CN108734218B - Information fusion method and device of multi-sensor system - Google Patents

Information fusion method and device of multi-sensor system Download PDF

Info

Publication number
CN108734218B
CN108734218B CN201810497192.XA CN201810497192A CN108734218B CN 108734218 B CN108734218 B CN 108734218B CN 201810497192 A CN201810497192 A CN 201810497192A CN 108734218 B CN108734218 B CN 108734218B
Authority
CN
China
Prior art keywords
value
sensor
current moment
target sensor
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810497192.XA
Other languages
Chinese (zh)
Other versions
CN108734218A (en
Inventor
张亚男
金正具
印思琪
姜永强
孙宏轩
许梦兴
郑财
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BOE Technology Group Co Ltd
Beijing BOE Optoelectronics Technology Co Ltd
Original Assignee
BOE Technology Group Co Ltd
Beijing BOE Optoelectronics Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by BOE Technology Group Co Ltd, Beijing BOE Optoelectronics Technology Co Ltd filed Critical BOE Technology Group Co Ltd
Priority to CN201810497192.XA priority Critical patent/CN108734218B/en
Publication of CN108734218A publication Critical patent/CN108734218A/en
Application granted granted Critical
Publication of CN108734218B publication Critical patent/CN108734218B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/251—Fusion techniques of input or preprocessed data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The embodiment of the invention provides an information fusion method and device of a multi-sensor system, relates to the field of sensor information processing, and can greatly improve the adaptability of an information fusion algorithm and the algorithm precision under complex conditions. The method comprises the following steps: establishing an observation equation according to a preset random fault rate; determining a target sensor which can perform information fusion at the current moment in the multi-sensor system; calculating a theoretical value of an observed value of the target sensor at the current moment according to an observation equation; calculating the value of the observation residual statistical property of the target sensor at the current moment; judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value; and acquiring the result of information fusion of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong.

