RU2685767C1 - Method and device for strap down inertial navigation - Google Patents

Abstract

FIELD: navigation.SUBSTANCE: present invention relates to strap down inertial navigation. Disclosed are method and device for strap down inertial navigation using measurements made by one or more inertial sensors. Method of strap down inertial navigation includes stages, at which: values of object linear acceleration, object angular velocity values, values of temperature of inertial sensor and value of magnetic field intensity, measured by one inertial sensor, then corrected received values of angular speed of object on correction factor for current temperature value and correction of accepted values of angular velocity of object to correction factor is performed. Determining current spatial orientation of object based on object angular speed; determining whether the object is moving, based on linear acceleration of object and values of angular speed of object using previously obtained first intellectual model of movement of object and when determining presence of movement of object, determining current spatial position of object based on values of linear acceleration of object and values of angular speed of object using previously obtained second intellectual model of movement of object, and when determining absence of object movement, setting current spatial position of object equal to previous spatial position of object.EFFECT: determining spatial position and orientation of object based on noisy and distorted measurements, made with one inertial sensor, without the need for coordination with external landmarks and / or use of data coming from outside in conditions of arbitrary spatial movement of object.42 cl, 8 dwg

Description

BACKGROUND OF INVENTION

The technical field to which the invention relates.

The present invention relates generally to the field of strapdown inertial navigation, and in particular, to a method and device for strapdown inertial navigation to determine the spatial orientation and spatial position of an object using data obtained by one or more inertial sensors.

Description of the prior art

The definition of spatial orientation and spatial position using inertial data is described, for example, in US patent US 9052202 B2, published 12/15/2011 entitled "USE OF INERTIAL SENSOR DATA TO IMPROVE MOBILE STATION POSITIONING", in which the calculated value is selected from a set of parallel operating models filter position of navigation, and this method performs both the determination of the absence of movement of the object, and the numerical evaluation of the measure of movement. The disadvantage of this approach is that, first, the values of the linear acceleration of the object and the angular velocity of the object from inertial sensors do not aggregate, i.e. This method allows to obtain an estimate of the spatial orientation of the object and the spatial position of the object only with a certain probability, and secondly, this method is fully focused on working with the global navigation satellite system.

Another method for determining spatial orientation and spatial position using inertial data is the approach described in US patent application US 20130158941 A1, published June 20, 2013, entitled "MOVING DIRECTION DETERMINATION WITH NOISY SIGNALS FROM INERTIAL NAVIGATION SYSTEMS ON MOBILE DEVICES" uses a vibration energy model to determine if there is motion, and also to estimate its direction. The disadvantages of this method are the low quality of motion estimation and inapplicability for cases of complex motion, since as a result of the work, the measurement of acceleration over time is not performed.

Another way to determine the spatial orientation of the object and the spatial position of the object using inertial data is the method described in the article SM LaValle, et al. and values from inertial sensors processed by the Kalman filter. The disadvantages of this method are, firstly, the practical inapplicability to assess the spatial position of the object, and secondly, the sensitivity of the calculated values of the yaw angle to changes in the magnetic field.

The following example of determining the spatial orientation of an object and the spatial position of an object using inertial data is the approach described in the article by X. Sun, et al. -4, 239-242 (2014). The essence of this approach is the use of an improved Zero-Velocity Update (ZUPT) correction method using the advanced Kalman filter in the moments of no movement. Serious disadvantages of this method are the need for a priori knowledge about the object stopping, as well as the impossibility of assessing the spatial orientation of the object.

Another approach that uses inertial data is the approach described in article D. Anguita et al., A Hardware-friendly Support Vector Machine for Embedded Automotive Applications, IJCNN, 1360-1364 (2007). This method is based on the classification of types of user movements using a set of inertial sensors mounted on a belt using the support vector machine method. This method does not estimate the spatial position of the object and the spatial orientation of the object, and is aimed only at the classification of the types of movement of the user, such as running, walking, climbing stairs, etc.

Thus, there is a need for methods and devices that more accurately assess both the spatial orientation of an object and the spatial position of an object using measurements of an inertial sensor for complex spatial movement of an object, arbitrary in character (speed, acceleration) and direction, without external reference points and / or use of incoming data from outside.

The present invention is intended to determine the spatial position of an object and the spatial orientation of an object based on noisy and distorted measurements made by one inertial sensor, without the need for coordination with external reference points and / or using incoming data from outside in an arbitrary spatial displacement of the object.

SUMMARY OF INVENTION

The aim of the present invention is the provision of a method and device for strapdown inertial navigation, which allow you to get at least one of the following advantages:

- ensuring sustainable determination of the spatial orientation of the object and the spatial position of the object for a long time based on noisy and distorted measurements made by one or more inertial sensors, without coordination with external reference points, without using additional data from devices that read information about the surrounding space, such as camera, global positioning system receiver, etc .;

- determination of the spatial orientation of the object and the spatial position of the object in terms of an arbitrary spatial displacement of the object according to its character (speed, acceleration) and direction of movement, such as individual steps with fixation of the foot, long smooth movement, etc .; irrespective of the spatial orientation of the object, that is, the spatial orientation of the object may not coincide with the direction of movement of the object;

- no need for factory calibration of the device, the ability to compensate for the deviation of the zero angular velocity of the gyroscope of the inertial sensor and compensate for the influence of the inhomogeneity of the magnetic field on the value of the angular velocity of the object directly in use

- the possibility of combining with other methods and devices of inertial navigation, for example, based on visual odometry, or global positioning systems for the purpose of long-term continuous determination of the spatial orientation of an object and the spatial position of an object in difficult environmental conditions, such as poor and / or rapidly changing lighting temperature change, magnetic field induction, etc .;

- The possibility of embedding in a variety of different devices, such as helmets, virtual / augmented / combined reality, air / ground / underwater drones, self-guided vehicles (scooter, bicycle, car, etc.), mechanized prostheses, etc.

This application discloses a strapless inertial navigation method, comprising the steps of: taking the linear acceleration values of an object measured by an accelerometer of an inertial sensor, the angular velocity of an object measured by an inertial sensor gyroscope, measured by an inertial sensor thermometer, and magnetic field values measured by an inertial sensor magnetometer, with the object’s linear acceleration values, angular velocity values The object, the temperature values of the inertial sensor and the magnetic field intensity acting on the inertial sensor magnetometer, are measured at at least one inertial sensor; if a deviation of the zero angular velocity of the gyroscope of the inertial sensor is detected, the accepted values of the angular velocity of the object are corrected by the correction factor for the current temperature value, and a set of correction factors for each value of the operating temperature of the inertial sensor is obtained in advance; if a change in the magnetic field strength is detected, correction of the accepted values of the angular velocity of the object by the correction factor is performed, and the value of the correction factor is calculated as a solution to the optimization problem; determine the current spatial orientation of the object based on the accepted values of the angular velocity, if no deviation of the zero angular velocity and / or a change in the magnetic field strength, or corrected values of the angular velocity of the object is detected, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected; determine whether the object is moving based on the accepted values of the linear acceleration of the object and the adopted values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the adopted values of the linear acceleration of the object and the corrected values of the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the previously obtained first intelligent model of change the object, the first intellectual model object changes the motion obtained in advance based on the characteristics of movement of at least one object of the same type as that of said object; and when determining whether the object has motion, determine the current spatial position of the object based on the accepted values of the linear acceleration of the object and the adopted values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the accepted values of the linear acceleration of the object and corrected values of the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the received second intellectual model of object motion, the second intellectual model of object motion obtained in advance based on the characteristic motion characteristics of at least one object of the same type as the said object, and when determining the absence of object motion set the current spatial position of the object equal to the previous spatial position object.

In an additional aspect, the method further comprises filtering the received values of the linear acceleration of the object and the received values of the angular velocity of the object to suppress noise, while filtering the received values of the linear acceleration of the object and the received values of the angular velocity of the object by means of weighted averaging using the sliding window method.

In another additional aspect, obtaining a correction factor for each value of the operating temperature of the inertial sensor comprises the steps of: determining whether the object is moving based on the linear acceleration values of the object and the angular velocity values of the object using the previously obtained first intelligent model of the change in the movement of the object; if it is determined that the object is not moving, form a set of accepted values of the angular velocity of the object for the current value of the operating temperature of the inertial sensor, and the optimal number of received values of the angular velocity of the object in the set is determined based on optimizing the ratio of the displacement of the average angular velocity of the object and the magnitude of the spread of the angular object speed; average the received values of the angular velocity of the object of the above set by the method of weighted average and calculate the correction factor for the current value of the inertial sensor operating temperature by solving the regression equation, while forming the set of the accepted values of the angular velocity of the object and calculating the correction factor for each value of the inertial sensor operating temperature.

In yet another additional aspect, the optimal number of accepted values of the angular velocity of an object in said set is determined by the method of intersection of the confidence intervals.

In yet another additional aspect, the calculation of the correction factors for the received value of the magnetic field strength for correcting the adopted values of the object’s angular velocity by varying the magnetic field strength includes the steps of: taking the magnetic field strength values, the magnetic field distribution model, and the inertial sensor magnetometer , with the model of the distribution of magnetic field strength and calibration parameters calculated zara her; the accepted values of the magnetic field strength are adjusted using the calibration parameters of the magnetometer; If the corrected values of the magnetic field strength satisfy the range of the accepted model of the distribution of the magnetic field strength, the correction coefficients of the accepted values of the angular velocity of the object are calculated as a measure of the deviation of the current vector of intensity from the model, if the corrected values of the magnetic field strength magnetic field, the calibration parameters are recalculated as orthogonalization and minimization of the mean value of the tension and cross-correlation between the axes.

In yet another additional aspect, the current spatial orientation of the object is calculated by numerical integration.

In yet another additional aspect, determining whether an object is moving is performed by comparing the value of the Euclidean norm of the angular velocity of an object with one constant and comparing the value of the Euclidean norm of linear acceleration of the object with another constant, and the above comparisons with constants are specified in the first intellectual model of the object’s movement, this determines that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object do not exceed the mentioned a constant, and it is determined that the object moves if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object exceed the mentioned constant.

In yet another additional aspect, determining whether an object moves is performed by comparing the value of the Euclidean norm of linear acceleration of an object with one adaptive threshold by the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold by the sliding window method and the method of calculating adaptive thresholds using the sliding window method is specified in the first intellectual model of the change in the motion of an object.

