CN104619015A - Conjugate gradient and simulated atomic transition-based indoor positioning method - Google Patents

Abstract

The invention discloses a conjugate gradient and simulated atomic transition-based wireless indoor positioning method. The indoor positioning method comprises the following steps: (1) randomly selecting N coordinate points as initial estimation coordinate group of an unknown node in a feasible domain range defined by adjacent beacon nodes; (2) taking each initial estimation coordinate in the group as a local initial point, calculating a local optimal value of a target function and the corresponding local optimal coordinate point according to the conjugate gradient method for the target function in parallel, judging whether the local optimal value meets a cut-off condition, and taking the corresponding coordinate as the global optimal coordinate of the unknown node to end the positioning if the condition is met; (3) performing the simulated atomic transition on the N target function local optimal values obtained by the parallel calculation, and jumping out of local optimization constraint of the step (2) to discover the global optimal coordinate point; (4) bombarding the N local optimal coordinate points obtained by the previous step to generate a new local estimation coordinate group, and transferring to (2). According to the method, the defect of long iteration time of the serial positioning method can be solved, and the positioning efficiency can be effectively improved.

Description

A kind of based on the indoor orientation method of conjugate gradient with simulation atomic transition

Technical field

The present invention relates to wireless sensor network field, particularly relate to a kind of wireless sensor network node indoor orientation method.

Background technology

Because wireless indoor location technology has wide utilization prospect in fields such as patient care, personal management, mall shoppings, wireless indoor location technology becomes the focus of people's research.In current wireless indoor location method, major part belongs to serial computing, the single solution of iteration optimization, optimizing process can be there is long, the shortcomings such as efficiency is low, and when wireless sensor network interior joint number is a lot, the localization method of this serial seems that efficiency is lower.

Summary of the invention

In order to solve iterative process length, inefficient problem that common serial localization method produces, the invention provides a kind of based on the wireless indoor location method of conjugate gradient with simulation atomic transition, under the prerequisite ensureing positioning precision, reduce positioning time, significantly improve location efficiency.

Provided by the invention a kind of based on the wireless indoor location method of conjugate gradient with simulation atomic transition, be improve localization method by the fusion of conjugate gradient method and simulation atomic transition.Its step comprises:

(1) within the scope of the feasible zone surrounded by contiguous beaconing nodes, the N number of coordinate of random selecting is as the original estimated coordinate colony of this unknown node;

(2) using each original estimated coordinate in colony as local and initial point, the local optimum of target function is obtained according to target function conjugate gradient method, and the suboptimization coordinate points of correspondence, judge whether local optimum meets cut-off condition, if meet, then using the global optimum coordinate points of the local optimum coordinate points of correspondence as unknown node, position fixing process terminates;

(3) to the local optimum of N number of target function that parallel computation obtains, carry out simulation atomic transition, to jump out the constraint of step (2) suboptimization to find global optimum's coordinate points, if reach maximum iteration time, terminate position fixing process;

(4) the N number of suboptimization coordinate colony after transition is carried out " bombardment ", produce new original estimated coordinate colony, forward step (2) to.

Contiguous beaconing nodes in described step (1) around unknown node has 3 at least.

Target function in described step (2) is defined as follows:

$f(x,y)=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(\sqrt{{(x-x_{\mathrm{mi}})}^2+{(y-y_{\mathrm{mi}})}^2}-{\hat{d}}_i)}^2$

In formula, (x _mi, y _mi) be the coordinate of i-th contiguous beaconing nodes of unknown node, be the measuring distance of unknown node to i-th contiguous beaconing nodes, m is the number of contiguous beaconing nodes.

Conjugate gradient method in described step (2) is a kind of method asking suboptimization, iteration speed fast, and its solution procedure is as follows:

(1) initial point X is chosen _k∈ R ⁿ, make k=1;

(2) compute gradient ▽ f (X _k), if ▽ f is (X _k)=0, then stop; Otherwise, calculate the direction of search $d_k=\left\{\begin{array}{cc}-\&dtri;f\left(X_k\right),& k=1\\ -\&dtri;f\left(X_k\right)+{\mathrm{\&beta;}}_k\&CenterDot;d_{k-1},& k\&GreaterEqual;2\end{array}\right.,$ Make ▽ f (X _k) ^td _k< 0, wherein parameter ${\mathrm{\&beta;}}_k=\frac{{|\&dtri;f\left(X_k\right)|}^2}{{|\&dtri;f\left(X_{k-1}\right)|}^2}$ Obtain; (3) X is made _k+1=X _k+ λ _kd _k, k=k+1, wherein step-length λ _k(λ _k> 0) obtained by Wolfe search criteria, go to step (2).