Description

Information fusion method and device of multi-sensor system
Technical Field
The invention relates to the field of sensor information processing, in particular to an information fusion method and device of a multi-sensor system.
Background
The complex systems such as an AR/VR (Augmented Reality/Virtual Reality) system, a medical health sensor intelligent system, a target tracking system, an image signal processing system and the like all belong to multi-sensor systems, and various multi-sensor systems need to adopt a specific algorithm to complete fusion processing of information acquired by each sensor. However, the existing multi-sensor information fusion method does not consider the situation that observation is random and unreliable, that is, observation of random faults, that is, the situation that the observed value of any sensor is likely to be wrong due to various random faults in the information fusion process, and meanwhile, the existing multi-sensor information fusion method does not consider the possibility that the system noise when the state value of the multi-sensor system is acquired and whether the observation noise of each sensor when the observed value is acquired have correlation; in the information fusion process of the actual multi-sensor system, random observation faults mostly exist, and meanwhile, certain correlation exists between the system noise of the actual multi-sensor system and the observation noise of each sensor, so that the introduced random observation faults are more consistent with the actual situation, and the existing multi-sensor information fusion method is not suitable for the actual situation.
Disclosure of Invention
The embodiment of the invention provides an information fusion method and device of a multi-sensor system, which can carry out optimal state estimation on the multi-sensor system more in a practical way, and greatly improve the adaptability of an information fusion algorithm and the algorithm precision under complex conditions.
In order to achieve the above purpose, the embodiment of the invention adopts the following technical scheme:
in a first aspect, an information fusion method and apparatus for a multi-sensor system are provided, including:
establishing an observation equation according to a preset random fault rate;
determining a target sensor which can perform information fusion at the current moment in the multi-sensor system;
calculating a theoretical value of an observed value of the target sensor at the current moment according to an observation equation;
calculating the value of the observation residual statistical property of the target sensor at the current moment;
judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value;
and acquiring the result of information fusion of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong.
Optionally, there is a correlation between system noise and observed noise in the multi-sensor system, which satisfies the following formula:
νi(k)＝βiw(k-1)+ηi(k)；
wherein, vi(k) For the observation noise, β, at time k for the ith sensor in a multi-sensor systemiA correlation parameter between the observed noise of the ith sensor in the multi-sensor system and the system noise of the multi-sensor system at the moment k, w (k-1) the system noise of the multi-sensor system at the moment k-1, etai(k) Random zero mean Gaussian white noise of the ith sensor at the moment k in the multi-sensor system; i is a positive integer.
Optionally, the state value of the multi-sensor system at any time contains system noise, and the observation value of any sensor in the multi-sensor system at any time contains observation noise.
Optionally, determining a target sensor capable of performing information fusion at the current time in the multi-sensor system includes:
and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
Optionally, the calculating, by the calculation formula, a value of the observation residual statistical characteristic of the target sensor at the current time includes:
and calculating the value of the observation residual statistical characteristic of the target sensor at the current moment according to the theoretical value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical characteristic calculation formula.
Optionally, the determining, according to a magnitude relationship between a value of an observation residual statistical characteristic of the target sensor at the current time and a preset chi-square distribution value, whether an error exists in an observation value of the target sensor at the current time includes:
when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the error exists in the observation value of the target sensor at the current moment;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
Optionally, obtaining a result of information fusion performed by the target sensor at the current time according to whether the observed value of the target sensor at the current time is incorrect includes:
acquiring a Kalman gain value of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong or not;
and calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current moment according to the Kalman gain value of the target sensor at the current moment.
Optionally, obtaining the kalman gain value of the target sensor at the current time according to whether the observation value of the target sensor at the current time is incorrect includes:
when the observation value of the target sensor at the current moment has an error, determining that the Kalman gain value of the target sensor at the current moment is zero;
and when the observed value of the target sensor at the current moment has no error, calculating the Kalman gain value of the target sensor at the current moment according to a Kalman gain calculation formula.
Optionally, the calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current time according to the kalman gain value of the target sensor at the current time includes:
according to the Kalman gain value of the target sensor at the current moment, sequentially calculating the update value of the state mean value predicted value of the target sensor according to the state mean value predicted value update formula by the target sensor in a sequence of sampling rate from large to small; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
sequentially calculating the update value of the target covariance of the target sensor according to the target sensor by the target sensor in the sequence of sampling rate from large to small according to the Kalman gain value of the target sensor at the current moment; the target covariance is an updated value of the target covariance of the target sensor whose sampling rate is the smallest among the target sensors.
In a second aspect, an information fusion apparatus of a multi-sensor system is provided, including:
the equation establishing module is used for establishing an observation equation according to a preset random fault rate;
the fusion judging module is used for determining a target sensor which can perform information fusion at the current moment in the multi-sensor system;
the observation calculation module is used for calculating a theoretical value of the observation value of the target sensor at the current moment determined by the fusion judgment module according to the observation equation established by the equation establishment module;
the residual statistical characteristic calculating module is used for calculating the value of the observation residual statistical characteristic of the target sensor at the current moment, which is determined by the fusion judging module;
the random fault judgment module is used for judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value calculated by the residual statistical characteristic calculation module;
and the fusion calculation module is used for acquiring the result of information fusion of the target sensor at the current moment according to the judgment result of the random fault judgment module.
Optionally, the apparatus further comprises a state calculation module;
the state calculation module is used for calculating the state value of the target sensor at the current moment before the observation calculation module calculates the theoretical value of the observed value of the target sensor at the current moment according to the observation equation established by the equation establishment module.
Optionally, the fusion judgment module is specifically configured to:
and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
Optionally, the residual statistical characteristic calculating module is specifically configured to:
and calculating the value of the observation residual statistical characteristic of the target sensor at the current moment according to the theoretical value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical characteristic calculation formula.
Optionally, the random fault determining module is specifically configured to:
when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the error exists in the observation value of the target sensor at the current moment;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
Optionally, the fusion calculation module includes a kalman gain calculation unit and a fusion processing unit;
the Kalman gain calculation unit is used for acquiring a Kalman gain value of the target sensor at the current moment according to the judgment result of the random fault judgment module;
and the fusion processing unit is used for calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current moment according to the Kalman gain value of the target sensor at the current moment calculated by the Kalman gain calculation unit.
Optionally, the kalman gain calculating unit is specifically configured to:
when the random fault judgment module determines that the observation value of the target sensor at the current moment has errors, determining that the Kalman gain value of the target sensor at the current moment is zero;
and when the random fault judgment module determines that the observed value of the target sensor at the current moment has no error, calculating the Kalman gain value of the target sensor at the current moment according to a Kalman gain calculation formula.
Optionally, the fusion processing unit includes a state mean estimate value operator unit and a target covariance calculation subunit;
the state mean value estimation value operator unit is used for sequentially calculating the update value of the state mean value prediction value of the target sensor according to the Kalman gain value of the target sensor at the current moment, which is calculated by the Kalman gain calculation unit, and the state mean value prediction value of the target sensor by sequentially updating the target sensor according to a state mean value prediction value updating formula in a descending order of sampling rate; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
the target covariance calculation subunit is used for calculating the update value of the target covariance of the target sensor according to the Kalman gain value of the target sensor at the current moment calculated by the Kalman gain calculation unit and the target sensor in turn according to a target covariance update formula in the sequence of sampling rate from large to small; the target covariance is an updated value of the target covariance of the target sensor whose sampling rate is the smallest among the target sensors.
In a third aspect, an embodiment of the present invention provides a computer device, including a memory, a processor; the memory has stored thereon a computer program operable on the processor, which when executed implements the method as provided by the first aspect.
In a fourth aspect, an embodiment of the present invention provides a computer storage medium storing a computer program, which when executed by a processor implements the method as provided in the first aspect.