In yet another additional aspect, the calculation of the adaptive threshold and the comparison with it of the Euclidean norm of the linear acceleration of an object are performed iteratively and contain steps in which: an adaptive threshold is established, it is preliminarily considered that the object does not move and the adaptive threshold is calculated in advance; taking a predetermined threshold for determining movement and a predetermined threshold for determining the absence of movement; comparing the value of the norm of the linear acceleration of an object with an adaptive threshold, and if the value of the norm of the linear acceleration of the object does not exceed the adaptive threshold, then it is determined that the object does not move; if the value of the norm of linear acceleration of an object exceeds the adaptive threshold, then determine the duration, how long the norm of the norm of linear acceleration of the object exceeds the adaptive threshold; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold exceeds the threshold for determining movement, then it is determined that the object moves, while the adaptive threshold does not change; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move; if it is determined that the object is not moving, recalculate the current adaptive threshold by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigning the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value covariance matrix; if it is determined that the object is moving, compare the value of the norm of the linear acceleration of the object with the adaptive threshold, and if the value of the norm of the linear acceleration of the object exceeds the adaptive threshold, then it is determined that the object continues to move; if the value of the norm of linear acceleration of an object does not exceed the adaptive threshold, then the completion of the movement is determined by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion, then determine that the object does not move, while the adaptive threshold is recalculated in accordance with the norm of the input values of the linear acceleration of the object.

In yet another additional aspect, the adaptive threshold is calculated as the value of the norm of linear acceleration of objects of the same type, averaged over a variety of different test data, at which movement is uniquely detected.

In yet another additional aspect, determining whether an object is moving is carried out using at least one previously trained recurrent neural network as the first intelligent model for changing the movement of an object.

In yet another additional aspect, the determination of whether an object is moving is performed using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of an object.

In yet another additional aspect, each recurrent neural network is a long short-term memory (LSTM).

In yet another additional aspect, determining the current spatial position of an object when determining the presence of an object’s movement includes the steps of: determining the direction vector of the object’s motion by calculating the average value of the linear acceleration of the object in the global coordinate system using the sliding window method and associating this average value to the coordinate system in which the vertical axis is parallel to the gravitational acceleration vector; estimate the speed of the object based on the values of the linear acceleration of the object using the second intellectual model of the movement of the object; projecting the estimated velocity of the object onto a certain vector of the direction of motion of the object and obtaining the current spatial position of the object by a method of numerical integration.

In yet another additional aspect, the vector of the direction of movement of the object is determined by the support vector machine (SVM).

In yet another additional aspect, an object direction vector is determined using a pre-trained recurrent neural network.

In yet another additional aspect, the linear velocity of the object is estimated based on the linear acceleration values of the object using the polynomial velocity model as the second intelligent model of the object's motion.

In yet another additional aspect, the linear velocity of the object is estimated based on the values of the linear acceleration of the object using a pre-trained recurrent neural network as the second intelligent model of the movement of the object.

In yet another additional aspect, an object's linear velocity is estimated based on the object's linear acceleration values using a combination of pre-trained models as a second intelligent model of the object's motion.

In yet another additional aspect, if the values of the linear acceleration of the object, the angular velocity of the object, the temperature values and the magnetic field strength values measured by more than one inertial sensor are taken, then the current spatial orientation of the object is independently determined on the basis of the adopted values measured by each sensor independently determine the current spatial position of the object on the basis of the adopted values measured by each sensor, with the aforementioned method additionally containing Stages include: compute the resulting current spatial orientation of the object based on all the current spatial orientations of the object, determined independently based on the received values from different inertial sensors; and calculate the resulting current spatial position of the object based on all current spatial positions of the object, determined independently based on the received values from different inertial sensors.

In yet another additional aspect, the calculation of the resulting current spatial orientation of the object and the calculation of the resulting current spatial position of the object are performed by weighted averaging of the current spatial orientations of the object and the current spatial positions of the object, determined independently based on the adopted values measured by different inertial sensors.

In addition, the present application discloses a strapdown inertial navigation device, comprising: a temperature correction unit adapted to, if a deviation of the zero angular velocity of the gyroscope of the inertial sensor is detected, to correct the angular velocity values of the object by the correction factor for the current temperature value, and a set of correction coefficients for each value of the operating temperature of the inertial sensor obtained in advance; a magnetic field correction unit, configured to, if a change in the magnetic field strength is detected, perform correction of the angular velocity values of the object by the correction factor, and the value of the correction factor is calculated as a solution to the optimization problem; a spatial orientation determination unit configured to determine the current spatial orientation of an object based on the angular velocity values, if no deviation of the zero angular velocity and / or a change in the magnetic field strength, or corrected angular velocity values of the object is detected, if a deviation of the zero angular velocity and / or change in magnetic field strength; a motion detection unit configured to determine whether the object is moving based on the linear acceleration values of the object and the angular velocity values of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the linear acceleration values of the object and corrected values the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the previously obtained first intel ektualnoy change model of the object, wherein the first intelligent model changes of the object is obtained in advance based on the characteristics of movement of at least one object of the same type as that of said object; and a spatial position determination unit, configured to determine the current spatial position of the object based on the linear acceleration values of the object and the values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field is detected, or the basis of the values of the linear acceleration of the object and the corrected values of the angular velocity of the object, if a deviation of zero angular velocity and / or a change in voltage is detected magnetic field using the previously obtained second intellectual model of the object's movement, the second intellectual model of the object's movement obtained in advance based on the characteristic features of the movement of at least one object of the same type as the said object, and, when determining the absence of the object's movement, the current spatial position of the object is equal to the previous spatial position of the object, while the values of the linear acceleration of the object are measured with an accelerometer inertial the sensor, the angular velocity of the object, measured by the gyro of the inertial sensor, the temperature of the inertial sensor, measured by the thermometer of the inertial sensor and the magnetic field strengths the intensity of the magnetic field acting on the magnetometer of the inertial sensor, measured by at least one inertial tchikom.

In an additional aspect, the device further comprises a filter adapted to filter the linear acceleration values of the object and the angular velocity values of the object to suppress noise, while filtering the linear acceleration values of the object and angular velocity values of the object is performed by weighted averaging using the sliding window method.

In another additional aspect, the temperature correction unit obtains a correction factor for each value of the inertial sensor operating temperature by: if it is determined that the object is not moving, generating a set of object angular velocity values for the current value of the inertial sensor operating temperature, said set is determined based on the optimization of the ratio of the displacement of the average value of the angular velocity of the object and the magnitude of the wasp values of the angular speed of the object; averaging the angular velocity values of the object of the above set using the weighted average method and calculating the correction factor for the current value of the inertial sensor operating temperature by solving the regression equation;

In yet another additional aspect, the optimal number of angular velocity values of an object in said set is determined by the method of intersection of confidence intervals.

In yet another additional aspect, the magnetic field correction unit calculates correction factors for the magnetic field strength value for correcting the object’s angular velocity values by varying the magnetic field strength by: adjusting the magnetic field strength values using the calibration parameters of the magnetometer; if the corrected magnetic field strengths satisfy the range from the magnetic field strength distribution model, calculate the correction factors of the object’s angular velocity as a measure of the deviation of the current intensity vector from the model, if the corrected magnetic field strengths do not match the range from the magnetic field distribution model gauge parameters as orthogonalization problems and minimizing the displacement of the mean o values of the intensity and cross-correlation between the axes, and the model of the distribution of the magnetic field strength and the calibration parameters are calculated in advance.

In yet another additional aspect, the spatial orientation determination unit determines the current spatial orientation of an object by computing with a numerical integration method.

In yet another additional aspect, the motion detection unit determines whether the object moves by comparing the value of the Euclidean norm of the angular velocity of the object with one constant and comparing the value of the Euclidean norm of linear acceleration of the object with another constant, and these comparisons with constants are specified in the first intellectual model of the object's movement , it is determined that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object is not greater than the operation mentioned constant, and it is determined that the object is moving, if the value of the Euclidean norm of angular velocity of the object and the value of the Euclidean norm of a linear acceleration of the object exceeds the said constant.

In yet another additional aspect, the motion detection unit determines whether an object moves by comparing the value of the Euclidean norm of the linear acceleration of an object with one adaptive threshold by the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold by the sliding window method adaptive thresholds and a method for calculating adaptive thresholds using the sliding window method are specified in the first intellectual model of the change in the motion of an object.

In yet another additional aspect, the calculation of the adaptive threshold and the comparison with it of the Euclidean norm of linear acceleration of an object is performed iteratively by: setting the adaptive threshold, it is pre-considered that the object does not move, and the adaptive threshold is calculated in advance; receiving a predetermined threshold for determining motion and a predetermined threshold for determining the absence of motion; comparing the value of the norm of linear acceleration of an object with an adaptive threshold, and if the value of the norm of linear acceleration of an object does not exceed the adaptive threshold, then it is determined that the object does not move; if the value of the norm of linear acceleration of the object exceeds the adaptive threshold, then the duration is determined, how long the norm of the linear acceleration of the object exceeds the adaptive threshold; this adaptive threshold does not change; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move; if it is determined that the object is not moving, the current adaptive threshold is recalculated by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigning the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value covariance matrix; if it is determined that the object is moving, the value of the norm of the linear acceleration of the object is compared with the adaptive threshold, and if the value of the norm of the linear acceleration of the object exceeds the adaptive threshold, then it is determined that the object continues to move; if the value of the norm of the linear acceleration of an object does not exceed the adaptive threshold, then the completion of the movement is determined by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion, that the object does not move, while the adaptive threshold is recalculated in accordance with the norm of the input values of the linear acceleration of the object.

In yet another additional aspect, the adaptive threshold is calculated as the value of the norm of linear acceleration of objects of the same type, averaged over many different test data, at which movement is uniquely detected.

In yet another additional aspect, the motion detection unit determines whether the object is moving, using at least one pre-trained recurrent neural network as the first intelligent model for changing the movement of the object.

In yet another additional aspect, the motion detection unit determines whether the object is moving, using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of the object.

In yet another additional aspect, each recurrent neural network is a long short-term memory (LSTM).

In yet another additional aspect, the spatial position determination unit determines the current spatial position of an object when determining whether an object moves by: determining the direction vector of an object by calculating the average value of the linear acceleration of an object in the global coordinate system using the sliding window method and linking this average value to the coordinate system in which the vertical axis is parallel to the gravitational acceleration vector; estimates of the velocity of the object based on the values of the linear acceleration of the object using the second intellectual model of the movement of the object; projecting the estimated velocity of the object onto a certain vector of the direction of motion of the object and obtaining the current spatial position of the object using a numerical integration method

In yet another additional aspect, the determination of the vector of the direction of motion of an object is performed by the support vector (SVM) method.

