straight line along the main direction, the average walking direction (

) And the closest main direction (

) And the difference between

) Is used as a measurement value of the error model as shown in Equation (13), and the measurement matrix is expressed by Equation (14).

As described above, according to the present invention, it is first determined whether the walker is walking in a straight line or not and whether the walker follows the main direction. Then, since the correction is performed for each determined type of motion, the moving direction estimation performance is improved, The position can be estimated more accurately.

In order to confirm the usefulness of the AHDE algorithm included in the pedestrian movement direction estimation method of the present invention as described above, the present inventors conducted two types of comparison experiments using an inertial measurement device mounted on a shoe. First, In the case where a pedestrian walks along a corridor (main direction) of a building in a building having a 360-m menopause path, as shown in FIG. 6, in the IEZ algorithm, the moving direction is diverged by the gyroscope bias The result shows that the position is divergent, whereas the experiment using HDE and AHDE algorithms shows the position return within 1m.

Next, in the experiment in which the pedestrian walked in various exercise types in an open space having a menopause of 340 m, the IEZ algorithm diverged as shown in FIG. 7, and the HDE algorithm showed a large return error, In the experiment using AHDE algorithm, relatively good position return performance was shown.

As described above, according to the present invention, since the motion type of the pedestrian is determined and then the appropriate correction is performed according to the determined motion type, the estimation performance of the moving direction of the pedestrian in the room is improved, It becomes possible to know more accurately.

Claims (5)

A method for estimating a moving direction of a pedestrian in a room,
A walking data acquiring step (S100) of acquiring walking data from a signal transmitted in accordance with a walking operation of a pedestrian from an inertial measuring device (IMU) mounted on a footwear of a pedestrian;
A step S200 of estimating the position of the pedestrian using the obtained walking data;
And a moving direction correction step (S300) of correcting the moving direction of the pedestrian so as to reduce a position estimation error of the pedestrian estimated by the pedestrian position estimation step (S200)
The movement direction correction step S300 includes a pedestrian movement type determination step S310 for determining the movement type of the pedestrian again and a measurement value update step S320 for updating the measurement value according to the determined movement type A method for estimating the moving direction of a pedestrian in a room using walking characteristics.
The method according to claim 1,
The walking direction of the pedestrian in the exercise type determining step S310 is determined by the difference between the current walking position and the previous walking position of the pedestrian according to Equation (6), and whether or not the pedestrian walks straight ahead toward the front is expressed by Equation And the difference between the present walking direction and the average value of the previous walking direction is found by the following equation

And determining whether the pedestrian movement direction is greater than a predetermined distance or not.
&Quot; (6) "

&Quot; (7) "

The method according to claim 1 or 2,
In the case where it is determined that the pedestrian walks non-linearly by the exercise type determination step (S310), the inertial navigation system error is estimated by the INS-EKF-ZUPT, and the pedestrian movement direction estimation Way.
The method according to claim 1 or 2,
(11) is used as a measurement value of the error model when the pedestrian is determined to walk in a straight line by the motion type determining step (S310), and the equation (12) is used as the measurement matrix of the filter. A Method for Estimating Pedestrian Moving Direction in Indoor.
&Quot; (11) "

&Quot; (12) "

The method according to claim 1 or 2,
(13) is used as a measurement value of the error model when the pedestrian is determined to walk in a straight line along the main direction by the exercise type determination step (S310), and the measurement matrix of the filter is expressed by the following expression A Method for Estimating Pedestrian Moving Direction in Indoor Using Walking Characteristics.
&Quot; (13) "

&Quot; (14) "

Images