Location cut-off condition in described step (2) is setting threshold value p, then ends when the target function obtained is less than threshold value p.

In described step (3), the local optimum functional value of the N number of target function obtained is a series of discrete, discrete numerical value.

Simulation atomic transition method in described step (3) be based on atom from the higher excitation state of energy to the transition process of the ground state of minimum energy to find the optimization procedure of the minimum distribution of self-energy, be a kind of new nonlinear optimization method.

Boltzmann probability distribution is obeyed in distribution in described step (3) in simulation atomic transition method:

$p=\left\{\begin{array}{cc}1& F\left(i\right)>F\left(j\right)\\ \exp [-(F\left(i\right)-F\left(j\right))/\mathrm{kT}]& F\left(i\right)\&le;F\left(j\right)\end{array}\right.$

In formula, k is Boltzmann constant, and T is the absolute temperature of system.

The calculating process of the simulation atomic transition method in described step (3) is as follows:

(1) in the N number of local optimum coordinate colony obtained, a m is selected ^*as global optimum's coordinate points of unknown node, the target function value of its correspondence is local optimum functional value is F (m ^*);

(2) another local optimum coordinate of Stochastic choice n ^*, the local optimum functional value of its correspondence is F (n ^*);

(3) F (m is compared ^*) and F (n ^*) size, if F (n ^*) < F (m ^*), then m ^*n can be transitted to ^*, and establish n ^*for global optimum's coordinate points of unknown node; Otherwise calculate their Boltzmann probability: p=exp{-[F (n ^*)-F (m ^*)]/kT}, and a random generation Probability p _r(0 < p _r< 1).If p > is p _r, then m ^*n can be transitted to ^*, and m ^*it is still global optimum's coordinate points of unknown node; Otherwise, p < p _rthen can not transition.

Suboptimization coordinate colony after what " bombardment " in described step (4) represented is to transition carries out a random perturbation:

$X_{k+1}=X_k^{*}+\mathrm{\&delta;x}$

In formula, X _k+1for the new original estimated coordinate colony obtained, for the local optimum coordinate that kth is secondary, δ x is random number.

Compared with existing serial location technology, a kind of wireless indoor location method based on conjugate gradient and simulation atomic transition provided by the invention achieves concurrent operation on algorithm, reduce the iterations of wireless sensor network node location, significantly improve location efficiency.

Accompanying drawing explanation

Fig. 1 is the flow chart that conjugate gradient method solves local optimum;

Fig. 2 is the deployment diagram of beaconing nodes under indoor environment and unknown node;

Fig. 3 is the general flow chart of indoor node localization method of the present invention.

Embodiment

In order to be illustrated more clearly in the technical scheme of the present application embodiment, be briefly described to accompanying drawing required in embodiment below.Obviously, described embodiment is only a part of embodiment of the present invention, instead of whole embodiments, and other, not making the invention obtained under creative work prerequisite, all belong to protection scope of the present invention.

Wireless sensor network (WSN) is exactly need to detect the sensor node arranging some in ambient air, humidity, personal information wait observation area at one.In these sensor nodes, there are the beaconing nodes of known position information and unknown node to be positioned.

As follows based on localization method implementation step of the present invention:

(1) within the scope of the feasible zone surrounded by contiguous beaconing nodes, the N number of coordinate of random selecting is as the original estimated coordinate colony of this unknown node;

(2) using each original estimated coordinate in colony as local and initial point, local optimum and the suboptimization coordinate points of target function is obtained according to target function conjugate gradient method, judge whether local optimum meets location condition, if meet, using the global optimum coordinate points of the local optimum coordinate points of correspondence as unknown node;

(3) local optimum of the N number of target function obtained according to parallel computation, carries out simulation atomic transition, to jump out the constraint of suboptimization;

(4) to the N number of suboptimization coordinate colony after transition through row " bombardment ", produce new original estimated coordinate colony, forward step (2) to;

Concrete process is as shown in the overview flow chart of Fig. 1.

Contiguous beaconing nodes in described step (1) around unknown node has 3 at least, as shown in flow chart 2.