The embodiment of the invention provides an information fusion method and device of a multi-sensor system, wherein the method comprises the following steps: establishing an observation equation according to a preset random fault rate; determining a target sensor which can perform information fusion at the current moment in the multi-sensor system; calculating the state value of the target sensor at the current moment according to the state equation and the dynamic moving average equation; calculating a theoretical value of an observed value of the target sensor at the current moment according to an observation equation; calculating the value of the observation residual statistical property of the target sensor at the current moment; judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value; and acquiring the result of information fusion of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong. According to the technical scheme provided by the embodiment of the invention, based on the conventional CKF algorithm, observation random faults are introduced into an observation equation, then a target sensor capable of being fused is judged, then the observation random faults are judged, and finally the result of information fusion of the multi-sensor system is obtained according to the judgment result. Because the technical scheme provided by the embodiment of the invention introduces the random observation fault when the observation equation is established at first, and then correspondingly judges the random fault to finally obtain the result after the information fusion of the multiple sensors, compared with the technical scheme provided by the embodiment of the invention in the prior art, the technical scheme provided by the embodiment of the invention is more suitable for the actual situation and has higher adaptability to the environment.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of an information fusion method of a multi-sensor system according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating an information fusion method for a multi-sensor system according to another embodiment of the present invention;
FIG. 3 is a schematic sampling diagram of a multi-sensor system according to an embodiment of the present invention;
fig. 4 is a schematic flow chart illustrating an information fusion method of a multi-sensor system according to another embodiment of the present invention;
fig. 5 is a schematic structural diagram of an information fusion device of a multi-sensor system according to an embodiment of the present invention;
fig. 6 is a comparison graph of the estimation effect of the four types of information fusion methods on the first-dimension position in the monte carlo simulation experiment provided by the embodiment of the present invention (with 10% of random faults observed, but only the fusion CKF method will detect and remove faults);
fig. 7 is a comparison graph of the estimation effect of the four types of information fusion methods on the second-dimensional position in the monte carlo simulation experiment provided by the embodiment of the present invention (with 10% of random faults observed, but only the fusion CKF method will detect and remove faults);
fig. 8 is a comparison graph of the estimation effect of the four types of information fusion methods on the third-dimensional position in the monte carlo simulation experiment provided by the embodiment of the present invention (with 10% of random faults observed, but only the fusion CKF method will detect and remove faults);
fig. 9 is a comparison graph of the estimation effect of the three types of information fusion methods on the first-dimension position in the monte carlo simulation experiment provided in the embodiment of the present invention (no random fault is observed);
fig. 10 is a comparison graph of the estimation effect of the three types of information fusion methods on the second-dimensional speed in the monte carlo simulation experiment provided in the embodiment of the present invention (no random fault is observed);
fig. 11 is a comparison graph of the estimation effect of the three types of information fusion methods on the third-dimensional acceleration in the monte carlo simulation experiment provided in the embodiment of the present invention (no random fault is observed);
fig. 12 is a comparison graph of the estimation effect of the three types of information fusion methods on the first-dimension position in the monte carlo simulation experiment provided in the embodiment of the present invention (with 10% of random faults observed, but with faults detected and removed);
fig. 13 is a comparison graph of the estimation effect of the three types of information fusion methods on the second-dimensional speed in the monte carlo simulation experiment provided in the embodiment of the present invention (with 10% of random faults observed, but faults will be detected and eliminated);
fig. 14 is a comparison graph of the estimation effect of the three types of information fusion methods on the third-dimensional acceleration in the monte carlo simulation experiment provided in the embodiment of the present invention (with 10% of random faults observed, but with faults detected and eliminated).
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that, in the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "e.g.," an embodiment of the present invention is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word "exemplary" or "such as" is intended to present concepts related in a concrete fashion.
It should be noted that, in the embodiments of the present invention, "of", "corresponding" and "corresponding" may be sometimes used in combination, and it should be noted that, when the difference is not emphasized, the intended meaning is consistent.
For the convenience of clearly describing the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first", "second", and the like are used for distinguishing the same items or similar items with basically the same functions and actions, and those skilled in the art can understand that the words "first", "second", and the like are not limited in number or execution order.
The existing multi-sensor information fusion method is low in degree of contact with actual conditions, and generally does not consider the condition that the observed value of a sensor does not exist due to the fact that random fault conditions appear in practice, so that the accuracy of the method is not enough for an actual multi-sensor system, and the method is not practical.
In view of the above problem, referring to fig. 1, an embodiment of the present invention provides an information fusion method for a multi-sensor system, including:
101. and establishing an observation equation according to a preset random fault rate.
In practice, the observation equation is established simultaneously with the state equation.
102. And determining a target sensor which can perform information fusion at the current moment in the multi-sensor system.
Illustratively, the step 102 is specifically: and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
103. And calculating a theoretical value of the observed value of the target sensor at the current moment according to an observation equation.
Specifically, because the observation random fault is introduced into the observation equation, the observation value calculated according to the equation is a theoretical value, and the observation value does not exist when the observation random fault occurs in practice and is different from the theoretical value, but an accurate observation value when the fault does not occur must be used when information is fused, so the theoretical value of the observation value is introduced, and a fault judgment process is carried out on the influence of the observation random fault in the subsequent method flow;
in addition, it should be noted that, in general, a theoretical value of an actual observed value is obtained by using a state value, so that, optionally, before the step 103, the method further includes: and calculating the state value of the target sensor at the current moment.
104. And calculating the value of the observation residual statistical property of the target sensor at the current moment.
Optionally, the step 104 includes: determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system;
wherein the sampling ratio refers to the ratio of the sampling rate of the sensor with the fastest sampling rate to the sum of the sampling rates of each sensor in the multi-sensor system; the value of the sampling ratio of each sensor is acquired before the step 101; illustratively, as shown in FIG. 3, the sampling ratio of sensor 1 is 1, the sampling ratio of sensor 2 is 3, and the sampling ratio of sensor 3 is 4.
For example, the time of each start of the multi-sensor system is set to 0, the time value of one second of operation is 1, and so on, and the time value of ks of operation is k.
105. And judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value.
Optionally, step 105 specifically includes: when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the error exists in the observation value of the target sensor at the current moment;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
106. And acquiring the result of information fusion of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong.
Illustratively, the step 106 specifically includes:
1061. and acquiring a Kalman gain value of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong or not.
Optionally, the 1061 step includes: when the observation value of the target sensor at the current moment has an error, determining that the Kalman gain value of the target sensor at the current moment is zero; and when the observed value of the target sensor at the current moment has no error, calculating the Kalman gain value of the target sensor at the current moment according to a Kalman gain calculation formula.
1062. And calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current moment according to the Kalman gain value of the target sensor at the current moment.
Optionally, the step 1062 includes: according to the Kalman gain value of the target sensor at the current moment, sequentially calculating the update value of the state mean value predicted value of the target sensor according to the state mean value predicted value update formula by the target sensor in a sequence of sampling rate from large to small; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
sequentially calculating the update value of the target covariance of the target sensor according to the target sensor by the target sensor in the sequence of sampling rate from large to small according to the Kalman gain value of the target sensor at the current moment; the target covariance is an updated value of the target covariance of the target sensor whose sampling rate is the smallest among the target sensors.
In practice, the information fusion of multiple sensors generally calculates a state mean value estimation value and a target covariance as final results, so the above 106 steps include steps 1061 and 1062; in addition, the parameter values required in the kalman gain calculation formula, the state mean prediction value updating formula and the target covariance updating formula may be obtained by using the existing CKF algorithm, and the like, which is not specifically limited in the present invention.