In yet another additional aspect, the determination of the vector of the direction of movement of the object is performed using a previously trained recurrent neural network.

In yet another additional aspect, the estimation of the linear velocity of the object based on the values of the linear acceleration of the object is performed using the polynomial velocity model as the second intelligent model of the motion of the object.

In yet another additional aspect, the estimation of the linear velocity of the object based on the values of the linear acceleration of the object is performed using a pre-trained recurrent neural network as the second intelligent model of the motion of the object.

In yet another additional aspect, the estimation of the linear velocity of the object based on the linear acceleration values of the object is performed using a combination of pre-trained models as the second intelligent model of the movement of the object.

In yet another additional aspect, if the object linear acceleration values, the object angular velocity values, the temperature values and the magnetic field strength values are measured by more than one inertial sensor, then the spatial orientation determining unit independently determines the current spatial orientation of the object based on the values measured by each sensor, and calculates the resulting current spatial orientation of the object based on all the current spatial orientations of the object, defined independently based on the received values from different inertial sensors, and the spatial position determination unit independently determines the current spatial position of the object based on the values measured by each sensor, and calculates the resulting current spatial position of the object based on all the current spatial positions of the object determined independently based on values from different inertial sensors.

In yet another additional aspect, the spatial orientation determination unit calculates the resulting current spatial orientation of an object by weighted averaging the current spatial orientations of the object, determined independently based on the values measured by different inertial sensors; and the spatial position determination unit calculates the resulting current spatial position of the object by means of weighted averaging of the current spatial positions of the object determined independently based on the received values measured by different inertial sensors.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other aspects, features and advantages of the present invention will be better understood from the subsequent detailed description given in conjunction with the accompanying drawings, in which:

FIG. 1 is a flowchart illustrating one embodiment of a free inertial navigation method.

FIG. 2 is a flowchart illustrating another embodiment of a free inertial navigation method.

FIG. 3 is a block diagram illustrating one embodiment of a strapdown inertial navigation device.

FIG. 4 is a block diagram illustrating another embodiment of a strapdown inertial navigation device.

FIG. 5 is a graph illustrating the choice of the optimal number of angular velocity values of an object when obtaining a correction factor for each value of the inertial sensor operating temperature.

FIG. 6 is a block diagram illustrating the use of the invention as a device for recognizing and authenticating a user.

FIG. 7 is a block diagram illustrating the use of the invention as a system for automatically controlling the movement of a vehicle.

FIG. 8 is a block diagram illustrating the use of the invention as a system for automatically controlling the movement of a vehicle in 3-dimensional space.

In the following description, unless otherwise indicated, like reference numerals are used to designate the same elements of the device or method steps when shown in different drawings, and their parallel description is not given.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE PRESENT INVENTION

The following description with reference to the accompanying drawings is provided to facilitate a complete understanding of the various embodiments of the present invention defined by the claims and its equivalents. The description includes various specific details to facilitate such an understanding, but these details should be considered only as exemplary. Accordingly, those skilled in the art will find that various changes and modifications of the various embodiments described in this application can be developed without departing from the scope of the present invention. In addition, descriptions of well-known functions and constructions can be omitted for clarity and brevity.

The terms and formulations used in the following description and claims are not limited to bibliographic meanings, but simply used by the creator of the present invention to provide a clear and consistent understanding of the present invention. Accordingly, it will be clear to those skilled in the art that the following description of various embodiments of the present invention is offered for illustration only.

It should be understood that the singular forms include plurality, unless the context clearly indicates otherwise.

Additionally, it should be understood that the terms "comprises", "comprising", "includes" and / or "including", when used in this application, means the presence of the stated features, values, operations, elements and / or components do not exclude the presence or addition of one or more other features, values, operations, elements, components, and / or their groups.

The method and device of the strapdown inertial navigation can be used both to track the movement of an object in 2-dimensional space, and to track the movement of an object in 3-dimensional space. Objects for which the method and device of spherical inertial navigation can be used are people, various vehicles, such as, for example, vehicles driven by a driver, and vehicles with automatic control (scooter, bicycle, car, etc.) , air / ground / underwater drones, etc. The method and device of strapless inertial navigation can also be used for mechanized prostheses, virtual / augmented / combined reality helmets, user recognition and authentication using specific samples of human movement, automatic control of vehicle movement.

In the following, various embodiments of the present invention are described in more detail with reference to the accompanying drawings.

One embodiment of the method of inertial free navigation is shown in FIG. one.

At step 101 of the strapdown inertial navigation method, the object linear acceleration values measured by the inertial sensor accelerometer, the object angular velocity values measured by the inertial sensor gyroscope, inertial sensor temperatures measured by the inertial sensor thermometer, and magnetic field intensity values measured by the inertial sensor magnetometer are obtained The values of the linear acceleration of the object, the values of the angular velocity of the object, the temperature values of the inertial sensor and the values of the magnetic field strength acting on the magnetometer of the inertial sensor are measured by at least one inertial sensor.

At step 102 of the strapdown inertial navigation method, if a deviation of the zero angular velocity of the inertial sensor gyroscope is detected, the object angular velocity values adopted at step 101 are corrected by the correction factor for the current temperature, and the set of correction factors for each inertial sensor operating temperature is obtained in advance.

для рабочей температуры

добиваются путем компенсации измерений на поправочный коэффициент

в виде

, где

означает скорректированное значение угловой скорости объекта для рабочей температуры

. The deviation of the zero angular velocity of the gyroscope is defined as a non-zero value of the angular velocity of the object, provided that the absence of movement of the object is detected. It is known that each individual copy of the gyroscope can have its own value of deviation in the measurement of the angular velocity, and this is different for different values of the temperature of the gyroscope. Improving the accuracy and stability of measurements of the angular velocity of the object

for operating temperature

Achieve by compensation of measurements on the correction factor

as

where

means the corrected value of the angular velocity of the object for the operating temperature

.

объекта добиваются путем компенсации измерений значения магнитного поля на поправочный коэффициент

в виде

, где

означает скорректированное значение угловой скорости объекта.At step 103 of the strapdown inertial navigation method, if a change in the magnetic field strength is detected, correction is made of the object's angular velocity values adopted at step 101 or corrected at step 102 by a correction factor. The value of the correction factor is calculated as a solution to the optimization problem. Improved accuracy and stability of angular velocity measurements

an object is obtained by compensating for measurements of the magnetic field value by a correction factor

as

where

means the corrected value of the angular velocity of the object.

At step 104 of the inertial navigation inertial navigation method, the current spatial orientation of the object is determined based on the values of angular velocity taken at step 101, if no deviation of the zero angular velocity and / or a change in magnetic field strength, or values of the angular velocity of the object corrected if a deviation of the zero angular velocity of the gyroscope and / or a change in the magnetic field strength is detected. The current spatial orientation of the object is calculated by the method of numerical integration, which is widely known in the art.

At step 105 of the inertial navigation inertial navigation method, it is determined whether the object is moving, based on the object's linear acceleration values taken at step 101 and the object's angular velocity values taken at step 101, if no deviation of the gyroscope's zero angular velocity and / or a change in magnetic field strength is detected, or based on the values of the linear acceleration of the object taken at step 101 and the values of the angular velocity of the object corrected at steps 102 and / or 103, if a deviation of zero angular velocity and / or changes is detected magnetic field strength, using the previously obtained first intelligent model for changing the motion of an object. The first intellectual model of changing the movement of an object is obtained in advance on the basis of the characteristic features of the movement of at least one object of the same type as the said object.

At step 106 of the method of storied inertial navigation, when determining the presence of motion of an object, determine the current spatial position of the object based on the values of linear acceleration of the object adopted at step 101 and the values of the object’s angular velocity adopted at step 101, if no deviation of the gyroscope’s zero angular velocity and / or change in intensity is detected the magnetic field, or based on the values of linear acceleration of the object taken at step 101 and corrected values of the angular velocity of the object at steps 102 and / or 103 m, if a deviation of zero angular velocity and / or a change in the magnetic field strength is detected using a previously obtained second intellectual model of the object's motion. The second intellectual model of the movement of an object is obtained in advance on the basis of the characteristic features of the movement of at least one object of the same type as the said object. When determining the absence of motion of an object, the current spatial position of the object is set equal to the previous spatial position of the object.

Another embodiment of the method of inertial free navigation is shown in FIG. 2 and includes steps 101-106 described above. In addition, shown in FIG. 2, an embodiment of the method of strapless inertial navigation comprises an additional step 107, in which the linear acceleration values of the object and the values of the angular velocity of the object adopted at step 101 are filtered to reduce noise. Filtering the received values of the linear acceleration of the object and the accepted values of the angular velocity of the object are performed by means of weighted averaging using the sliding window method. In addition, any other suitable filtering method can be used to filter the received values of the linear acceleration of the object and the accepted values of the angular velocity of the object. Filtering by weighted averaging by the sliding window method is widely known in the art and its detailed description is not required.

угловой скорости

объекта для каждого значения рабочей температуры

инерциального датчика определяют, движется ли объект, на основании значений линейного ускорения объекта и значений угловой скорости объекта с использованием предварительно полученной первой интеллектуальной модели изменения движения объекта. Если определено, что объект не движется, формируют набор принятых значений угловой скорости

объекта для текущего значения рабочей температуры

инерциального датчика, причем оптимальное количество принятых значений угловой скорости

объекта в упомянутом наборе определяется на основании оптимизации соотношения смещения

среднего значения угловой скорости объекта и величины разброса var значений угловой скорости объекта. Затем усредняют принятые значения угловой скорости

объекта упомянутого набора методом взвешенного среднего и вычисляют поправочный коэффициент

для текущего значения рабочей температуры

инерциального датчика посредством решения регрессионного уравнения.Upon receipt of a correction factor

angular velocity

object for each value of operating temperature

the inertial sensor determines whether the object is moving, based on the values of the linear acceleration of the object and the values of the angular velocity of the object using the previously obtained first intelligent model for changing the movement of the object. If it is determined that the object is not moving, form a set of accepted values of angular velocity

object for the current value of the working temperature

inertial sensor, and the optimal number of accepted values of angular velocity

the object in the said set is determined based on the optimization of the displacement ratio

the average value of the angular velocity of the object and the magnitude of the var variation of the angular velocity of the object. Then average the accepted values of the angular velocity

the object of the above set by the method of weighted average and calculate the correction factor

for the current operating temperature

inertial sensor by solving a regression equation.