Target function in described step (2) is defined as follows:

$f(x,y)=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{(\sqrt{{(x-x_{\mathrm{mi}})}^2+{(y-y_{\mathrm{mi}})}^2}-{\hat{d}}_i)}^2$

In formula, (x _mi, y _mi) be the coordinate of i-th contiguous beaconing nodes of unknown node, be the measuring distance of unknown node to i-th beaconing nodes, m is the number of contiguous beaconing nodes.

Conjugate gradient method in described step (2) is that one asks suboptimization alternative manner fast, and its solution procedure is as follows:

(1) initial point X is chosen _k∈ R ⁿ, make k=1;

(2) compute gradient ▽ f (X _k), if ▽ f is (X _k)=0, then stop; Otherwise, calculate the direction of search $d_k=\left\{\begin{array}{cc}-\&dtri;f\left(X_k\right),& k=1\\ -\&dtri;f\left(X_k\right)+{\mathrm{\&beta;}}_k\&CenterDot;d_{k-1},& k\&GreaterEqual;2\end{array}\right.,$ Make ▽ f (X _k) ^td _k< 0, wherein parameter ${\mathrm{\&beta;}}_k=\frac{{|\&dtri;f\left(X_k\right)|}^2}{{|\&dtri;f\left(X_{k-1}\right)|}^2}$ Obtain; (3) X is made _k+1=X _k+ λ _kd _k, k=k+1, wherein step-length λ _k(λ _k> 0) obtained by Wolfe search criteria, go to step (2).

Concrete process as shown in Figure 3.

Location cut-off condition in described step (2) is setting threshold value p, then ends when the target function obtained is less than threshold value p.

In described step (3), the local optimum functional value of the N number of target function obtained is a series of discrete, discrete numerical value.

Simulation atomic transition method in described step (3) be simulation atom from the higher excitation state of energy to the optimization procedure that the transition process of the ground state of minimum energy distributes to find self energy minimum energy, be a kind of new non-linear optimum inversion method.

Boltzmann probability distribution is obeyed in distribution in described step (3) in simulation atomic transition method:

$p=\left\{\begin{array}{cc}1& F\left(i\right)>F\left(j\right)\\ \exp [-(F\left(i\right)-F\left(j\right))/\mathrm{kT}]& F\left(i\right)\&le;F\left(j\right)\end{array}\right.$

In formula, k is Boltzmann constant, and T is the absolute temperature of system.

The calculating process of the simulation atomic transition method in described step (3) is as follows:

(1) in the N number of local optimum coordinate colony obtained, a m is selected ^*as global optimum's coordinate points of unknown node, the target function value of its correspondence is local optimum functional value is F (m ^*);

(2) another local optimum coordinate of Stochastic choice n ^*, the local optimum functional value of its correspondence is F (n ^*);

(3) F (m is compared ^*) and F (n ^*) size, if F (n ^*) < F (m ^*), then m ^*n can be transitted to ^*, and establish n ^*for global optimum's coordinate points of unknown node; Otherwise calculate their Boltzmann probability: p=exp{-[F (n ^*)-F (m ^*)]/kT}, and a random generation Probability p _r(0 < p _r< 1), if p > is p _r, then m ^*n can be transitted to ^*, and m ^*it is still global optimum's coordinate points of unknown node; Otherwise, p < p _rthen can not transition.

Suboptimization coordinate colony after what " bombardment " in described step (4) represented is to transition carries out a random perturbation:

$X_{k+1}=X_k^{*}+\mathrm{\&delta;x}$

In formula, X _k+1for the new original estimated coordinate colony obtained, for the local optimum coordinate that kth is secondary, δ x is random number.

Meet and be less than given threshold value p or simulation atomic transition iterations and reach the coordinate that coordinate points corresponding to global optimum that maximum iteration time obtains is unknown node to be positioned.

Compared with existing serial location technology, a kind of wireless indoor location method based on conjugate gradient and simulation atomic transition provided by the invention achieves concurrent operation on algorithm, reduce the iterations of wireless sensor network node location, significantly improve location efficiency.

Claims (10)

1. based on the wireless indoor location method of conjugate gradient with simulation atomic transition, it is characterized in that, described method is the indoor orientation method merged based on conjugate gradient and simulation atomic transition.

