CN109783844B - Multi-sensor management method based on tracking precision and energy consumption index - Google Patents

Abstract

The invention discloses a multi-sensor management method based on tracking precision and energy consumption indexes, which converts a sensor management problem into a multi-evaluation decision problem, and constructs a corresponding sensor identification framework and a super power set space by abstractly mapping each sensor node in a sensor network to a focal element in a DSmT theory so as to realize the modeling of a sensor. And then, obtaining a decision matrix under multiple evaluation indexes according to the tracking precision and the energy consumption index of each sensor combination. And finally, acquiring the optimal sensor combination at the current moment by using a BF-TOPSIS multi-evaluation index strategy to serve as a sensor for tracking the target at the next moment, thereby realizing the tracking of the target. The invention completes the modeling of multi-sensor management by the DSmT theory, thereby reducing the uncertainty problem in sensor selection; and through a multi-evaluation index decision strategy, the optimal sensor at the current moment is selected to the greatest extent, and the method is stable in performance and strong in transportability.

Description

Multi-sensor management method based on tracking precision and energy consumption index

Technical Field

The invention belongs to the technical field of information fusion, and particularly relates to a multi-sensor management method based on tracking accuracy and energy consumption indexes.

Background

With the mass production of cheap sensors and the great improvement of monitoring performance of the cheap sensors, the wireless sensor network is unprecedentedly developed in various fields (such as environment monitoring, industrial perception, fault diagnosis, military use, robot control and the like). Among them, the target tracking problem based on the sensor network is one of the research hotspots of the sensor network. However, due to physical limitations of the sensor network (e.g., communication bandwidth of the sensor network, battery capacity carried by the sensor, etc.), only a small number of sensors can be used for target tracking at a certain time. Therefore, it is necessary to design a proper sensor management strategy, and to select a proper subset of sensors (instead of all sensors working simultaneously) to be applied to the target tracking problem in a unit time.

From the perspective of methodology, the current sensor management method research mainly focuses on the research based on the information theory, and most of the researches are performed by utilizing the maximization of information increment after the information increment statistics is established. However, the disadvantage of this kind of information theory method is that the uncertainty problem of sensor management under a single index is not considered, that is, there are often inconsistency and even conflict situations between the optimal sensor management scheme under the information increment condition and the sensor management scheme under the tracking accuracy index, or the management result under the energy consumption minimization index.

Disclosure of Invention

The purpose of the invention is as follows: aiming at the defects, the invention provides a multi-sensor management method based on tracking precision and energy consumption indexes, which completes the modeling of multi-sensor management through a DSmT theory and reduces the uncertainty problem in sensor selection; and through a multi-evaluation index decision strategy, the optimal sensor at the current moment is selected to the greatest extent, and the method is stable in performance and strong in transportability.

The technical scheme is as follows: the invention provides a multi-sensor management method based on tracking precision and energy consumption indexes, which comprises the following steps:

(1) for each sensor in the sensor network, a sensor authentication framework is constructed: a ═ A₁,A₂,…,A_M}；

(2) Constructing a hyperpower set space D under the identification framework based on the DSmT theory^A＝{A，∪，∩｝；

(3) Performing target tracking on each sensor combination in the super power set through a Kalman filtering algorithm to obtain tracking precision;

(4) calculating the energy consumption of each sensor through an energy function according to the switching process of the sensors;

(5) constructing a decision matrix according to the tracking precision obtained in the step (3) and the energy consumption obtained in the step (4);

(6) and acquiring the best sensor representation through a BF-TOPSIS multi-evaluation strategy, and carrying out target tracking at the current moment by using the sensor combination.