The formation of a set of accepted values of the angular velocity of the object and the calculation of the correction factor are performed for each value of the operating temperature of the inertial sensor.

находится как решение полиномиальной аппроксимации m-го порядка

. Вектор коэффициентов линейной полиномиальной аппроксимации

вычисляются из регрессионной модели

, где

есть вектор измерений отклонения нулевой угловой скорости гироскопа, содержащий оптимальное количество

элементов.Correction factor

is found as a solution to the mth order polynomial approximation

. The vector of linear polynomial approximation coefficients

calculated from the regression model

where

is the measurement vector of the deviation of the zero angular velocity of the gyroscope containing the optimal amount

items.

для каждого значения рабочей температуры

, регулируется исходя из достижения баланса между точностью аппроксимации (малое смещение) результата и достаточной сглаженностью (низкая дисперсия). Этот баланс достигается с помощью статистического критерия, который известен как пересечение доверительных интервалов (intersection of confidence intervals, ICI), впервые описанный в A. Goldenshluger, A. Nemirovski, On spatial adaptive estimation of non-parametric regression. Math Meth Statistics 6, 135-170 (1997), как аргумент минимизации среднеквадратичной функции потерь

, где

означает смещение среднего значения угловой скорости объекта, а var означает величину разброса значений угловой скорости объекта (дисперсию).The number of values of deviations of the gyroscope zero angular velocity used to determine the correction factor

for each value of working temperature

, is adjusted on the basis of achieving a balance between the approximation accuracy (small displacement) of the result and sufficient smoothness (low dispersion). This balance is achieved using a statistical criterion, which is known as the intersection of confidence intervals (ICI), first described in A. Goldenshluger, A. Nemirovski, On-line adaptive estimation of non-parametric regression. Math Meth Statistics 6, 135-170 (1997), as an argument for minimizing the root-mean-square loss function

where

means the shift of the average value of the angular velocity of the object, and var means the magnitude of the variation of the angular velocity of the object (variance).

значений угловой скорости объекта формируется вектор y значений угловой скорости объекта для условии, когда определено, что движения нет. Например, вектор y составляется из значений, евклидова норма угловой скорости объекта и евклидовой нормы линейного ускорения объекта которого менее соответствующими заранее определенными порогов

,

. Данные пороги

и

определяются на модельных данных, как среднее значение нормы линейного ускорения объекта и среднее значение угловой скорости объекта для заранее определенных случаем, когда сторонними средствами выявлено или задано, что объект не движется.Following [V. Katkovnik, K. Egiazarian and J. Astola, Local Approximation Techniques and Signal Processing, SPIE Press Monograph Vol. PM157 (2006)], to calculate the optimal amount

values of the angular velocity of the object is formed by the vector y of the values of the angular velocity of the object for the condition when it is determined that there is no movement. For example, the vector y is composed of values, the Euclidean norm of the angular velocity of the object and the Euclidean norm of the linear acceleration of the object of which are less than the corresponding predefined thresholds

,

. These thresholds

and

are determined on the model data as the average value of the norm of the linear acceleration of the object and the average value of the angular velocity of the object for a predetermined case when it is detected or specified by third-party means that the object does not move.

}_i, где результат свертки обозначен {

}. Пространственные фильтры {

}_i определяются через количество значений угловой скорости объекта {h_i} в векторе y (то есть длину вектора y) для рабочей температуры

. Тогда оптимальное количество значений

угловой скорости объекта в y для рабочей температуры t_k определяется как максимальное число значений в y, допускающее непустой доверительный интервал

, где

,

определяет стандартное отклонение, параметр

определяет порог доверительного интервала.The convolution of the vector of the accumulated values of the angular velocity of the object y with a set of filters {

} _i , where the result of the convolution is denoted by {

}. Spatial filters {

} _{i are} determined by the number of angular velocity values of the object {h _i } in the vector y (that is, the length of the vector y ) for the operating temperature

. Then the optimal number of values

the angular velocity of the object in y for the operating temperature t _k is defined as the maximum number of values in y allowing a non-empty confidence interval

where

,

determines the standard deviation parameter

determines the threshold of the confidence interval.

среднего значения угловой скорости объекта и величины разброса var значений угловой скорости объекта, как аргумент минимизации среднеквадратичной функции потерь

.FIG. 5 shows the choice of the optimal number of angular velocity values of an object based on the ratio of the displacement

the average value of the angular velocity of the object and the magnitude of the var spread of the values of the angular velocity of the object, as an argument for minimizing the rms loss function

.

для принятого значения напряженности

магнитного поля для коррекций принятых значений угловой скорости

объекта по изменению напряженности

магнитного поля принимают значения напряженности

магнитного поля, модель распределения напряженности магнитного поля и предварительные калибровочные параметры

и

магнетометра инерциального датчика, причем модель распределения напряженности магнитного поля и калибровочные параметры

и

вычислены заранее. Затем производится корректировка принятых значений напряженности

магнитного поля с использованием калибровочных параметров

и

магнетометра. Если скорректированные значения напряженности

магнитного поля удовлетворяют диапазону из принятой модели распределения напряженности магнитного поля, производится расчет поправочных коэффициентов

принятых значений угловой скорости

объекта как мера отклонения текущего вектора напряженности от значения упомянутой модели, если скорректированные значения напряженности

магнитного поля не удовлетворяют диапазону из принятой модели распределения напряженности магнитного поля, производится перерасчет калибровочных параметров

и

как задача ортогонализации и минимизации смещения среднего значения напряженности и кросс-корреляции между осями.When calculating correction factors

for the accepted value of tension

magnetic field for the correction of the accepted values of angular velocity

object to change the tension

Magnetic field strength values

magnetic field distribution model, magnetic field strength and preliminary calibration parameters

and

magnetometer inertial sensor, and the model of the distribution of magnetic field strength and calibration parameters

and

calculated in advance. Then the accepted values of tension are corrected.

magnetic field using calibration parameters

and

magnetometer If the corrected tension values

magnetic fields satisfy the range of the accepted model of the distribution of magnetic field strength, the correction factors are calculated

accepted values of angular velocity

of an object as a measure of the deviation of the current vector of intensity from the value of the model mentioned, if the corrected values of the intensity

the magnetic field does not satisfy the range of the adopted model of the distribution of the magnetic field strength, recalculation of the calibration parameters

and

as a task of orthogonalization and minimization of the displacement of the average value of tension and cross-correlation between the axes.

магнитного поля в i-й момент времени корректируется как

проекция на сферу единичного радиуса

с центром сферы в центре координат, то есть математическое ожидание

, где

и

принятые калибровочные параметры магнетометра инерциального датчика, минимизирующие неортогональность осей, смещение среднего значения напряженности и кросс-корреляцию между осями. Рассмотрим текущее скорректированное значение напряженности

магнитного поля и модельное значение

, заранее определенное в условиях однородного магнитного поля. Поправка пространственной ориентации объекта выражается через разницу направлений значений напряженности

и

магнитного поля в плоскости, ортогональной вектору ускорения свободного падения

, где верхний индекс

и

означает проекция векторов напряженности магнитного поля на соответствующие оси координат x и z.Accepted value of tension

the magnetic field at the i-th instant is corrected as

projection onto a sphere of unit radius

with the center of the sphere in the center of coordinates, that is, the expectation

where

and

adopted calibration parameters of the inertial sensor magnetometer, minimizing the non-orthogonality of the axes, the shift of the average value of the tension and the cross-correlation between the axes. Consider the current adjusted tension value.

magnetic field and model value

pre-determined in a uniform magnetic field. The correction of the spatial orientation of the object is expressed through the difference in the directions of the tension values

and

the magnetic field in a plane orthogonal to the gravitational acceleration vector

where superscript

and

means the projection of the magnetic field strength vectors to the corresponding x and z coordinate axes.

объекта добиваются путем компенсации измерений значения магнитного поля на поправочный коэффициент

в виде

, где

означает скорректированное значение угловой скорости объекта для скорректированного значения напряженности

магнитного поля, а поправочный коэффициент

выражается как

и отражает отклонение текущего вектора напряженности магнитного поля от модельного значения при условии, что скорректированное значение напряженности магнитного поля удовлетворяет диапазону из принятой модели распределения напряженности магнитного поля. Например, в качестве модели распределения напряженности магнитного поля используют нормальное распределение. Здесь

есть коэффициент сглаживания,

есть матрица поворота согласования для согласования с текущей пространственной ориентации объекта. Improved accuracy and stability of angular velocity measurements

an object is obtained by compensating for measurements of the magnetic field value by a correction factor

as

where

means the corrected value of the angular velocity of the object for the corrected value of tension

magnetic field, and the correction factor

expressed as

and reflects the deviation of the current vector of magnetic field strength from the model value, provided that the corrected value of the magnetic field strength satisfies the range from the adopted model of the distribution of the magnetic field strength. For example, the normal distribution is used as a model for the distribution of magnetic field strength. Here

there is a smoothing factor,

There is a rotation rotation matrix for matching with the current spatial orientation of the object.

и

выполняют как решение оптимизационной задачи

из соображений, что значение напряженности магнитного поля не выходит за пределы 2-х стандартных отклонений от среднего значения. Матрица

формируется из допустимых значений напряженности магнитного поля, путем их объединения по столбцам.Recalculation of calibration parameters

and

perform as a solution to the optimization problem

from considerations that the value of the magnetic field strength does not exceed 2 standard deviations from the average value. Matrix

formed from the permissible values of the magnetic field strength, by combining them in columns.

The determination of whether an object is moving can be made by comparing the value of the Euclidean norm of the angular velocity of an object with one constant and comparing the value of the Euclidean norm of linear acceleration of the object with another constant. The mentioned comparisons and the mentioned constants are set in the first intellectual model of the change in the motion of an object on the basis of the characteristic features of the motion of at least one object of the same type as the said object. It is determined that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object do not exceed the mentioned constants, respectively, and determines that the object moves if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object exceed mentioned constants, respectively.

The determination of whether an object is moving can also be made by comparing the value of the Euclidean norm of linear acceleration of an object with one adaptive threshold by the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold by the sliding window method, and the method of the above comparison with adaptive thresholds and method calculations of adaptive thresholds using the sliding window method are specified in the first intelligent model for changing the motion of an object.