2. indoor orientation method according to claim 1, is characterized in that, comprises the steps:

(1) within the scope of the feasible zone surrounded by contiguous beaconing nodes, the N number of coordinate of random selecting is as the original estimated coordinate colony of this unknown node;

(2) using each original estimated coordinate in colony as local and initial point, local optimum and the suboptimization coordinate points of target function is obtained according to target function conjugate gradient method, judge whether local optimum meets cut-off condition, if meet, then using the global optimum coordinate points of the local optimum coordinate points of correspondence as unknown node, location is terminated;

(3) local optimum of the N number of target function obtained according to parallel computation, carry out simulation atomic transition, to jump out the constraint of suboptimization to find global optimum's coordinate points, when reaching maximum iteration time, return the position coordinates of current optimum coordinates point as unknown node, location is terminated;

(4) to the N number of suboptimization coordinate colony after transition through row " bombardment ", produce new original estimated coordinate colony, forward step (2) to.

3. the indoor orientation method according to claim 1-2, is characterized in that, the contiguous beaconing nodes of described unknown node has 3 at least.

4. the indoor orientation method according to claim 1-2, is characterized in that, described location cut-off condition is setting threshold value p or iterations Kmax.

5. the indoor orientation method according to claim 1-2, is characterized in that, described target function is as follows:

$f(x,y)=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{(\sqrt{{(x-x_{\mathrm{mi}})}^2+{(y-y_{\mathrm{mi}})}^2}-{\hat{d}}_i)}^2---\left(1\right)$

In formula, (x _mi, y _mi) be the coordinate of i-th contiguous beaconing nodes of unknown node, be the measuring distance of unknown node to i-th beaconing nodes, m is the number of contiguous beaconing nodes.

6. the indoor orientation method according to claim 1-3, is characterized in that, described conjugate gradient method is a kind of conventional linear iterative algorithm asking local minimum problem fast, and its solution procedure is as follows:

(1) initial point X is chosen _k∈ R ⁿ, make k=1;

(2) compute gradient if then stop; Otherwise, calculate the direction of search $d_k=\left\{\begin{array}{cc}-\&dtri;f\left(X_k\right),& k=1\\ -\&dtri;f\left(X_k\right)+{\mathrm{\&beta;}}_k\&CenterDot;d_{k-1},& k\&GreaterEqual;2\end{array}\right.,$ Make ▽ f (X _k) ^td _k< 0, wherein parameter ${\mathrm{\&beta;}}_k=\frac{{|\&dtri;f\left(X_k\right)|}^2}{{|\&dtri;f\left(X_{k-1}\right)|}^2};$

(3) X is made _k+1=X _k+ λ _kd _k, k=k+1, wherein step-length λ _k(λ _k> 0) obtained by Wolfe search criteria, go to step (2).

7. the indoor orientation method according to claim 1-6, is characterized in that, the local optimum functional value of the N number of target function obtained is a series of discrete, discrete numerical value.

8. the indoor orientation method according to claim 1-7, is characterized in that, described simulation atomic transition method obeys Boltzmann probability distribution:

$p=\left\{\begin{array}{cc}1& F\left(i\right)>F\left(j\right)\\ \exp [-(F\left(i\right)-F\left(j\right))/\mathrm{kT},& F\left(i\right)\&le;F\left(j\right)\end{array}\right.---\left(2\right)$

In formula, k is Boltzmann constant, and T is the absolute temperature of system.

9. the indoor orientation method according to claim 1-8, is characterized in that, the process of described simulation atomic transition method is as follows:

(1) in the N number of local optimum coordinate colony obtained, a m is selected ^*as global optimum's coordinate points of unknown node, the target function value of its correspondence is local optimum functional value is F (m ^*);

(2) another local optimum coordinate of Stochastic choice n ^*, the local optimum functional value of its correspondence is F (n ^*);

(3) F (m is compared ^*) and F (n ^*) size, if F (n ^*) < F (m ^*), then m ^*n can be transitted to ^*, and establish n ^*for global optimum's coordinate points of unknown node; Otherwise calculate their Boltzmann probability: p=exp{-[F (n ^*)-F (m ^*)]/kT}, and a random generation Probability p _r(0 < p _r< 1), if p > is p _r, then m ^*n can be transitted to ^*, and m ^*it is still global optimum's coordinate points of unknown node; Otherwise, p < p _rthen can not transition.

10. the indoor orientation method according to claim 1-9, is characterized in that, the suboptimization coordinate colony after what described " bombardment " represented is to transition carries out a random perturbation:

$X_{k+1}=X_k^{*}+\mathrm{\&delta;x}---\left(3\right)$

In formula, X _k+1for the new original estimated coordinate colony obtained, for the local optimum coordinate that kth is secondary, δ x is random number.