The embodiment of the invention provides an information fusion method of a multi-sensor system, which comprises the following steps: establishing an observation equation according to a preset random fault rate; determining a target sensor which can perform information fusion at the current moment in the multi-sensor system; calculating the state value of the target sensor at the current moment according to the state equation and the dynamic moving average equation; calculating a theoretical value of an observed value of the target sensor at the current moment according to an observation equation; calculating the value of the observation residual statistical property of the target sensor at the current moment; judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value; and acquiring the result of information fusion of the target sensor at the current moment according to the condition that whether the observed value of the target sensor at the current moment is wrong. According to the technical scheme provided by the embodiment of the invention, based on the conventional CKF algorithm, observation random faults are introduced into an observation equation, then a target sensor capable of being fused is judged, then the observation random faults are judged, and finally the result of information fusion of the multi-sensor system is obtained according to the judgment result. Because the technical scheme provided by the embodiment of the invention introduces the random observation fault when the observation equation is established at first, and then correspondingly judges the random fault to finally obtain the result after the information fusion of the multiple sensors, compared with the technical scheme provided by the embodiment of the invention in the prior art, the technical scheme provided by the embodiment of the invention is more suitable for the actual situation and has higher adaptability to the environment.
In order to more clearly illustrate a specific flow of the multi-sensor information fusion method provided by the embodiment of the present invention, referring to fig. 2, an embodiment of the present invention provides another information fusion method of a multi-sensor system, including:
201. and establishing a state equation of the multi-sensor system and establishing an observation equation according to a preset random fault rate.
Specifically, the state equation is:
x(k+1)＝A(k)x(k)+w(k)；
the observation equation is:
yi(ki)＝γi(ki)[Ci(ki)xi(ki)]+νi(ki)；
x(k)∈Rnis the state value of the multi-sensor system at the time k, x (k +1) is the state value of the multi-sensor system at the time k +1, A (k) epsilon Rn*nA system state transition matrix of the multi-sensor system at the moment k, and w (k) system noise of the multi-sensor system at the moment k; wherein n denotes the dimension of the multi-sensor system, for example, if the parameters detected by a certain multi-sensor system have speed, acceleration and displacement, the dimension of the multi-sensor system is 3;
yi(ki)∈Rmfor the theoretic value of the observed value of the ith sensor at the kth time to which the ith sensor belongs in the multi-sensor system, because it is assumed in the present embodiment that the observed random fault is data loss, γ used for expressing the observed random fault in the observation equationi(ki) The epsilon R is a random sequence (which can be changed according to the type of observed random faults) with the value of 0 or 1 and satisfying Bernoulli distribution, and is mainly used for expressing the observed fault rate, gamma, of the multi-sensor system in an observation equationi(ki) When the observation value is 0, the observation shows that gamma is generated when the observation has faultsi(ki) When the observation value is 1, the observation does not have a fault, and the proportion of 0 in all the values is the observation fault rate; ci(ki)∈Rm*nIs an observation matrix, x, of the ith sensor in the multi-sensor system at the kth moment to which the ith sensor belongsi(ki) For the state value v of the ith sensor in the multi-sensor system at the kth time to which the ith sensor belongsi(ki) Observing noise at the kth moment of the ith sensor in the multi-sensor system; m refers to the dimension of the sensor, and m is less than or equal to n;
since there is a certain correlation between the system noise and the observation noise in the multi-sensor system in the actual multi-sensor system, in the present embodiment:
νi(k)＝βiw(k-1)+ηi(k)；
νi(k) for the observation noise, β, at time k for the ith sensor in a multi-sensor systemiObserving noise and multisensory for the ith sensor in a multisensor systemThe correlation parameter of the system noise of the multi-sensor system at the moment k, w (k-1) is the system noise of the multi-sensor system at the moment k-1, etai(k) Random zero mean Gaussian white noise of the ith sensor at the moment k in the multi-sensor system; i is a positive integer; the correlation parameters are specifically determined according to the actual condition or determined by experiential personnel and historical data during simulation experiments;
optionally, the state value of the multi-sensor system at any time contains system noise, and the observation value of any sensor in the multi-sensor system at any time contains observation noise.
For example, the time of each start of the multi-sensor system is set to 0, the time value of one second of operation is 1, and so on, and the time value of ks of operation is k.
202. And determining a target sensor which can perform information fusion at the current moment in the multi-sensor system.
Referring to fig. 3, a multi-sensor system with three different sensors is taken as an example, in the figure, vertical dotted straight lines with the same height belong to the same sensor; because the sampling rates of the three sensors are different, the information of the fixed sensors cannot be fused at each moment, so that the information of which sensors at the current moment can be fused and interpreted at the moment, and various judgment modes exist in practice;
the remainder function is:
Figure GDA0002689421910000121
Figure GDA0002689421910000122
the sampling ratio of the p-th target sensor in the target sensors is shown, k is a time value, and p is a positive integer; since the fastest sampling rate sensor must be engaged each timeIf the sampling ratio of the sensor can be divided by the time value of the current time, the sensor and the sensor with the highest rate transmission rate are sampled at the current time, and information fusion can be performed.
Wherein the sampling ratio refers to the ratio of the sampling rate of the sensor with the fastest sampling rate to the sum of the sampling rates of each sensor in the multi-sensor system; the value of the sampling ratio of each sensor is acquired before the step 101; illustratively, as shown in FIG. 3, the sampling ratio of sensor 1 is 1, the sampling ratio of sensor 2 is 3, and the sampling ratio of sensor 3 is 4.
203. And calculating the state value of the target sensor at the current moment according to the state equation and the dynamic moving average equation.
Specifically, each sensor in the multi-sensor system is asynchronous and multi-rate, and when multi-sensor information fusion is performed, because information fusion data only has one transfer once, and the transfer is reflected on data updating of the sensor with the fastest transmission rate, only the state value of the sensor with the fastest transmission rate is calculated by using a state equation in the information fusion process at each moment, and the rest sensors are calculated by using a dynamic sliding average equation; so, referring to fig. 4, step 203 comprises:
2031. and calculating the state value of the first target sensor at the current moment according to the state value of the first target sensor at the previous moment and the state equation.
The first target sensor is the target sensor with the fastest sampling rate in the target sensors.
2032. And calculating the state values of the target sensors except the first target sensor at the current moment according to the state value of the first target sensor at the current moment and the state transition matrix in the state equation and the dynamic moving average equation.
The dynamic moving average equation is:
Figure GDA0002689421910000123
wherein n isiIs the sampling ratio, x, of the ith sensor in a multi-sensor system1(niki) For the first target sensor at the nth sensor of the ith sensorikiThe state value at each moment.
204. And calculating a theoretical value of the observed value of the target sensor at the current moment according to the state value of the target sensor at the current moment and the observation equation.
Specifically, because the observation random fault is introduced into the observation equation, the observation value calculated according to the equation is a theoretical value, and in practice, the observation value does not exist when the observation random fault occurs and is different from the theoretical value, but an accurate observation value when the fault does not occur must be used during information fusion, so the theoretical value of the observation value is introduced, and a fault judgment process is also performed on the influence of the observation random fault in the subsequent method flow.
205. And according to the state mean value estimation value and the target covariance of the target sensor at the previous moment, calculating a state mean value predicted value, an observation variance predicted value and a target covariance predicted value of the target sensor at the current moment according to a cubature Kalman filtering CKF algorithm.
Wherein, the state mean value estimated value is the estimated value of the mean value of the state values; the target covariance is the covariance of the mean value of the state values and the observed value; the state mean value predicted value is the predicted value of the mean value of the state values; observing the predicted value as the predicted value of the observed value; the observation variance prediction value is a prediction value of the variance of the observation value; the target covariance predicted value is the predicted value of the target covariance;
it should be noted that, when calculating the state mean predicted value, the observation variance predicted value, and the target covariance predicted value of the target sensor at the first time, the state mean estimated value and the target covariance of the target sensor at the previous time are all preset values, and are obtained before step 101.
206. And calculating the value of the observation residual statistical characteristic of the target sensor at the current moment according to an observation residual statistical characteristic calculation formula.
Since observation random faults are introduced in the multi-sensor information fusion process in the embodiment of the present invention, fault judgment needs to be performed to determine the kalman gain before each sensor is subjected to subsequent update of the state mean value estimation value and the target covariance, and a method of calculating a residual statistical property is specifically adopted in the embodiment of the present invention to judge whether there is an observation random fault, so that, optionally, step 106 in the embodiment of the present invention specifically is:
calculating the value of the observation residual statistical property of the target sensor at the current moment according to the theoretical value and the observation predicted value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical property calculation formula;
the observation residual statistic characteristic calculation formula is as follows:
Figure GDA0002689421910000131
Figure GDA0002689421910000132
wherein the content of the first and second substances,
Figure GDA0002689421910000141
for the value of the residual statistical property at time k for the p-th target sensor,
Figure GDA0002689421910000142
is a theoretical value of the observed value of the p-th target sensor at the time k,
Figure GDA0002689421910000143
for the observed predicted value of the p-th target sensor at the moment k,
Figure GDA0002689421910000144
covariance of p-th target sensor at time k, wlIs the weight of the volume points, l is the number of the volume points,
Figure GDA0002689421910000145
is the theoretical value of the observed value of the ith volume point of the pth target sensor at the time k,
Figure GDA0002689421910000146
the noise variance is observed at time k for the p-th target sensor.
207. And judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value.