The calculation of the adaptive threshold and the comparison with it of the Euclidean norm of linear acceleration of an object to determine whether an object is moving is performed iteratively and includes the following operations. A pre-calculated adaptive threshold is set as an adaptive threshold, and it is pre-considered that the object does not move. Then, a predetermined threshold for determining movement and a predetermined threshold for determining the absence of movement are taken. The threshold for determining movement and the threshold for determining the absence of movement are specified in the first intellectual model of the change in the movement of an object based on the characteristic features of the movement of at least one object of the same type as the said object. A comparison is made of the value of the norm of the linear acceleration of an object with an adaptive threshold, and if the value of the norm of the linear acceleration of an object does not exceed the adaptive threshold, then it is determined that the object does not move. If the value of the norm of linear acceleration of an object exceeds the adaptive threshold, then I determine the duration, how long the norm of the norm of linear acceleration of an object exceeds the adaptive threshold. If the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold exceeds the threshold for determining motion, then it is determined that the object moves, and the adaptive threshold does not change. If the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move. If it is determined that the object is not moving, recalculate the current adaptive threshold by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigning the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value covariance matrix. If it is determined that the object is moving, compare the value of the norm of the linear acceleration of the object with the adaptive threshold, and if the value of the norm of the linear acceleration of the object exceeds the adaptive threshold, then determine that the object continues to move. If the value of the norm of the linear acceleration of the object does not exceed the adaptive threshold, the completion of the movement is determined by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion that the object does not move, while the adaptive threshold is recalculated in accordance with the norm of the input values of the linear acceleration of the object.

The adaptive threshold can be calculated as the value of the norm of linear acceleration of objects of the same type, averaged over many different test data, at which movement is uniquely detected.

The determination of whether an object is moving can be carried out using at least one previously trained recurrent neural network as the first intelligent model for changing the movement of an object. The determination of whether an object is moving can also be carried out using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of an object. Each recurrent neural network can be a long-term short-term memory (LSTM). Pre-training is performed on the basis of the characteristic features of the movement of at least one object of the same type as the said object.

The proposed approach uses recurrent neural networks (RNN), that is, a type of neural networks where the connections between elements, the so-called states, form a directed sequence. Thanks to the transfer of states, it is possible to process a series of events in time or successive spatial chains. Recurrent networks can use their internal memory to process sequences of arbitrary length, so the RNN networks are applicable in tasks where the entire sequence is divided into segments, for example, in handwriting recognition, speech recognition or motion recognition tasks.

In this invention, a special kind of RNN architecture can be used for motion detection, namely, Long short-term memory (LSTM). Unlike traditional recurrent neural networks, the LSTM network is well adapted to learning on the tasks of classifying, processing and predicting time series in cases where important events are separated in time with indefinite duration and boundaries. The relative immunity to the duration of the time breaks is due to the possibility of forgetting irrelevant information, which gives LSTM an advantage over alternative recurrent neural networks, hidden Markov models, and other learning methods for sequences in various applications.

An LSTM network is an artificial neural network that contains LSTM modules instead of or in addition to other network modules. An LSTM module is a recurrent network module capable of storing values for both short and long periods of time. The key to this feature is that the LSTM module does not use the activation function within its recurrent components. Thus, the stored value is not blurred in time, and the gradient or penalty does not disappear when using the method of back propagation of error in time during the training of the network. LSTM modules are grouped into “blocks” containing various LSTM modules. Such a device is typical for "deep" multilayer neural networks and contributes to the implementation of parallel computing using the appropriate equipment. In the formulas below, each variable written in lowercase italics denotes a dimension vector equal to the number of LSTM modules in a block.

In this invention, grouping into blocks allows making a sequential decision on the presence of movement: first, according to the many-to-many scheme, when groups of features are distinguished from the data obtained from inertial sensors, and then they are gradually reduced down to the last unit implementing the many to one ”, where the final decision is made on the presence or absence of movement.

LSTM blocks contain three or four “gates” that are used to control the flow of information at the inputs and at the memory outputs of the data blocks. These valves are implemented as a logistic function for calculating the value in the range [0, 1]. Multiplication by this value is used to partially allow or prohibit the flow of information into and out of memory. For example, the “input valve” controls the measure of the occurrence of a new value in the memory, and the “forgetting valve” controls the measure of the value stored in the memory. The “output valve” controls the measure of the extent to which the value stored in memory is used when calculating the output activation function for a block.

The weights in the LSTM block, expressed in the matrixes of the parameters W and U , are used to specify the direction of operation of the indicated gates through the transformation of the input x _t and output from the previous time step of the vectors h _t-1 . Traditional LSTM with forgetting valves with ₀ = 0 and h ₀ = 0 can be represented as

обозначает поэлементное произведение Адамара, σ _g , σ _c и σ _h означают функцию активации: сигмоидальную, на основе гиперболического тангенса и гиперболического тангенса, соответственно. x _t , h _t и c _t означают входной, выходной вектора и вектор состояний, соответственно. i _t , o _t и f _t - вектора входного, выходного вентилей и вентиля забывания, соответственно. W , U - матрицы параметров, а b - вектор смещения, определяющие согласно нижнему индексу _f, _i, _o и _c преобразование состояния, входного, выходного векторов и вентиля забывания.Where

denotes the elementwise Hadamard product, σ _g , σ _c and σ _h mean the activation function: sigmoidal, based on the hyperbolic tangent and the hyperbolic tangent, respectively. x _t , h _t and c _t denote the input, output vector and state vector, respectively. i _t , o _t and f _t are the input, output, and forgetting vectors, respectively. W , U are matrixes of parameters, and b is the displacement vector, which determines, according to the subscript _f , _i , _o and _c, the state transformation of the input, output vectors and the forgetting valve.

Scale training allows the LSTM unit to learn the function that determines the rule for managing the memory. LSTM blocks are usually trained using the time-error propagation method.

In this invention, LSTM networks can be used as the first intelligent model for changing the motion of an object and the second intelligent model of an object's motion, as a set of LSTM modules sequentially connected first according to the many-to-many scheme to highlight the features of the movement up to the last block that implements “many-to-one” scheme for deciding on the presence or absence of object motion for the first intellectual model of the object motion change and for estimating the synthetic object velocity for the second engine tele-model of the movement of the object.

Determining the current spatial position of the object at step 106 in determining the presence of movement of the object contains the following operations. Determine the vector of the direction of movement of the object by calculating the average values of the linear acceleration of the object in the global coordinate system using the sliding window method and linking this average value to the coordinate system in which the vertical axis is parallel to the gravitational acceleration vector. Then, the object velocity is estimated based on the linear acceleration values of the object using the second intellectual model of the object's motion, the estimated object velocity is projected onto a certain vector of the object's direction of motion, and the current spatial position of the object is obtained by numerical integration.

The linear velocity of the object is estimated based on the linear acceleration values of the object using the polynomial velocity model as the second intelligent model of the object's motion.

,

и

в k-й момент времени

соответственно вдоль осей x, y и z с использованием локальной полиномиальной аппроксимации m-го порядка в виде

, где коэффициенты линейной полиномиальной аппроксимации

,

и

для осей x, y и z вычисляются как аргумент минимизации соответствующей функции

,

и

соответственно.Let there be a set { π _j } of temporal sequences of linear velocity variation in a 3-dimensional space, characteristic of the movement of an object of the type under consideration. The proposed approach calculates the components of the synthetic linear velocity of the object.

,

and

at the k-th instant

respectively along the x, y and z axes using the local m-th order polynomial approximation in the form

where the coefficients of linear polynomial approximation

,

and

for the x, y and z axes are calculated as the argument to minimize the corresponding function

,

and

respectively.

, где

означает усреднение нескольких предыдущих k-му моменту времени

значений линейного ускорения объекта

, p=0,1,… методом скользящего окна. Тогда значение текущей пространственной позиции объекта вычисляется в вычисленном направлении движения объекта

методом численного интегрирования.The vector of the direction of movement of the object determines the measure of movement in a 3-dimensional space and is calculated as

where

means averaging several previous kth time point

object linear acceleration values

, p = 0,1, ... using the sliding window method. Then the value of the current spatial position of the object is calculated in the calculated direction of the object.

numerical integration method.

The vector of the direction of movement of the object can be determined by the support vector machine (SVM).

The vector of the direction of motion of the object can be determined using a pre-trained recurrent neural network.

The linear velocity of the object can be estimated based on the values of the linear acceleration of the object using a pre-trained recurrent neural network as the second intelligent model of the movement of the object.

The linear velocity of an object can also be estimated based on the linear acceleration values of the object using a combination of pre-trained models as the second intelligent model of object motion.

Pre-training is performed on the basis of the characteristic features of the movement of at least one object of the same type as the said object.

In the method of free-form inertial navigation, it is also possible to receive the object linear acceleration values, the object angular velocity values, temperature values and magnetic field strength values measured by more than one inertial sensor at step 101. Then, at step 104, the current spatial orientation of the object is independently determined based on the received values measured by each sensor, and at step 106 independently determine the current spatial position of the object based on the received values measured each by the sensor. The method further comprises the following operations. The resulting current spatial orientation of the object is calculated based on all current spatial orientations of the object, determined independently based on received values from different inertial sensors, and the resulting current spatial position of the object is calculated based on all current spatial positions of the object, determined independently based on received values from different inertial sensors .

The calculation of the resulting values of the current spatial orientation of the object and the calculation of the current spatial position of the object can be done by weighted averaging the current spatial orientations of the object and the current spatial positions of the object, determined independently based on the received values measured by different inertial sensors or by other methods known in the art.

The strapdown inertial navigation device operates according to the strapdown inertial navigation method, and the blocks of the said device respectively perform all the operations of the method described above. Therefore, a detailed description of the operations performed by the units of the strapdown inertial navigation device is not given to avoid repetition.

One embodiment of a strapdown inertial navigation device is shown in FIG. 3. In the present embodiment, the strapdown inertial navigation device includes a temperature correction unit 201, a magnetic field correction unit 202, a spatial orientation determination unit 203, a motion detection unit 204 and a spatial position determination unit 205.

Another embodiment of a strapdown inertial navigation device is shown in FIG. 4. In the present embodiment, the strapdown inertial navigation device contains the above blocks 201-205 and filter 206 for filtering the object linear acceleration values and the object angular velocity values for noise suppression. The filter 206 filters the linear acceleration values of the object and the angular velocity values of the object by means of a weighted averaging using the sliding window method. In addition, any other suitable filtering method can be used to filter the received values of the linear acceleration of the object and the accepted values of the angular velocity of the object.