Further, the specific steps of constructing the sensor identification framework in the step (1) are as follows:

the sensors in the raw sensor network have a₁,A₂,…,A_MThe sensor identification framework is directly formed by the sensors: a ═ A₁,A₂,…,A_M}。

Further, the specific steps of constructing the over-power set space in the step (2) are as follows:

A₁,A₂,…,A_M∈D^A (2.1)

a in the formula (2.1)_iI ∈ {1, …, M } represents a virtual sensor; d^ARepresents a power set space;

if A is_i,A_j∈D^AThen, then

A_i∩A_j∈D^A,A_i∪A_j∈D^A (2.2)

A in the formula (2.2)_i∩A_jIntersection operation constitutes a combined focal element, A_i∪A_jThe union set operation forms the extracted focal elements, and all the focal elements form a power set space, namely D^A＝{A,∪,∩}。

Further, the specific steps of obtaining the tracking accuracy in the step (3) are as follows:

X(k+1|k)＝F(k)X(k|k), (3.1)

P(k+1|k)＝F(k)P(k|k)F^T(k)+Q(k), (3.2)

γ(k+1)＝Z(k+1)-H(k+1)X(k+1|k), (3.3)

S(k+1)＝H(k+1)P(k+1|k)H^T(k+1)+R(k+1), (3.4)

K(k+1)＝P(k+1|k)H^T(k+1)S^-1(k+1), (3.5)

X(k+1|k+1)＝X(k+1|k)+K(k+1)γ(k+1), (3.6)

P(k+1|k+1)＝P(k+1|k)-K(k+1)·S(k+1)K^T(k+1). (3.7)

wherein X (k | k) is a state variable at time k, F (k) is a state transition matrix, H (k) is a measurement matrix, Q (k) and R (k) are a state noise covariance and an observation noise covariance, S (k) is an intermediate variable, K (k) is a Kalman gain, P (k | k) is an estimation covariance, Z (k +1) is measured data, and γ (k +1) is a current sensor A_iThe resulting tracking accuracy.

Further, the specific step of calculating the energy consumption of each sensor in the step (4) is as follows:

if the current task sensor A_iConsider the sensor activated at the next moment to be A_jAt that time, if i ≠ j, then sensor A_iThe consumed data transmission energy is calculated by the formula

Wherein e is_tAnd e_dIs a sensor A_iOf the transmitted data parameter, r_ijIs a sensor A_iAnd A_jEuclidean distance between, b_cα is an exponential coefficient, and is generally set to 2, for the number of bytes of data transfer.

Sensor A_jThe data reception energy consumption function of

E_r(A_j)＝e_rb_c (4.2)

Wherein e is_rIs a sensor A_jReceived energy consumption of b_cIs the number of bytes of data transfer. Similarly, sensor A_iThe energy consumption function of the monitoring and data processing is E_s(A_j)＝e_sb_s. Thus, sensor A is selected_jThe calculation formula of (2) is as follows:

Where e₀＝(e_t+e_r)b_c+e_sb_s,e₁＝e_db_c,e_s,b_sdata parameters are processed for the sensors.

The energy consumption was obtained according to the above equation (4.3).

Further, the specific steps of constructing the decision matrix in the step (5) are as follows:

and (3) constructing a decision matrix according to the tracking precision and the energy consumption corresponding to each sensor combination calculated in the formula (3.3) and the formula (4.3):

wherein, C₁Representing a tracking accuracy index, C₂Representing an energy consumption index, A_iRepresenting virtual sensors, S_j(i) Representative index C_jThe score of the ith sensor.

Further, the specific steps of obtaining the best sensor representation through a BF-TOPSIS multi-evaluation strategy in the step (6), and performing target tracking at the current time by using the sensor combination are as follows:

S_ijrepresents the score of the ith sensor under the jth index, | D^AI represents the number of elements in the hyperpower set;

ω in equations (6.8) and (6.9) represents the weight of the j-th index, and is generally set to 1/N, N is the number of indexes, and m is^best(A_i) To fully support A_iBasic probability assignment of (i.e. m)^best(A_i)＝m(A_i)＝1；m^worst(A_i) To completely not support A_iBasic probability assignment of (i.e. of)

And finally, sorting the sensors according to the values obtained by calculation in the step (6.10), wherein the sensors correspond to C (A)_i,A^best) The smaller the value, the most likely the sensor is selected.

Obtaining optimal sensor A^*And (4) repeating the formulas (3.1) - (3.7) in the step (3) so as to realize the tracking of the target.

The research of the prior art shows that the sensor management problem needs to comprehensively consider all the situations under the premise of multiple evaluation indexes, and the uncertainty caused by the management of the sensor is fully reduced from the perspective of fusion.