Exemplary, possible errors in the observed value include, but are not limited to, any of: the observation value caused by data loss does not exist, the actual observation value and the theoretical value of the observation value caused by sensor failure are different, and the actual observation value and the theoretical value of the observation value are different because the observation process in the sensor observation is incorrect.
Optionally, referring to fig. 4, the step 207 includes:
2071. and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the error exists in the observation value of the target sensor at the current moment.
In particular, values of observed residual gage characteristics of the target sensor
Figure GDA0002689421910000147
Obeying chi-square distribution with the degree of freedom m, wherein m is the observation dimension of the target sensor, so that a preset chi-square distribution value can be determined from a chi-square distribution table after the confidence degree is preset
Figure GDA0002689421910000148
When in use
Figure GDA0002689421910000149
It indicates that the observation is faulty, and the observation value of the target sensor at the current moment is wrong.
2072. And when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
When in use
Figure GDA00026894219100001410
It indicates that the observation is not faulty, and the observation value of the target sensor at the current moment is not wrong.
208. And acquiring a Kalman gain value of the target sensor at the current moment according to the condition that whether the observed value of the target sensor is wrong or not.
Optionally, referring to fig. 4, step 208 includes:
2081. and when the observed value of the target sensor at the current moment is wrong, determining that the Kalman gain value of the target sensor at the current moment is zero.
2082. When the observed value of the target sensor at the current moment has no error, calculating a Kalman gain value of the target sensor at the current moment according to an observation variance predicted value and a target covariance predicted value of the target sensor at the current moment and a Kalman gain calculation formula;
the Kalman gain calculation formula is as follows:
Figure GDA0002689421910000151
Figure GDA0002689421910000152
for the kalman gain at time k for the p-th target sensor,
Figure GDA0002689421910000153
for the target covariance prediction value of the p-th target sensor at time k,
Figure GDA0002689421910000154
and predicting the observed variance of the p-th target sensor at the k moment.
Illustratively, the two steps 2081 and 2082 may be specifically expressed by the following equations:
Figure GDA0002689421910000155
209. and calculating a state mean value estimation value and a target covariance after information fusion of the target sensor at the current moment according to a state mean value prediction value, a Kalman gain value, a theoretical value of an observation value, an observation prediction value, a state mean value prediction value, a target covariance prediction value and an observation variance prediction value of the target sensor at the current moment and a state mean value estimation value and a target covariance after information fusion of the target sensor at the current moment.
Optionally, referring to fig. 4, step 209 includes:
2091. according to the state mean value predicted value, the Kalman gain value, the theoretical value of the observed value and the observation predicted value of the target sensor at the current moment, sequentially calculating the update value of the state mean value predicted value of the target sensor according to the state mean value predicted value update formula of the target sensor in the order from large to small of the sampling rate; and taking the updated value of the state mean value predicted value of the target sensor with the minimum sampling rate in the target sensors as the state mean value estimated value after information fusion of the target sensors.
Illustratively, the state mean prediction value update formula is:
Figure GDA0002689421910000156
wherein the content of the first and second substances,
Figure GDA0002689421910000157
the updated value of the state mean value predicted value of the p-th target sensor at the time k,
Figure GDA0002689421910000158
for the kalman gain value of the p-th target sensor at time k,
Figure GDA0002689421910000159
and predicting the state mean value of the p-th target sensor at the k moment.
2092. According to a target covariance predicted value, a Kalman gain value and an observation variance predicted value of a target sensor at the current moment, sequentially calculating an update value of the target covariance of the target sensor according to a target covariance update formula by the target sensor in a sequence of sampling rates from high to low; and the update value of the target covariance of the target sensor with the minimum sampling rate in the target sensors is the target covariance after information fusion of the target sensors.
Illustratively, the target covariance update formula is:
Figure GDA0002689421910000161
wherein the content of the first and second substances,
Figure GDA0002689421910000162
for the updated value of the target covariance for the p-th target sensor at time k,
Figure GDA0002689421910000163
for the target covariance prediction value of the p-th target sensor at time k,
Figure GDA0002689421910000164
and predicting the observed variance of the p-th target sensor at the k moment.
It should be noted that, the parameters in the two formulas of the state mean prediction value updating formula and the target covariance updating formula need to be obtained by means of the CKF algorithm, so that the following iterative relationship exists: the state mean value estimated value of the p-th target sensor in the target sensors at the previous moment is an updated value of the state mean value predicted value of the p-1-th target sensor in the target sensors at the current moment;
the target covariance of the p-th target sensor in the target sensors at the previous moment is the updated value of the target covariance of the p-1-th target sensor in the target sensors at the current moment;
the sampling rate of the p-th target sensor in the target sensors is only greater than that of the p-1 target sensors in the target sensors; the sampling rate of the p-1 th target sensor in the target sensors is only smaller than that of the p-th target sensor in the target sensors; p is a positive integer.
According to the method and the process, the state mean value estimation value and the target covariance after information fusion of the multi-sensor system at any moment can be obtained, and the information fusion estimation of the multi-sensor in the multi-sensor system is completed.
The information fusion method of the multi-sensor system provided by the above specific embodiment includes: establishing a state equation of the multi-sensor system and establishing an observation equation according to a preset random fault rate; determining a target sensor which can perform information fusion at the current moment in the multi-sensor system; calculating the state value of the target sensor at the current moment according to the state equation and the dynamic moving average equation; calculating a theoretical value of an observed value of the target sensor at the current moment according to the state value of the target sensor at the current moment and an observation equation; according to the state mean value estimation value and the target covariance of the target sensor at the previous moment, calculating a state mean value predicted value, an observation variance predicted value and a target covariance predicted value of the target sensor at the current moment according to a cubature Kalman filtering CKF algorithm; the target covariance is the covariance of the state mean and the observed value; calculating the value of the observation residual statistical characteristic of the target sensor according to an observation residual statistical characteristic calculation formula; judging whether random faults exist in the observation of the target sensor according to the size relation between the observation residual statistical characteristic value of the target sensor and the preset chi-square distribution value; acquiring a Kalman gain value of a target sensor at the current moment according to the condition that whether random faults occur in observation of the target sensor; and calculating a state mean value estimation value and a target covariance after information fusion of the target sensor at the current moment according to a state mean value prediction value, a Kalman gain value, an observation prediction value, a state mean value prediction value, a target covariance prediction value and an observation variance prediction value of the target sensor at the current moment and a state mean value estimation value and a target covariance after information fusion of the target sensor at the current moment. It can be seen that, in the technical scheme provided by the above specific embodiment, based on the existing CKF algorithm, an observation random fault is introduced into an observation equation, then a target sensor capable of fusion is determined, then the observation random fault is determined to determine a kalman gain value, then a sequential fusion algorithm applied to information fusion of asynchronous multi-rate sensors is used to fuse the target sensors in order from large to small sampling rates, and finally a state mean value estimation value and a target covariance after fusion are calculated. Because the technical scheme introduces random observation faults when an observation equation is established at first, then correspondingly judges the random faults, and finally calculates the result after the multi-sensor information fusion by using the sequential fusion algorithm, the technical scheme provided by the embodiment of the invention has stronger environmental adaptability, is more suitable for the actual situation and has higher precision compared with the technical scheme provided by the prior art.
In order to more vividly illustrate the superiority of the information fusion algorithm provided by the above embodiment compared with the existing information fusion algorithm, the following description is made with the actual monte-carlo simulation experiment result:
in order to evaluate the estimation effect of the state mean value, a root mean square error RMSE and a root mean square error TARMSE of an average time are introduced, the two parameters are indexes for measuring the effect of the information fusion algorithm on the state mean value estimation value, the smaller the value is, the better the estimation effect is,
Figure GDA0002689421910000171
Figure GDA0002689421910000172
wherein T is the length of the simulation signal of the Monte Carlo simulation, M is the number of Monte Carlo simulations, x (k) is the state value of the multi-sensor system at time k,
Figure GDA0002689421910000173
is the first*And (3) state mean value estimated values obtained at the time k of the sub-Monte Carlo simulation.
In a simulation experiment, an UKF (Unscented Kalman Filter) method, a CKF (Cubage Kalman Filter) method and a fused UKF method are used for index comparison with the method (the fused CKF method) provided by the embodiment of the invention, wherein the length of a simulation signal is set to be 600, the state dimension is three-dimensional, the first dimension is position, the second dimension is speed, the third dimension is acceleration, and the observation tracks the three-dimensional state, namely the observation is also three-dimensional;
firstly, simulation experiments are carried out by using an original UKF method, a CKF method, a fusion UKF method and the fusion CKF method provided by the embodiment of the invention, 10% of observed random faults are introduced in the four methods, wherein the UKF method, the CKF method and the fusion UKF method do not eliminate the observed random faults, and the experiments can obtain the results of table 1, table 2 and fig. 6-8:
TABLE 1
Figure GDA0002689421910000181
TABLE 2
Figure GDA0002689421910000182
As can be seen from table 1, the last estimation effect obtained by the fused CKF method provided in the embodiment of the present invention is significantly better than that of the UKF method, the CKF method, and the fused UKF method in the presence of an observed random fault, and as can be seen from table 2, compared with the fused UKF method provided in the embodiment of the present invention, the method provided in the embodiment of the present invention has a shorter operation time and higher efficiency as compared with the method provided in the embodiment of the present invention;
as can be seen by referring to the RMSE index curves of the simulation experiments shown in fig. 6 to 8, the index curve of the integrated CKF method with fault detection and elimination is more stable, which indicates that compared with the conventional UKF method, CKF method and integrated UKF method, the method provided by the embodiment of the present invention can eliminate the influence of observing random faults and improve the system stability, and is more suitable for actual complex environments.
In addition, in order to further show the advantages of the technical solutions provided by the embodiments of the present invention, the method for detecting and eliminating faults in the embodiments of the present invention is introduced into the existing UKF method, CKF method and fusion UKF method, and the experimental results shown in table 3, table 4, table 5, table 6 and fig. 9-fig. 14 below can be obtained;
as can be seen by referring to tables 3 and 4 below, the estimation accuracy of the state mean value of the information fusion method provided by the embodiment of the present invention is significantly better than that of the existing UKF method and CKF method no matter whether observation random faults are set or not, and meanwhile, as can be seen by referring to the RMSE index curve of the simulation experiment of fig. 9-14, in the case of three dimensions, the RMSE curve of the information fusion algorithm provided by the embodiment of the present invention is also the lowest compared with the existing UKF method and CKF method, i.e., the estimation effect of the state mean value is optimal, and the estimation effect of the state mean value is equivalent compared with the existing fusion UKF method;
TABLE 3
Figure GDA0002689421910000191
TABLE 4
Figure GDA0002689421910000192
As can be seen from tables 5 and 6, the time required for the result of the simulation experiment of the information fusion method provided by the embodiment of the present invention to be significantly shorter than that of the existing fusion UKF method, no matter whether a random fault is observed or not;
TABLE 5
Figure GDA0002689421910000193
The UKF method and the CKF method are single-sensor algorithms, so the running time is shorter than that of the fused UKF and the fused CKF for a multi-sensor system.
TABLE 6
Figure GDA0002689421910000194
Therefore, in general, the information fusion method of the multi-sensor system provided by the embodiment of the invention has better effect than the existing method.
Referring to fig. 5, an embodiment of the present invention further provides an information fusion apparatus 01 of a multi-sensor system, including:
an equation establishing module 501, configured to establish an observation equation according to a preset random failure rate;
a fusion judgment module 502, configured to determine a target sensor in the multi-sensor system that can perform information fusion at the current time;
the observation calculation module 503 is configured to calculate a theoretical value of the observation value of the target sensor at the current time, which is determined by the fusion determination module 502, according to the observation equation established by the equation establishment module 501;
a residual statistical characteristic calculating module 504, configured to calculate a value of an observation residual statistical characteristic of the target sensor determined by the fusion determining module 502 at the current time;
a random fault determination module 505, configured to determine whether an observed value of the target sensor at the current time is incorrect according to a magnitude relationship between a value of an observed residual statistical characteristic of the target sensor at the current time, which is calculated by the residual statistical characteristic calculation module 504, and a preset chi-square distribution value;
and a fusion calculation module 506, configured to obtain a result of information fusion performed by the target sensor at the current time according to the judgment result of the random fault judgment module 505.
Optionally, because the observed value and theoretical value of the sensor are actually calculated and the state value of the sensor is needed, the apparatus further includes a state calculating module 507;
the state calculating module 507 is configured to calculate a state value of the target sensor at the current time before the observation calculating module 503 calculates a theoretical value of the observed value of the target sensor at the current time according to the observation equation established by the equation establishing module 501.
Optionally, the fusion determining module 502 is specifically configured to:
and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
Optionally, the residual statistical characteristic calculating module 504 is specifically configured to:
and calculating the value of the observation residual statistical characteristic of the target sensor at the current moment according to the theoretical value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical characteristic calculation formula.
Optionally, the random fault determining module 505 is specifically configured to:
when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the error exists in the observation value of the target sensor at the current moment;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
Optionally, the fusion calculation module 506 includes a kalman gain calculation unit 5061 and a fusion processing unit 5062;
the kalman gain calculation unit 5061 is configured to obtain a kalman gain value of the target sensor at the current time according to the determination result of the random fault determination module 505;
the fusion processing unit 5062 is configured to calculate a state mean value estimation value and a target covariance of the target sensor after information fusion at the current time according to the kalman gain value of the target sensor at the current time calculated by the kalman gain calculation unit.
Optionally, the kalman gain calculation unit 5061 is specifically configured to:
when the random fault judgment module 505 determines that the observation value of the target sensor at the current moment has an error, determining that the kalman gain value of the target sensor at the current moment is zero;
when the random fault judgment module 505 determines that the observation value of the target sensor at the current moment has no error, the kalman gain value of the target sensor at the current moment is calculated according to the kalman gain calculation formula.
Optionally, the fusion processing unit 5062 includes a state mean estimate operator unit 50621 and a target covariance calculation subunit 50622;
the state mean value estimation value operator unit 50621 is used for sequentially calculating the update value of the state mean value prediction value of the target sensor according to the Kalman gain value of the target sensor at the current moment, which is calculated by the Kalman gain calculation unit, and the state mean value prediction value update formula of the target sensor sequentially according to the sampling rate from high to low; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
the target covariance calculation subunit 50622 is configured to calculate, according to the kalman gain value of the target sensor at the current time, which is calculated by the kalman gain calculation unit 5061, update values of the target covariance of the target sensor by sequentially calculating the target sensor according to a target covariance update formula in an order from a large sampling rate to a small sampling rate; the target covariance is an updated value of the target covariance of the target sensor whose sampling rate is the smallest among the target sensors.
It should be noted that, the information fusion apparatus of a multi-sensor system provided in the embodiment of the present invention may further include an obtaining module for obtaining parameters of the multi-sensor system during operation, such as a sampling ratio of a sensor, a sampling rate of the sensor, and the like, and may also directly obtain parameters required by the module, where the obtaining module is not limited specifically here.
The information fusion device of the multi-sensor system provided by the embodiment of the invention comprises: the equation establishing module is used for establishing an observation equation according to a preset random fault rate; the fusion judging module is used for determining a target sensor which can perform information fusion at the current moment in the multi-sensor system; the observation calculation module is used for calculating a theoretical value of an observation value of the target sensor at the current moment according to the observation equation established by the equation establishment module; the residual statistical characteristic calculation module is used for calculating the value of the observation residual statistical characteristic of the target sensor at the current moment, which is determined by the fusion judgment module, according to an observation residual statistical characteristic calculation formula; the random fault judgment module is used for judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value calculated by the residual statistical characteristic calculation module; and the fusion calculation module is used for acquiring the result of information fusion of the target sensor at the current moment according to the judgment result of the random fault judgment module. Therefore, the technical scheme provided by the embodiment of the invention can introduce observation random faults into an observation equation on the basis of the conventional CKF algorithm, then judge the target sensor capable of being fused, then judge the observation random faults, and finally obtain the information fusion result of the multi-sensor system according to the judgment result. Because the technical scheme provided by the embodiment of the invention introduces the random observation fault when the observation equation is established at first, and then correspondingly judges the random fault to finally obtain the result after the information fusion of the multiple sensors, compared with the technical scheme provided by the embodiment of the invention in the prior art, the technical scheme provided by the embodiment of the invention is more suitable for the actual situation and has higher adaptability to the environment.
The embodiment of the application provides computer equipment which comprises a memory and a processor. The memory is stored with a computer program that can be run on a processor, and the processor implements the information fusion method of the multi-sensor system when executing the computer program. Wherein the storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
An embodiment of the present application provides a computer-readable medium, which stores a computer program, and the computer program, when executed by a processor, implements the information fusion method of the multi-sensor system.
The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (14)