The temperature correction unit 201 performs correction of the angular velocity values of the object by a correction factor for the current temperature value, if a deviation of the zero angular velocity of the gyroscope of the inertial sensor is detected. A set of correction factors for each value of the operating temperature of the inertial sensor is obtained in advance.

The magnetic field correction unit 202 corrects the angular velocity of the object by a correction factor if a change in the magnetic field strength is detected. The value of the correction factor is calculated as a solution to the optimization problem.

The spatial orientation determination unit 203 determines the current spatial orientation of the object based on the angular velocity values, if no deviation of the zero angular velocity and / or a change in the magnetic field strength, or angular velocity values of the object corrected by the blocks 201 and / or 202 are detected, if a deviation of the zero angular speeds and / or changes in the magnetic field.

The motion detection unit 204 determines whether the object moves based on the linear acceleration values of the object and the angular velocity values of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the linear acceleration values of the object and angular velocity values of the object, corrected by blocks 201 and / or 202, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the previously obtained first intellectual Second model of the object changes. The first intellectual model of the change in the motion of an object is obtained in advance on the basis of the characteristic features of the motion of at least one object of the same type as the said object.

The spatial position determining unit 205 in determining the presence of movement of an object determines the current spatial position of the object based on the linear acceleration values of the object and the angular velocity values of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or values of the angular velocity of the object, corrected by blocks 201 and / or 202, if a deviation of zero angular velocity and / or a change in the intensity is detected rotten field, using a previously obtained second intellectual model of the object. A second intellectual model of the movement of an object is obtained in advance on the basis of the characteristic features of the movement of at least one object of the same type as the said object. When determining the absence of movement of an object, the spatial position determination unit 205 sets the current spatial position of the object to be equal to the previous spatial position of the object.

The free inertial navigation device uses the object's linear acceleration values measured by an inertial sensor accelerometer, object angular velocity values measured by an inertial sensor gyroscope, inertial sensor temperatures measured by an inertial sensor thermometer, and magnetic field intensity values measured by an inertial sensor magnetometer, and the linear acceleration values of the object the angular velocity of the object, the temperature of the inertial sensor and the magnetic field strengths acting on the inertial sensor magnetometer are measured by at least one inertial sensor including at least an accelerometer, a gyroscope, a thermometer and a magnetometer.

The temperature correction unit 201 obtains a correction factor for each value of the operating temperature of the inertial sensor when determining that the object is not moving. The temperature correction unit 201 generates a set of angular velocity values of the object for the current value of the inertial sensor operating temperature, and the optimal number of angular velocity values of the object in said array is determined based on optimizing the ratio of the offset of the mean angular velocity of the object and the magnitude of the angular velocity variation of the object. Then, the temperature correction unit 201 averages the angular velocity values of the object of the above set by the weighted average method and calculates a correction factor for the current value of the operating temperature of the inertial sensor by solving the regression equation. The formation of a set of values of the angular velocity of the object and the calculation of the correction factor is performed for each value of the operating temperature of the inertial sensor. The temperature correction unit 201 determines the optimal number of angular velocity values of the object in the above-mentioned set by the method of intersection of the confidence intervals.

The magnetic field correction unit 202 calculates correction factors for the magnetic field strength value for correcting the angular velocity values of the object from changes in the magnetic field strength by the following operations.

The magnetic field correction unit 202 corrects the magnetic field strength values using the calibration parameters of the magnetometer. Calibration parameters are calculated in advance. If the corrected values of the magnetic field strength satisfy the range from the model of the distribution of the magnetic field strength that was calculated in advance, the magnetic field correction unit 202 calculates the correction factors for the angular velocity of the object as a measure of the deviation of the current intensity vector from the value of the model. If the corrected values of the magnetic field strength do not satisfy the range from the magnetic field strength distribution model, the magnetic field correction unit 202 recalculates the calibration parameters as an orthogonalization task and minimizing the displacement of the average value of the strength and cross-correlation between the axes.

The spatial orientation determination unit 203 determines the current spatial orientation of the object by computing with a numerical integration method.

Motion detection unit 204 determines whether an object is moving by comparing the value of the Euclidean norm of the angular velocity of an object with one constant and comparing the value of the Euclidean norm of linear acceleration of an object with another constant, and these comparisons with constants and constants are specified in the first intellectual model of the object’s movement based on characteristic features of the movement of at least one object of the same type as the said object. Motion detection unit 204 determines that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object do not exceed the above constant, and determines that the object moves if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of linear acceleration object exceed the constant mentioned.

Motion detection unit 204 can also determine whether an object is moving by comparing the value of the Euclidean norm of linear acceleration of an object with one adaptive threshold using the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold using the sliding window method. The method of the mentioned comparison with adaptive thresholds and the method of calculating adaptive thresholds using the sliding window method are specified in the first intellectual model of the change in the motion of an object.

Motion detection unit 204 calculates the adaptive threshold and compares with it the Euclidean norm of the linear acceleration of the object iteratively. For this, motion detection unit 204 establishes an adaptive threshold, it being pre-considered that the object does not move, and the adaptive threshold is calculated in advance. Then, motion detection unit 204 receives a predetermined threshold for determining motion and a predetermined threshold for determining no motion and compares the value of the norm of the linear acceleration of the object with the adaptive threshold. The threshold for determining movement and the threshold for determining the absence of movement are specified in the first intellectual model of the change in the movement of an object based on the characteristic features of the movement of at least one object of the same type as the said object. If the value of the norm of the linear acceleration of an object does not exceed the adaptive threshold, then it is determined that the object does not move. If the value of the norm of linear acceleration of the object exceeds the adaptive threshold, then the duration is determined, how long the norm of the norm of linear acceleration of the object exceeds the adaptive threshold. If the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold exceeds the threshold for determining movement, then it is determined that the object moves, while the adaptive threshold does not change. If the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move. If the object is determined not to move, block 204 determines the movement of the current adaptive threshold by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigns the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value between the calculated average weighted value and the maximum value of the covariance matrix. If it is determined that the object is moving, block 204 motion detection compares the value of the norm of the linear acceleration of the object with an adaptive threshold. If the value of the norm of the linear acceleration of the object exceeds the adaptive threshold, then block 204 motion detection determines that the object continues to move. If the value of the norm of the linear acceleration of an object does not exceed the adaptive threshold, then motion detection unit 204 determines the completion of motion by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion , then the motion detection unit 204 determines that the object does not move, and the adaptive threshold is recalculated in accordance with the norm of the input values accelerate the object. The adaptive threshold is calculated as the value of the norm of linear acceleration of objects of the same type, averaged over a variety of different test data, at which movement is uniquely detected.

Motion detection unit 204 may also contain at least one pre-trained recurrent neural network and determine if the object is moving using at least one pre-trained recurrent neural network as the first intelligent model for changing the movement of the object.

Motion detection unit 204 may also comprise a cascade of pre-trained recurrent neural networks and determine whether an object is moving using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of an object.

Each recurrent neural network can be a long-term short-term memory (LSTM).

The spatial position determination unit 205 determines the current spatial position of an object when determining the presence of an object's motion by determining the object's direction vector by calculating the average value of the linear acceleration values of the object in the global coordinate system using the sliding window method and associating this average value to the coordinate system in which the vertical axis is parallel the vector of the acceleration of free fall, the estimated speed of the object based on the values of the linear acceleration an object using the second intellectual model of the object's motion, projecting the estimated object velocity onto a certain vector of the object's direction of motion and obtaining the current spatial position of the object by the method of numerical integration.

The spatial position determination unit 205 can determine the object direction vector of the object by the support vector machine (SVM).

The spatial position determination unit 205 may comprise a pre-trained recurrent neural network and determine the direction vector of the object using a pre-trained recurrent neural network.

The spatial position determining unit 205 may estimate the linear velocity of the object based on the linear acceleration values of the object using the polynomial velocity model as the second intelligent model of the motion of the object.

The spatial position determination unit 205 may comprise a pre-trained recurrent neural network and estimate the object linear velocity based on the object's linear acceleration values using the pre-trained recurrent neural network as the second intelligent model of the object's motion.

The spatial position determination unit 205 may comprise a combination of pre-trained models and estimate the linear velocity of the object based on the linear acceleration values of the object using a combination of pre-trained models as the second intelligent model of the movement of the object.

Values of linear acceleration of an object, values of an angular velocity of an object, values of temperature and values of magnetic field strength measured by more than one inertial sensor can be received into the device of free-form inertial navigation.

When data is received from more than one inertial sensor, the spatial orientation determination unit 203 independently determines the current spatial orientation of the object based on the values measured by each sensor, and calculates the resulting current spatial orientation of the object based on all the current spatial orientations of the object determined independently based on the received values different inertial sensors. The spatial position determining unit 205 independently determines the current spatial position of the object based on the values measured by each sensor, and calculates the resulting current spatial position of the object based on all the current spatial positions of the object determined independently based on the received values from different inertial sensors.

When data is received from more than one inertial sensor, the spatial orientation determining unit 203 may calculate the resulting current spatial orientation of the object by weighted averaging the current spatial orientations of the object, determined independently based on the values measured by different inertial sensors; and the spatial position determination unit 205 may calculate the resulting current spatial position of the object by weighted averaging of the current spatial positions of the object, determined independently based on the received values measured by different inertial sensors.

FIG. 6 is a block diagram illustrating the use of the invention as a device for recognizing and authenticating a user for specific samples of human movement. The user recognition and authentication device includes a temperature correction unit 201, a magnetic field correction unit 202, a spatial orientation determination unit 203, a motion definition unit 204, and a spatial position determination unit 205 that fully correspond to the blocks 201-205 described above. In addition, the device recognition and authentication of the user may contain the above-described filter 206.

The device recognition and authentication of the user further comprises a block 207 for extracting the vector of signs of movement of the object. The values of the spatial orientation of the object obtained in block 203 and the values of the spatial position of the object obtained in block 205 are received in block 207 of the selection of the vector of signs of motion of the object, which recognizes and authenticates the user by specific patterns of human movement.

FIG. 7 is a block diagram illustrating the use of the invention as a system for automatically controlling the movement of a vehicle (“cruise control”). The automatic vehicle motion control system includes a temperature correction unit 201, a magnetic field correction unit 202, a spatial orientation determination unit 203, a motion detection unit 204, and a spatial position determination unit 205 that fully correspond to the blocks 201-205 described above. In addition, the system for automatically controlling the movement of the vehicle may contain the filter 206 described above.