Considering that uncertainty exists in sensor management, and a strategy for selecting a sensor can be regarded as a decision problem under multiple evaluation indexes, here, a new idea given by the patent is to model the sensor itself under a decision-smart Theory (DSmT) theoretical framework, construct an identification framework based on the sensor, and then realize sensor management under a tracking problem through a Belief Function based on the Technique for Order Preference by Similarity to Ideal Solution (BF-TOPSIS) multiple evaluation index strategy.

According to the tracking precision and the energy consumption of the sensor combination, the management of the sensors is realized through a multi-evaluation index strategy. The method can firstly model the uncertainty problem of the sensor combination: mathematically describing different forms of sensor operation by intersection operations between focal elements in DSmT theory, (1) a single sensor operation corresponds to a single sub-focal element; (2) the simultaneous operation of the multiple sensors corresponds to extracting focal elements; (3) it cannot be determined to correspond to the confocal point. Meanwhile, all possible sensor combinations can be evaluated by utilizing the provided multi-evaluation index decision method, and the evaluation indexes are the tracking accuracy and the energy consumption. In addition, the sensor management method can be used for real-time tracking of single targets.

By adopting the technical scheme, the invention has the following beneficial effects:

the multi-sensor management method based on multiple evaluation indexes (tracking accuracy and energy consumption) can well solve the uncertain problem in the traditional sensor management method and can also solve the unreasonable phenomenon that a single evaluation index carries out sensor management.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a diagram illustrating the activation results of a single sensor at different times in an exemplary embodiment;

FIG. 3 is a diagram illustrating the activation results of multiple sensors at different times according to various criteria in an exemplary embodiment;

FIG. 4 is a diagram illustrating the energy consumption of a sensor under a single tracking accuracy indicator in an exemplary embodiment;

FIG. 5 is a diagram illustrating the energy consumption of a sensor under multiple criteria in an exemplary embodiment.

Detailed Description

The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.

As shown in fig. 1, the multi-sensor management method based on tracking accuracy and energy consumption index according to the present invention specifically includes the following steps:

1. multi-sensor modeling based on DSmT theory

Suppose the tracking system contains M independent sensors, denoted A_i(i ═ 1, …, M). According to the DSmT theory, the identification framework Α ═ a of the tracking system was constructed₁,A₂,A₃,…,A_i,…,A_MAccordingly, a multi-sensor based over-power set space D can be constructed according to the over-power set space definition in the DSmT theory^A{ a, [ u }. Wherein, the single sub-focal element A_iDefined as only one sensor a at the present moment_iThe method is applied to target tracking; combined focus element A_i∩A_jI e (1, …, M), j e (1, …, M) is defined as the current time sensor A_iAnd sensor A_jWorking at the same time; extracting coke element A_i∪A_jI e (1, …, M), j e (1, …, M) is defined as the sensor A which can not be determined at the current moment_iWorking or sensor a_jAnd (6) working.

2. Multi-sensor evaluation method based on tracking precision index

Aiming at the problem of single target tracking, at each moment, a Kalman Filter Kalman Filter (KF) is utilized to track a target by combining the measurement error of each sensor, and it needs to be noted that when the sensor is a single sub-focal element, the Kalman filtering algorithm is only used once; and if the target is the composite focal element, tracking the target by using the sequential Kalman filtering, and constructing a corresponding evaluation vector according to the tracking precision of each sensor combination after traversing the tracking effect of all the sensor combinations. The specific KF steps are as follows:

X(k+1|k)＝F(k)X(k|k),

P(k+1|k)＝F(k)P(k|k)F^T(k)+Q(k),

γ(k+1)＝Z(k+1)-H(k+1)X(k+1|k),

S(k+1)＝H(k+1)P(k+1|k)H^T(k+1)+R(k+1),

K(k+1)＝P(k+1|k)H^T(k+1)S^-1(k+1),

X(k+1|k+1)＝X(k+1|k)+K(k+1)γ(k+1),

P(k+1|k+1)＝P(k+1|k)-K(k+1)·S(k+1)K^T(k+1).

wherein X (-) is a state estimation value, P (-) is an estimation covariance, Q (-) is a state noise covariance, R (-) is an observation noise covariance, K (-) is a Kalman gain, F is a state transition matrix, and H is a measurement matrix.