1. An information fusion method of a multi-sensor system is characterized by comprising the following steps:
establishing an observation equation according to a preset random fault rate;
determining a target sensor which can perform information fusion at the current moment in the multi-sensor system;
calculating a theoretical value of an observed value of the target sensor at the current moment according to the observation equation;
calculating the value of the observation residual statistical characteristic of the target sensor at the current moment;
judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual error statistical characteristic value of the target sensor at the current moment and a preset chi-square distribution value;
acquiring a result of information fusion of the target sensor at the current moment according to the condition whether the observed value of the target sensor at the current moment is wrong or not;
the obtaining of the result of information fusion of the target sensor at the current moment according to whether the observation value of the target sensor at the current moment is wrong includes:
acquiring a Kalman gain value of the target sensor at the current moment according to the condition whether the observed value of the target sensor at the current moment is wrong or not; calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current moment according to a Kalman gain value of the target sensor at the current moment;
the calculating of the state mean value estimation value and the target covariance after information fusion of the target sensor at the current moment according to the Kalman gain value of the target sensor at the current moment comprises:
according to the Kalman gain value of the target sensor at the current moment, sequentially calculating the update value of the state mean value predicted value of the target sensor according to a state mean value predicted value update formula by the target sensor in a sequence of sampling rate from large to small; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
calculating the target covariance update value of the target sensor by the target sensor according to the Kalman gain value of the target sensor at the current moment in a sampling rate sequence from large to small; the target covariance is an updated value of a target covariance of a target sensor of which a sampling rate is smallest among the target sensors.
2. The method of claim 1, wherein there is a correlation between system noise and observed noise in the multi-sensor system that satisfies the following equation:
νi(k)＝βiw(k-1)+ηi(k)；
wherein, vi(k) For the observed noise, β, at time k for the ith sensor in the multi-sensor systemiA correlation parameter between the observed noise of the ith sensor in the multi-sensor system and the system noise of the multi-sensor system at the moment k, w (k-1) the system noise of the multi-sensor system at the moment k-1, etai(k) Random zero mean Gaussian white noise of the ith sensor at the moment k in the multi-sensor system; i is a positive integer.
3. The method of claim 2, wherein the state values of the multi-sensor system at any one time comprise the system noise, and wherein the observations of any one of the sensors of the multi-sensor system at any one time comprise observation noise.
4. The method of claim 1, wherein determining the target sensors in the multi-sensor system that can perform information fusion at the current time comprises:
and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
5. The method of claim 1, wherein the calculating the value of the observed residual statistical property of the target sensor at the current time comprises:
and calculating the value of the observation residual statistical property of the target sensor at the current moment according to the theoretical value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical property calculation formula.
6. The method of claim 1, wherein the determining whether the observed value of the target sensor at the current moment has an error according to a magnitude relationship between the value of the observed residual statistical characteristic of the target sensor at the current moment and a preset chi-square distribution value comprises:
when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the observation value of the target sensor at the current moment has errors;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
7. The method of claim 1, wherein the obtaining the kalman gain value of the target sensor at the current time according to whether the observed value of the target sensor at the current time is incorrect comprises:
when the observation value of the target sensor at the current moment has an error, determining that the Kalman gain value of the target sensor at the current moment is zero;
and when the observed value of the target sensor at the current moment has no error, calculating the Kalman gain value of the target sensor at the current moment according to a Kalman gain calculation formula.
8. An information fusion apparatus of a multi-sensor system, comprising: the system comprises an equation establishing module, a fusion judging module, an observation calculating module, an observation residual statistical characteristic calculating module, a random fault judging module and a fusion calculating module;
the equation establishing module is used for establishing an observation equation according to a preset random fault rate;
the fusion judging module is used for determining a target sensor which can perform information fusion at the current moment in the multi-sensor system;
the observation calculation module is used for calculating a theoretical value of the observation value of the target sensor at the current moment determined by the fusion judgment module according to the observation equation established by the equation establishment module;
the residual statistical characteristic calculating module is used for calculating the value of the observation residual statistical characteristic of the target sensor at the current moment, which is determined by the fusion judging module;
the random fault judgment module is used for judging whether the observed value of the target sensor at the current moment has errors or not according to the magnitude relation between the observed residual statistical characteristic value of the target sensor at the current moment and the preset chi-square distribution value calculated by the residual statistical characteristic calculation module;
the fusion calculation module is used for acquiring the result of information fusion of the target sensor at the current moment according to the judgment result of the random fault judgment module;
the fusion calculation module comprises a Kalman gain calculation unit and a fusion processing unit;
the Kalman gain calculation unit is used for acquiring a Kalman gain value of the target sensor at the current moment according to a judgment result of the random fault judgment module;
the fusion processing unit is used for calculating a state mean value estimation value and a target covariance of the target sensor after information fusion at the current moment according to the Kalman gain value of the target sensor at the current moment calculated by the Kalman gain calculation unit;
the fusion processing unit comprises a state mean value estimation value operator unit and a target covariance calculation subunit;
the state mean value estimated value operator unit is used for calculating the updated value of the state mean value predicted value of the target sensor according to the Kalman gain value of the target sensor at the current moment, which is calculated by the Kalman gain calculation unit, and the target sensor is sequentially updated according to a state mean value predicted value updating formula in a sequence of sampling rate from large to small; the state mean value estimation value is an updated value of a state mean value predicted value of a target sensor with the minimum sampling rate in the target sensors;
the target covariance calculation subunit is configured to calculate, according to the kalman gain value of the target sensor at the current time, which is calculated by the kalman gain calculation unit, update values of the target covariance of the target sensor sequentially according to a target covariance update formula by the target sensor in a descending order of a sampling rate; the target covariance is an updated value of a target covariance of a target sensor of which a sampling rate is smallest among the target sensors.
9. The apparatus according to claim 8, wherein the fusion determination module is specifically configured to:
and determining a target sensor capable of carrying out information fusion at the current moment according to the moment value of the current moment and the sampling ratio of each sensor in the multi-sensor system.
10. The apparatus of claim 8, wherein the residual statistical property calculation module is specifically configured to:
and calculating the value of the observation residual statistical property of the target sensor at the current moment according to the theoretical value of the observation value of the target sensor at the current moment and the sampling ratio of the target sensor and an observation residual statistical property calculation formula.
11. The apparatus according to claim 8, wherein the random failure determination module is specifically configured to:
when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is larger than the preset chi-square distribution value, determining that the observation value of the target sensor at the current moment has errors;
and when the absolute value of the observation residual statistical characteristic value of the target sensor at the current moment is less than or equal to the preset chi-square distribution value, determining that no error exists in the observation value of the target sensor at the current moment.
12. The apparatus of claim 8, wherein the kalman gain calculation unit is specifically configured to:
when the random fault judgment module determines that the observation value of the target sensor at the current moment has an error, determining that the Kalman gain value of the target sensor at the current moment is zero;
and when the random fault judgment module determines that the observed value of the target sensor at the current moment has no error, calculating the Kalman gain value of the target sensor at the current moment according to a Kalman gain calculation formula.
13. A computer device comprising a memory, a processor; the memory has stored thereon a computer program operable on the processor, which when executed implements the method of any of claims 1-7.
14. A computer-readable medium, in which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
CN201810497192.XA 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system Active CN108734218B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810497192.XA CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810497192.XA CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Publications (2)