The automatic vehicle motion control system further comprises a vehicle motion correction unit 208. The spatial orientation values obtained in block 203 and the spatial position values of the object obtained in block 205 are received in vehicle motion correction block 208, which performs vehicle motion refinement using additional spatial position and spatial orientation values obtained from external independent sensors, such as camera, global navigation system sensor, etc., in order to prevent an abnormal vehicle emergency.

FIG. 8 is a block diagram illustrating the use of the invention as a system for automatically controlling the movement of any vehicle in 3-dimensional space, including unmanned versions. The automatic vehicle motion control system in 3-dimensional space contains a temperature correction unit 201, a magnetic field correction unit 202, a spatial orientation determination unit 203, a motion determination unit 204 and a spatial position determination unit 205, which fully correspond to blocks 201-205, described above. In addition, the system for automatically controlling the movement of the vehicle may contain the filter 206 described above.

The system of automatic control of the movement of the vehicle in 3-dimensional space further comprises a block 209 for correcting the 3-dimensional movement of the vehicle. The spatial orientation values of the object obtained in block 203 and the spatial position values of the object obtained in block 205 are received by the 3-dimensional vehicle movement correction block 209, which corrects the 3-dimensional movement of the vehicle using additional spatial position and spatial orientation values obtained from external independent sensors, such as a camera, a global navigation system sensor, etc., in order to prevent an abnormal emergency; tnogo means.

The above descriptions of embodiments of the invention are illustrative, and configuration modifications and implementations are within the scope of the present description. For example, although embodiments of the invention have been described in general in connection with FIGS. 1-8, the descriptions given are exemplary. Although the subject matter of the invention is described in a language characteristic of constructive features or methodological operations, it is understood that the subject matter of the invention defined by the appended claims is not necessarily limited to the specific features or operations described above. Moreover, the specific features and operations described above are disclosed as exemplary forms of implementing the claims.

Accordingly, it is assumed that the scope of the invention is limited only by the following claims.

Claims (105)

1. A method of free inertial navigation, comprising the steps of: take the values of the linear acceleration of the object measured by the accelerometer of the inertial sensor, the angular velocity of the object measured by the gyroscope of the inertial sensor, the temperature of the inertial sensor measured by the thermometer of the inertial sensor, and the magnetic field intensity measured by the magnetometer of the inertial sensor, and the values of the linear acceleration of the object angular values object velocity, inertial sensor temperature values and magnetic field strength values, guide for the inertial sensor magnetometer, are measured by at least one inertial sensor; if a deviation of the zero angular velocity of the gyroscope of the inertial sensor is detected, the accepted values of the angular velocity of the object are corrected by the correction factor for the current temperature value, and a set of correction factors for each value of the operating temperature of the inertial sensor is obtained in advance; if a change in the magnetic field strength is detected, correction of the accepted values of the angular velocity of the object by the correction factor is performed, and the value of the correction factor is calculated as a solution to the optimization problem; determine the current spatial orientation of the object based on the accepted values of the angular velocity, if no deviation of the zero angular velocity and / or a change in the magnetic field strength, or corrected values of the angular velocity of the object is detected, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected; determine whether the object is moving based on the accepted values of the linear acceleration of the object and the adopted values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the adopted values of the linear acceleration of the object and the corrected values of the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the previously obtained first intelligent model of change the object, the first intellectual model object changes the motion obtained in advance based on the characteristics of movement of at least one object of the same type as that of said object; and when determining the presence of movement of an object, determine the current spatial position of the object based on the accepted values of the linear acceleration of the object and the adopted values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or on the basis of the adopted values of the linear acceleration of the object and corrected values the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using a second intellectual model of the object’s motion obtained in person, the second intellectual model of the object’s motion obtained in advance based on the characteristic motion characteristics of at least one object of the same type as the said object, and when determining the absence of the object’s motion, the current spatial position of the object is set equal to the previous spatial position object. 2. A method according to claim 1, further comprising the step of filtering the received values of the linear acceleration of the object and the received values of the angular velocity of the object to suppress noise, however, the filtering of the accepted values of the linear acceleration of the object and the adopted values of the angular velocity of the object is performed by means of weighted averaging using the sliding window method. 3. A method according to claim 1, wherein obtaining a correction factor for each value of the operating temperature of the inertial sensor comprises the steps of: determining whether the object is moving based on the linear acceleration values of the object and the angular velocity values of the object using the previously obtained first intelligent model for changing the movement of the object; if it is determined that the object is not moving, form a set of accepted values of the angular velocity of the object for the current value of the operating temperature of the inertial sensor, and the optimal number of received values of the angular velocity of the object in the set is determined based on optimizing the ratio of the displacement of the average angular velocity of the object and the magnitude of the spread of the angular object speed; averaging the received values of the angular velocity of the object of the above set by the method of weighted average and calculating the correction factor for the current value of the inertial sensor operating temperature by solving the regression equation, however, the formation of a set of the accepted values of the angular velocity of the object and the calculation of the correction factor are performed for each value of the operating temperature of the inertial sensor. 4. The method according to claim 3, in which the optimal number of received values of the angular velocity of the object in the above-mentioned set is determined by the method of intersection of the confidence intervals. 5. A method according to claim 1, wherein the calculation of the correction factors for the received value of the magnetic field strength for correcting the received values of the angular velocity of the object from the change in the magnetic field strength comprises the steps of: take the values of the magnetic field strength, the model of the distribution of the magnetic field strength and the preliminary calibration parameters of the magnetometer of the inertial sensor, and the model of the distribution of the magnetic field strength and the calibration parameters are calculated in advance; the accepted values of the magnetic field strength are adjusted using the calibration parameters of the magnetometer; If the corrected values of the magnetic field strength satisfy the range of the accepted model of the distribution of the magnetic field strength, the correction coefficients of the accepted values of the angular velocity of the object are calculated as a measure of the deviation of the current vector of intensity from the model, if the corrected values of the magnetic field strength magnetic field, the calibration parameters are recalculated as orthogonalization and minimization of the mean value of the tension and cross-correlation between the axes. 6. A method according to claim 1, in which the current spatial orientation of the object is calculated by the method of numerical integration. 7. The method according to claim 1 or 3, in which the determination of whether the object moves is performed by comparing the value of the Euclidean norm of the angular velocity of the object with one constant and comparing the value of the Euclidean norm of the linear acceleration of the object with another constant, and the above comparisons with constants are specified in the first intellectual model of changing the movement of an object, it is determined that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object do not exceed the above constant, and it is determined that the object moves if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object exceed mentioned constant. 8. The method according to claim 1 or 3, wherein determining whether an object moves is performed by comparing the value of the Euclidean norm of the linear acceleration of an object with one adaptive threshold by the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold by the sliding window method, moreover, the method of the above comparison with adaptive thresholds and the method of calculating adaptive thresholds using the sliding window method are specified in the first intellectual model of the change in the motion of an object. 9. The method according to claim 8, in which the calculation of the adaptive threshold and the comparison with it of the Euclidean norm of the linear acceleration of the object is performed iteratively and contains steps in which: establish an adaptive threshold, and previously it is considered that the object does not move and the adaptive threshold is calculated in advance; taking a predetermined threshold for determining movement and a predetermined threshold for determining the absence of movement; comparing the value of the norm of the linear acceleration of an object with an adaptive threshold, and if the value of the norm of the linear acceleration of the object does not exceed the adaptive threshold, then it is determined that the object does not move; if the value of the norm of linear acceleration of an object exceeds the adaptive threshold, then determine the duration, how long the norm of the norm of linear acceleration of the object exceeds the adaptive threshold; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold exceeds the threshold for determining movement, then it is determined that the object moves, while the adaptive threshold does not change; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move; if it is determined that the object is not moving, recalculate the current adaptive threshold by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigning the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value covariance matrix; if it is determined that the object is moving, compare the value of the norm of the linear acceleration of the object with the adaptive threshold, and if the value of the norm of linear acceleration of an object exceeds the adaptive threshold, then it is determined that the object continues to move; if the value of the norm of linear acceleration of an object does not exceed the adaptive threshold, then the completion of the movement is determined by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion, then determine that the object does not move, while the adaptive threshold is recalculated in accordance with the norm of the input values of the linear acceleration of the object. 10. The method according to claim 9, in which the adaptive threshold is calculated as averaged over a variety of different test data, the value of the norm of linear acceleration of objects of the same type at which movement is uniquely detected. 11. The method according to claim 1 or 3, wherein determining whether the object is moving is carried out using at least one previously trained recurrent neural network as the first intelligent model for changing the movement of the object. 12. The method according to claim 11, wherein determining whether an object moves is performed using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of an object. 13. The method according to claim 11 or 12, in which each recurrent neural network is a long short-term memory (LSTM). 14. The method according to claim 1, wherein determining the current spatial position of the object when determining the presence of motion of the object comprises the steps of: determine the vector of the direction of movement of the object by calculating the average values of the linear acceleration of the object in the global coordinate system using the sliding window method and linking this average value to the coordinate system in which the vertical axis is parallel to the gravitational acceleration vector; estimate the speed of the object based on the values of the linear acceleration of the object using the second intellectual model of the movement of the object; projecting the estimated velocity of the object onto a certain vector of the direction of motion of the object and obtaining the current spatial position of the object by a method of numerical integration. 15. A method according to claim 14, in which determine the direction vector of the object by the method of reference vectors (SVM). 16. A method according to claim 14, in which determine the vector of the direction of movement of the object using a pre-trained recurrent neural network. 17. A method according to any one of claims. 14-16, in which the linear velocity of the object is estimated based on the values of the linear acceleration of the object using the polynomial velocity model as the second intelligent model of the motion of the object. 18. A method according to any one of claims. 14-16, in which the linear velocity of the object is estimated based on the values of the linear acceleration of the object using the pre-trained recurrent neural network as the second intelligent model of the movement of the object. 19. A method according to any one of claims. 14-16, in which the linear velocity of the object is estimated based on the values of the linear acceleration of the object using a combination of pre-trained models as the second intelligent model of the movement of the object. 20. A method according to claim 1, wherein, if the values of the linear acceleration of the object, the values of the angular velocity of the object, the temperature values and the magnetic field strength values measured by more than one inertial sensor are taken, then the current spatial orientation of the object is independently determined based on the received values measured by each sensor, and independently determine the current spatial position of the object based on the adopted values measured by each sensor, wherein the method further comprises the steps of: calculate the resulting current spatial orientation of the object based on all the current spatial orientations of the object, determined independently based on the received values from different inertial sensors; and calculate the resulting current spatial position of the object based on all the current spatial positions of the object, determined independently based on the received values from different inertial sensors. 21. A method according to claim 20, in which the calculation of the resulting values of the current spatial orientation of the object and the calculation of the current spatial position of the object are made by weighted averaging the current spatial orientations of the object and the current spatial positions of the object, determined independently based on the adopted values measured by different inertial sensors. 22. The device is free-flow inertial navigation, containing: a temperature correction unit, configured to, if a deviation of the zero angular velocity of the gyroscope of the inertial sensor is detected, correct the angular velocity values of the object by the correction factor for the current temperature value, with a set of correction factors for each value of the inertial sensor operating temperature obtained in advance; a magnetic field correction unit, configured to, if a change in the magnetic field strength is detected, perform correction of the angular velocity values of the object by the correction factor, and the value of the correction factor is calculated as a solution to the optimization problem; a spatial orientation determination unit configured to determine the current spatial orientation of an object based on the angular velocity values, if no deviation of the zero angular velocity and / or a change in the magnetic field strength, or corrected angular velocity values of the object is detected, if a deviation of the zero angular velocity and / or change in magnetic field strength; a motion detection unit configured to determine whether the object is moving based on the linear acceleration values of the object and the angular velocity values of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, or based on the linear acceleration values of the object and corrected values the angular velocity of the object, if a deviation of the zero angular velocity and / or a change in the magnetic field strength is detected, using the previously obtained first intel ektualnoy change model of the object, wherein the first intelligent model changes of the object is obtained in advance based on the characteristics of movement of at least one object of the same type as that of said object; and a spatial position determination unit, configured to determine the current spatial position of the object based on the linear acceleration values of the object and the values of the angular velocity of the object, if no deviation of the zero angular velocity and / or a change in the magnetic field is detected, or based on values of the linear acceleration of the object and the adjusted values of the angular velocity of the object, if a deviation of zero angular velocity and / or a change in voltage is detected magnetic field using the previously obtained second intellectual model of the object’s movement, the second intellectual model of the object’s movement obtained in advance based on the characteristic motion characteristics of at least one object of the same type as the said object, and the spatial position of the object is equal to the previous spatial position of the object, the values of the linear acceleration of the object are measured by the accelerometer of the inertial sensor, the angular velocity of the object is measured by the gyroscope of the inertial sensor, the temperature of the inertial sensor is measured by the inertial sensor thermometer and the magnetic field of the inertial sensor, and the values of the linear acceleration of the object, the values of the angular velocity of the object, temperature of the inertial sensor and the value of the magnetic field strength acting on the mag etometr inertial sensor, measured at at least one inertial sensor. 23. The device according to claim 22, further comprising a filter configured to filter the values of the linear acceleration of the object and the values of the angular velocity of the object to suppress noise, at the same time filtering the values of the linear acceleration of the object and the values of the angular velocity of the object are performed by means of weighted averaging using the sliding window method. 24. The device according to claim 22, in which the temperature correction unit obtains a correction factor for each value of the operating temperature of the inertial sensor by: if it is determined that the object is not moving, forming a set of angular velocity of the object for the current value of the operating temperature of the inertial sensor, and the optimal number of angular velocity of the object in the above set is determined based on optimizing the ratio of the displacement of the average angular velocity of the object and the magnitude of the spread of the angular velocity of the object; averaging the values of the angular velocity of the object of the above set by the method of weighted average and calculating the correction factor for the current value of the operating temperature of the inertial sensor by solving the regression equation, the formation of a set of values of the angular velocity of the object and the calculation of the correction coefficient is performed for each value of the operating temperature of the inertial sensor. 25. The device according to claim 24, in which the optimal number of values of the angular velocity of the object in the above-mentioned set is determined by the method of intersection of the confidence intervals. 26. The device according to claim 22, in which the correction unit for the magnetic field calculates the correction factors for the magnetic field strength value for the correction of the angular velocity values of the object from the change in the magnetic field strength by: adjusting the magnetic field strengths using the calibration parameters of the magnetometer; if the corrected magnetic field strengths satisfy the range from the magnetic field strength distribution model, calculate the correction factors of the object’s angular velocity as a measure of the deviation of the current intensity vector from the model, if the corrected magnetic field strengths do not match the range from the magnetic field distribution model gauge parameters as orthogonalization problems and minimizing the displacement of the mean about the values of tension and cross-correlation between the axes, moreover, the model of the distribution of magnetic field strength and calibration parameters are calculated in advance. 27. The device according to claim 22, in which the block determining the spatial orientation determines the current spatial orientation of the object by computing the method of numerical integration. 28. The device according to claim 22 or 24, in which the motion detection unit determines whether the object is moving by comparing the value of the Euclidean norm of the angular velocity of the object with one constant and comparing the value of the Euclidean norm of the linear acceleration of the object with another constant, and these comparisons with constants are given in the first intellectual model of changing the movement of an object, it was determined that the object does not move if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object do not exceed the above constant, and it is determined that the object moves if the value of the Euclidean norm of the angular velocity of the object and the value of the Euclidean norm of the linear acceleration of the object exceed mentioned constant. 29. The device according to claim 22 or 24, in which the motion detection unit determines whether the object is moving, by comparing the value of the Euclidean norm of linear acceleration of an object with one adaptive threshold by the sliding window method and comparing the value of the Euclidean norm of the angular velocity of the object with another adaptive threshold by the sliding method windows, moreover, the method of the above comparison with adaptive thresholds and the method of calculating adaptive thresholds using the sliding window method are specified in the first intellectual model of the change in the motion of an object. 30. The device according to clause 29, in which the calculation of the adaptive threshold and the comparison with it of the Euclidean norm of the linear acceleration of the object is performed iteratively by: establishing an adaptive threshold, and it is pre-considered that the object does not move and the adaptive threshold is calculated in advance; receiving a predetermined threshold for determining motion and a predetermined threshold for determining the absence of motion; comparing the value of the norm of linear acceleration of an object with an adaptive threshold, and if the value of the norm of linear acceleration of an object does not exceed the adaptive threshold, then it is determined that the object does not move; if the value of the norm of the linear acceleration of an object exceeds the adaptive threshold, then the duration is determined, how long the norm of the norm of linear acceleration of the object exceeds the adaptive threshold; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold exceeds the threshold for determining movement, then it is determined that the object moves, while the adaptive threshold does not change; if the duration of finding the norm of the linear acceleration of an object of a more adaptive threshold does not exceed the threshold for determining movement, then it is determined that the object does not move; if it is determined that the object is not moving, recalculating the current adaptive threshold by calculating the average weighted values of the linear acceleration of the object in the sliding window, and then assigning the current value of the adaptive threshold to the maximum value between the current value of the adaptive threshold and the maximum value between the calculated average weighted value and the maximum value of the covariance matrices; if it is determined that the object is moving, comparing the value of the norm of the linear acceleration of the object with the adaptive threshold, and if the value of the norm of the linear acceleration of the object exceeds the adaptive threshold, then it is determined that the object continues to move; if the value of the norm of the linear acceleration of an object does not exceed the adaptive threshold, then the completion of the movement is determined by comparing the duration, how long the norm of the linear acceleration of the object is less than the threshold for determining no motion, and if the duration of finding a less adaptive threshold exceeds the threshold for determining no motion, that the object does not move, while the adaptive threshold is recalculated in accordance with the norm of the input values of the linear acceleration of the object. 31. The device according to claim 30, in which the adaptive threshold is calculated as the averaged over the set of various test data value of the norm of linear acceleration of objects of the same type at which movement is uniquely detected. 32. The device according to claim 22 or 24, in which the block motion detection determines whether the object is moving, using at least one pre-trained recurrent neural network as the first intelligent model for changing the movement of the object. 33. The device according to claim 32, in which the block motion detection determines whether the object is moving, using a cascade of pre-trained recurrent neural networks as the first intelligent model for changing the movement of the object. 34. The device according to claim 32 or 33, in which each recurrent neural network is a long short-term memory (LSTM). 35. The device according to claim 22, in which the block determining the spatial position determines the current spatial position of the object when determining the presence of movement of the object by: determining the vector of the object's motion by calculating the average value of the linear acceleration of the object in the global coordinate system using the sliding window method and linking this average value to the coordinate system in which the vertical axis is parallel to the gravitational acceleration vector; estimates of the velocity of the object based on the values of the linear acceleration of the object using the second intellectual model of the movement of the object; projecting the estimated velocity of the object onto a certain vector of the direction of motion of the object and obtaining the current spatial position of the object using a numerical integration method 36. The device according to p. 35, in which the determination of the vector of the direction of movement of the object is performed by the method of reference vectors (SVM). 37. The device according to claim 35, in which the determination of the direction vector of the movement of the object is performed using a pre-trained recurrent neural network. 38. Device according to any one of paragraphs. 35-37, in which the estimation of the linear velocity of the object based on the values of the linear acceleration of the object is performed using the polynomial velocity model as the second intelligent model of the motion of the object. 39. Device according to any one of paragraphs. 35-37, in which the estimation of the linear velocity of the object based on the values of the linear acceleration of the object is performed using a pre-trained recurrent neural network as the second intelligent model of the motion of the object. 40. Device according to any one of paragraphs. 35-37, in which the estimation of the linear velocity of the object based on the values of the linear acceleration of the object is performed using a combination of pre-trained models as the second intelligent model of the movement of the object. 41. The device according to claim 22, in which, if the values of the linear acceleration of the object, the values of the angular velocity of the object, the values of temperature and the values of the magnetic field strength are measured by more than one inertial sensor, then the spatial orientation determination unit independently determines the current spatial orientation of the object based on the values measured by each sensor, and calculates the resulting current spatial orientation of the object based on all the current spatial orientations of the object determined independently based on the received values from different inertial sensors, and the spatial position determination unit independently determines the current spatial position of the object based on the values measured by each sensor, and calculates the resulting current spatial position of the object based on all the current spatial positions of the object, determined independently based on the received values from different inertial sensors. 42. The device according to claim 41, in which the spatial orientation determination unit calculates the resulting current spatial orientation of the object by means of weighted averaging of the current spatial orientations of the object, determined independently based on the values measured by different inertial sensors; and the spatial position determination unit calculates the resulting current spatial position of the object through weighted averaging of the current spatial positions of the object, determined independently based on the received values measured by different inertial sensors.

Images