Then, the tracking error of the target is calculated according to the measured value:

γ_i(k)＝|Z_i(k)-Z_i(k)|

3. multi-sensor evaluation method based on energy consumption index

Energy consumption is a cost function for target tracking using sensors, and in the target tracking problem, energy consumption is mainly concentrated on data communication (including data transmission and data reception) between sensors, and the sensors detect and process information inside the sensors. This patent uses an energy function model as follows:

if the current task sensor A_iConsider the sensor activated at the next moment to be A_jAt that time, if i ≠ j, then sensor A_iThe consumed data transmission energy is calculated by the formula

Wherein e is_tAnd e_dIs a sensor A_iOf the transmitted data parameter, r_ijIs a sensor A_iAnd A_jEuclidean distance between, b_cIs the number of bytes of data transfer.

Sensor A_jThe data reception energy consumption function of

E_r(A_j)＝e_rb_c

Wherein e is_rIs a sensor A_jThe received energy consumption of (2). Similarly, sensor A_iThe energy consumption function of the monitoring and data processing is E_s(A_j)＝e_sb_s. Thus, sensor A is selected_jThe energy consumption of (a) is calculated as follows:

Where e₀＝(e_t+e_r)b_c+e_sb_s,e₁＝e_db_c。

4. BF-TOPSIS-based multi-sensor management method

And (4) constructing a Decision Matrix (DM) according to the tracking precision and the energy consumption of each sensor combination obtained in the steps (2) and (3), and expressing the Decision Matrix by a mathematical symbol S.

Then, the evidence support and objection levels of each sensor combination are calculated, the Sup_j(A_i) The support degree is expressed in a decision matrix, and the jth evaluation index is corresponding to A_iThe degree of support of (c); inf_j(A_i) The degree of objection represents the j-th evaluation index for A_iOf the optical fiber.

Next, we can calculate the probability assignment for each sensor combination according to the support and objection degrees.

Finally, the interval distance function is combined

And

using the Ranking Index (Ranking Index) C (A)_i,A^best) And determining the overall support degree of each sensor combination as a unique evaluation index of the selected sensor.

The effect of the invention can be further illustrated by the following matlab simulation experiment.

Setting simulation experiment scenes: assume that the monitored area has 4 fixed position sensors, S1 (35m,100m), S2 (60m,160m), S3 (30m,20m) and S4 (55m,60 m). The starting position of the single target is (x)₀,y₀) (10m ), the corresponding motion velocity and acceleration are v ═ respectively (10m )v_x,v_y) (5m/s,6m/s) and a ═ a_x,a_y)＝(0m²/s,1m²And/s) to perform matlab simulation experiments by using the method of the present invention, and obtain the experimental results of fig. 2-5, wherein fig. 2-3 are the distribution results of the sensors at different times, and fig. 4-5 are the results of the energy consumption of the sensors in the simulation experiments.

2-3, it can be seen that, based on sensor management under multiple evaluation indexes, switching frequency between sensors is greatly reduced, which brings the advantage that communication between sensors is reduced, thereby ensuring that energy consumption of the sensors is reduced.

As can be seen from fig. 4 to 5, the sensor management under multiple evaluation indexes enables the energy consumption of each sensor in the simulation experiment to be respectively reduced as follows compared with the sensor management under a single tracking index: sensor 1: 19.67%; sensor 2: 59.24%; sensor 3: 27.43%; sensor 4: 10.59%.

Claims (2)

1. A multi-sensor management method based on tracking accuracy and energy consumption indexes is characterized by comprising the following steps:

(1) for each sensor in the sensor network, a sensor authentication framework is constructed: a ═ A₁,A₂,…,A_M}；

(2) Constructing a hyperpower set space D under the identification framework based on the DSmT theory^A＝{A,∪,∩}；

(3) Performing target tracking on each sensor combination in the super power set through a Kalman filtering algorithm to obtain tracking precision;

(4) calculating the energy consumption of each sensor through an energy function according to the switching process of the sensors;

(5) constructing a decision matrix according to the tracking precision obtained in the step (3) and the energy consumption obtained in the step (4);

(6) and acquiring the best sensor representation through a BF-TOPSIS multi-evaluation strategy, and carrying out target tracking at the current moment by using the sensor combination.

The specific steps of constructing the sensor identification frame in the step (1) are as follows:

the sensors in the raw sensor network have a₁,A₂,…,A_MThe sensor identification framework is directly formed by the sensors: a ═ A₁,A₂,…,A_M}。

The specific steps of constructing the overpowering set space in the step (2) are as follows:

A₁,A₂,…,A_M∈D^A

a in the formula (2.1)_iI ∈ {1, …, M } represents a virtual sensor; d^ARepresents a power set space;

if A is_i,A_j∈D^AThen, then

A_i∩A_j∈D^A,A_i∪A_j∈D^A

A in the formula (2.2)_i∩A_jIntersection operation constitutes a combined focal element, A_i∪A_jThe union set operation forms the extracted focal elements, and all the focal elements form a power set space, namely D^A＝{A,∪,∩}。

The specific steps for obtaining the tracking accuracy in the step (3) are as follows:

X(k+1|k)＝F(k)X(k|k),

P(k+1|k)＝F(k)P(k|k)F^T(k)+Q(k),

γ(k+1)＝Z(k+1)-H(k+1)X(k+1|k),

S(k+1)＝H(k+1)P(k+1|k)H^T(k+1)+R(k+1),

K(k+1)＝P(k+1|k)H^T(k+1)S^-1(k+1),

X(k+1|k+1)＝X(k+1|k)+K(k+1)γ(k+1),

P(k+1|k+1)＝P(k+1|k)-K(k+1)·S(k+1)K^T(k+1)；

wherein X (k | k) is a state variable at time k, F (k) is a state transition matrix, H (k) is a measurement matrix, Q (k) and R (k) are a state noise covariance and an observation noise covariance, S (k) is an intermediate variable, K (k) is a Kalman gain, P (k | k) is an estimation covariance, Z (k +1) is measured data, and γ (k +1) is a current sensor A_iThe resulting tracking accuracy.

The specific steps of calculating the energy consumption of each sensor in the step (4) are as follows:

if the current task sensor A_iConsider the sensor activated at the next moment to be A_jAt that time, if i ≠ j, then sensor A_iThe consumed data transmission energy is calculated by the formula

Wherein e is_tAnd e_dIs a sensor A_iOf the transmitted data parameter, r_ijIs a sensor A_iAnd A_jEuclidean distance between, b_cAlpha is an exponential coefficient and is set as 2;

sensor A_jThe data reception energy consumption function of

E_r(A_j)＝e_rb_c

Wherein e is_rIs a sensor A_jReceived energy consumption of b_cThe number of bytes for data transfer; similarly, sensor A_iThe energy consumption function of the monitoring and data processing is E_s(A_j)＝e_sb_s(ii) a Thus, sensor A is selected_jThe calculation formula of (2) is as follows:

wherein e₀＝(e_t+e_r)b_c+e_sb_s,e₁＝e_db_c,e_s,b_sProcessing data parameters for the sensor;

obtaining the energy consumption according to the formula;

the specific steps of constructing the decision matrix in the step (5) are as follows:

according to the calculated tracking precision and energy consumption corresponding to each sensor combination, a decision matrix is constructed and obtained:

wherein, C₁Representing a tracking accuracy index, C₂Representing an energy consumption index, A_iRepresenting virtual sensors, S_j(i) Representative index C_jThe score of the ith sensor.

2. The multi-sensor management method based on tracking accuracy and energy consumption index as claimed in claim 1, wherein the specific steps of obtaining the best sensor representation through BF-TOPSIS multi-evaluation strategy in the step (6) and using the sensor combination to track the target at the current time are as follows:

S_ijrepresents the score of the ith sensor under the jth index, | D^AI represents the number of elements in the hyperpower set;

in the above formula, ω represents the weight of the jth index and is set to 1/N, N is the number of indexes, and m is^best(A_i) To fully support A_iBasic probability assignment of (i.e. m)^best(A_i)＝m(A_i)＝1；m^worst(A_i) To completely not support A_iBasic probability assignment of (i.e. of)

Finally, according to the value obtained by calculation in the formula, sorting the sensors, and obtaining the C (A) corresponding to the sensors_i,A^best) The smaller the value, the most likely the sensor is selected;

obtaining optimal sensor A^*After that, repeatAnd (3) tracking the target.

Images