Publication Number Publication Date
CN108734218A CN108734218A (en) 2018-11-02
CN108734218B true CN108734218B (en) 2021-01-15

Family

ID=63934951

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810497192.XA Active CN108734218B (en) 2018-05-22 2018-05-22 Information fusion method and device of multi-sensor system

Country Status (1)

Country Link
CN (1) CN108734218B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109848988B (en) * 2019-01-24 2022-12-06 深圳市普森斯科技有限公司 Scanning matching method and system based on historical multi-frame point cloud information fusion
CN109978055B (en) * 2019-03-26 2021-04-23 京东方科技集团股份有限公司 Information fusion method and system for multi-sensor system, computer device and medium
CN111062359B (en) * 2019-12-27 2023-03-31 广东海洋大学深圳研究院 Two-stage Kalman filtering fusion method based on noise sequential decorrelation
CN112134544B (en) * 2020-09-25 2024-03-19 郑州佛光发电设备股份有限公司 Filtering method, device, equipment and computer readable storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101630445A (en) * 2009-06-10 2010-01-20 民航数据通信有限责任公司 Method for processing data of asynchronous multiple sensors based on noise error and system thereof
CN101696906A (en) * 2009-10-28 2010-04-21 上海电机学院 Multi-source information fusion based state monitoring system and method of wind-driven generator
US20130246006A1 (en) * 2012-03-13 2013-09-19 King Fahd University Of Petroleum And Minerals Method for kalman filter state estimation in bilinear systems
CN103575278A (en) * 2013-10-08 2014-02-12 北京理工大学 Robust filtering method aiming at parameter uncertainty and observation loss
CN103714045A (en) * 2014-01-09 2014-04-09 北京理工大学 Information fusion estimation method for asynchronous multi-rate non-uniform sampled observation data
CN104297557A (en) * 2014-10-08 2015-01-21 北京航空航天大学 United navigation autonomous integrity monitoring method applicable to free flight of plurality of aircraft

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101630445A (en) * 2009-06-10 2010-01-20 民航数据通信有限责任公司 Method for processing data of asynchronous multiple sensors based on noise error and system thereof
CN101696906A (en) * 2009-10-28 2010-04-21 上海电机学院 Multi-source information fusion based state monitoring system and method of wind-driven generator
US20130246006A1 (en) * 2012-03-13 2013-09-19 King Fahd University Of Petroleum And Minerals Method for kalman filter state estimation in bilinear systems
CN103575278A (en) * 2013-10-08 2014-02-12 北京理工大学 Robust filtering method aiming at parameter uncertainty and observation loss
CN103714045A (en) * 2014-01-09 2014-04-09 北京理工大学 Information fusion estimation method for asynchronous multi-rate non-uniform sampled observation data
CN104297557A (en) * 2014-10-08 2015-01-21 北京航空航天大学 United navigation autonomous integrity monitoring method applicable to free flight of plurality of aircraft

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
一类噪声相关多传感器系统的新型序贯式融合滤波;宁涛 等;《大连理工大学学报》;20160331;第56卷(第2期);第208页至第211页 *
基于广义M估计的鲁棒容积卡尔曼滤波目标跟踪算法;吴昊 等;《物理学报》;20151231;第64卷(第21期);全文 *
基于异步多速率观测的噪声去相关融合算法;甘林海 等;《传感器与微系统》;20151231;第34卷(第8期);全文 *
基于状态转换的顺序式异步数据融合算法;葛泉波 等;《系统工程与电子技术》;20081231;第30卷(第12期);全文 *
多传感器自适应容积卡尔曼滤波融合算法;敖志刚 等;《计算机应用研究》;20140515;第31卷(第5期);第1312至第1315页 *
抗差与自适应组合的卡尔曼滤波算法在动态导航中的研究;李刚 等;《重庆邮电大学学报（自然科学版）》;20150228;第27卷(第1期);全文 *
船舶动力定位系统多传感器信息融合方法研究;徐树生;《万方数据知识服务平台》;20140417;全文 *

Also Published As

Publication number Publication date
CN108734218A (en) 2018-11-02

Similar Documents

Publication Publication Date Title
CN108734218B (en) Information fusion method and device of multi-sensor system
CN109474472A (en) A kind of fault detection method based on the more cell space filtering of holohedral symmetry
EP2098882A2 (en) Location measurement method based on predictive filter
CN112883820B (en) Road target 3D detection method and system based on laser radar point cloud
CN108875178B (en) Sensor arrangement method for reducing structural modal identification uncertainty
JP2008276785A (en) Fault detection and isolation method of sensor, and fault accommodation method of sensor using this
CN105571614B (en) magnetic sensor calibration method and device
CN108594250B (en) Point cloud data denoising method and device
JP2017194460A (en) Navigation system and method for error correction
KR101390776B1 (en) Localization device, method and robot using fuzzy extended kalman filter algorithm
CN112313536B (en) Object state acquisition method, movable platform and storage medium
CN112797967B (en) Random drift error compensation method of MEMS gyroscope based on graph optimization
CN109996205A (en) Data Fusion of Sensor method, apparatus, electronic equipment and storage medium
US20150032409A1 (en) Method and apparatus for calculating azimuth
CN112989625A (en) Method for eliminating abnormal value of UWB sensor
EP3770637B1 (en) Object recognizing device
CN112363617A (en) Method and device for acquiring human body action data
CN115727871A (en) Track quality detection method and device, electronic equipment and storage medium
CN114120252B (en) Automatic driving vehicle state identification method and device, electronic equipment and vehicle
CN112344966B (en) Positioning failure detection method and device, storage medium and electronic equipment
CN112285650B (en) Method, system and storage medium for positioning unknown wave velocity sound emission source in presence of abnormal TDOA
CN115235527A (en) Sensor external parameter calibration method and device and electronic equipment
US20130191071A1 (en) System and method for automatic modal parameter extraction in structural dynamics analysis
CN110175412B (en) Oil reservoir data fitting method, oil reservoir data analysis method and device
CN107976206B (en) MEMS gyroscope performance evaluation method based on information entropy

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant