CN106257301B - Distributed space time correlation model trace tracking method based on statistical inference - Google Patents
Distributed space time correlation model trace tracking method based on statistical inference Download PDFInfo
- Publication number
- CN106257301B CN106257301B CN201610315568.1A CN201610315568A CN106257301B CN 106257301 B CN106257301 B CN 106257301B CN 201610315568 A CN201610315568 A CN 201610315568A CN 106257301 B CN106257301 B CN 106257301B
- Authority
- CN
- China
- Prior art keywords
- boundary
- time
- beaconing nodes
- unknown node
- region
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
- G01S5/0278—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves involving statistical or probabilistic considerations
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
- G01S5/0294—Trajectory determination or predictive filtering, e.g. target tracking or Kalman filtering
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Probability & Statistics with Applications (AREA)
- Mobile Radio Communication Systems (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The present invention relates to a kind of distributed space time correlation model trace tracking method based on statistical inference, according to the equidistant latticed deployment beaconing nodes in longitudinal and transverse direction in located space, by beaconing nodes information preservation at each unknown node；It receives unknown node and sends the beaconing nodes of notification information with fixed frequency transmitting positioning signal, unknown node receives simultaneously is respectively formed multiple time serieses according to beaconing nodes；Each unknown node builds boundary time sequence respectively, detects boundary crossover event and determines corresponding time point；Each unknown node builds zone time window statistic, infers current region where it；Infer track and boundary intersection position；Track is formed, aggregation node is as a result uploaded.The method of the present invention considers RSSI location informations Probability Characteristics in track following problem with spatiotemporal data structure as a whole, finds that boundary is jumped event and area information in a manner of space time information statistical inference, realizes track following.
Description
Technical field
The present invention relates to a kind of wireless sensor network monitoring technology, specifically a kind of distribution based on statistical inference
Formula space time correlation model trace tracking method.
Background technology
It simply, not by sighting condition is influenced due at low cost, deployment, wireless sensor network WSN (Wireless
Sensor Network) it is widely used in all kinds of Indoor Monitoring Systems, including safety production monitoring of coal mine system, subway are applied
The real-time security early warning system of work, indoor precisely navigation system etc..The indoor application of wireless sensor network is also in rapidly expanding
Among exhibition.
For many indoor applications, the movement locus of target is most basic information, is to be based on location-based service LBS
The basis that (Location Based Service) is realized.Such as in safety production monitoring of coal mine system, personnel, mobile device
Track is important monitoring content；According to personnel and equipment, the fortune of equipment and equipment in the real-time security early warning system of subway work
Dynamic trajectory predictions simultaneously trigger pre-warning signal.
Researcher proposes a variety of wireless sensor network track followings (or prediction) algorithm, these algorithms are all right
The movement locus of target is obtained on the basis of target positioning.This kind of algorithm can be divided into two stages, position, obtain to target first
Spatiotemporal data structure is carried out again after to discrete target position information, obtains the movement locus time sequence of (or prediction) target
Row, the as a result, accuracy of target trajectory are largely determined by the precision of location algorithm.
It is positioned specific to target, due to received signal strength indicator RSSI (Received Signal Strength
Indication it) is easily obtained, the location technology based on RSSI is constantly subjected to lasting concern；Indoor application due to barrier,
The limitation of sighting condition, many occasions of location technology based on RSSI are by as most important location technology.Essentially,
Location technology based on RSSI belongs to the location technology based on ranging, and the conventional practice is first to establish the ranging model of RSSI,
The distance between unknown node and beaconing nodes are conversed according to measured RSSI value in position fixing process, list three side positioning equations
Solution obtains the current location of unknown node, continues the rail that series of discrete point obtained by this process just constitutes unknown node
Mark.The main problem of this method be RSSI ranging model by site environment, multipath, diffraction, measuring technique etc. it is a variety of because
Element influence and lead to have strong time-varying characteristics, positioning accuracy is relatively low, and position success rate is not high.
Existing wireless sensor network trace tracking method is to form track by consecutive tracking, does not make full use of one
Determine opening up for the probability distribution statistical feature whithin a period of time of neighbor beacon node locating information in spatial dimension and wireless network
The local spatial information that structure is included is flutterred, existing method is poor to environmental suitability, and error is easy to accumulate, the precision of track following
It is low.
Invention content
Poor, error accumulation, track essence for wireless sensor network trace tracking method environmental suitability in the prior art
The deficiencies of relatively low is spent, does not accumulate and carries the technical problem to be solved in the present invention is to provide a kind of good environmental adaptability, error
The distributed space time correlation model trace tracking method based on statistical inference of high track following accuracy.
In order to solve the above technical problems, the technical solution adopted by the present invention is：
The present invention is based on the distributed space time correlation model trace tracking methods of statistical inference, include the following steps：
According to the equidistant latticed deployment beaconing nodes in longitudinal and transverse direction in located space, by beaconing nodes information
{NodeID, (x, y) } and it is stored at each unknown node, in unknown node forming region information and boundary information, wherein NodeID
For beaconing nodes Id, (x, y) is its coordinate；
It receives unknown node and sends the beaconing nodes of notification information with fixed frequency transmitting positioning signal, unknown node connects
It receives and is respectively formed multiple time series R according to beaconing nodesi；
Each unknown node builds boundary time sequence respectively, detects boundary crossover event and determines corresponding time point；Respectively not
Know node disjoint dynamic cutting time window, build zone time window statistic, is inferred according to zone time window statistic
Current region；
Infer track and boundary intersection position；Track is formed, aggregation node is as a result uploaded.
The unknown node receives and is respectively formed multiple time series R according to beaconing nodesiFor：Unknown node will receive
The positioning signal that multiple independent beaconing nodes are sent, is divided into multigroup according to beaconing nodes, and every group forms sequence according to time order and function
Row, as time series.
The structure boundary time sequence is：As unit of beaconing nodes group, the boundary RSSI times are built as follows
Sequence：
Rb=(rb1,rb2,rb3...rbk...rbs) (4)
Wherein, it includes beaconing nodes number, r that m, which is boundary,ikFor k-th of RSSI value of i-th of beaconing nodes, b is boundary
Number, rbkFor k-th of value of the boundary time sequence, k is the serial number in time series, and s is boundary time sequence length.
The detection boundary crossover event is：
Judge boundary time sequence in T2Inside whether there is primary more apparent ascending and descending process, if it is, hair
Raw boundary crossover event；Maximum point in ascending and descending process indicates time point of the unknown node by boundary；Corresponding region
Time window T1The period between event time point is crossed over for adjacent boundary.
Infer that current region is realized by following algorithm according to zone time window statistic：
argmax(W(S,T1)) (23)
Wherein, W (S, T1) it is time window T1For interior beaconing nodes using region as unit RSSI sizes, RSSI is to receive signal
Intensity instruction, W (S, T1) it is a statistic, n is that region S includes beaconing nodes number, and i is beaconing nodes, and j is in time series
Component number, rijFor corresponding RSSI values, l is length of time series, l and time window size T1It is in direct ratio：
L=T1*f (3)
Wherein f is positioning signal tranmitting frequency；
argmax(W(S,T1)) indicate to find out statistic W (S, T1) maximum region.
It is realized by following formula with boundary intersection position the deduction track：
Wherein, A is track and top horizontal boundary intersection point, and B is beaconing nodes, and C is track and longitudinal boundary intersection point, and D is
Track and following horizontal boundary intersection point, E are lower left corner beaconing nodes, t1Event time point, t are corresponded to for A2Event time is corresponded to for C
Point, t3Event time point is corresponded to for D, v is the movement speed of unknown node, and BC+CE=1 indicates not considering the actual area length of side
Q, solving equations can obtain C points and account for length of side percentage, to determine C points position according to region length of side q, can similarly acquire D points position；
Line behind the position C, D is acquired, the movement locus of the personnel for wearing unknown node or machinery is obtained.
Space time correlation model based on statistical inference is defined as follows：
1) region：The plane subdivision that two dimensional surface is divided by four adjacent in length and breadth beaconing nodes, each lattice
For a region, the lattice length of side is q；
2) boundary：The plane line of demarcation that two dimensional surface is made of the beaconing nodes of horizontal, every horizontal line/
Ordinate (Bh/Bv) it is a line circle；
3) unknown node receives the positioning signal from each beaconing nodes and constitutes one according to the priority of arrival time
RSSI time serieses are：
Ri=(ri1,ri2,ri3...rik...ril)(1)
In a period of time, location information that each beaconing nodes are sent to current unknown node according to arrival time priority
Form a Ri, i expression beaconing nodes ID numbers；Lower target component k is implicit to indicate relative time information, and l is length of time series；
4) space time correlation：Infer area information corresponding with time window according to area of space forming region statistic, press
Boundary time Sequence Detection boundary crossover event is formed according to boundary；The included beaconing nodes number m in boundary is true according to field condition
Fixed, selection can most reflect that the beaconing nodes with unknown node relative position relation are calculated；
5) distributed computing method：Each unknown node in the location information of locally storage beaconing nodes, simultaneously examine by selected boundary
Boundary crossover event is surveyed, divides time window according to boundary crossover event time click and sweep, in local computing range statistics amount and counts
Infer corresponding region information in time window, and then trace information is calculated, the above each unknown node of calculating process is independent
It carries out, and result of calculation is uploaded into aggregation node；
6) time window：Distributed computing method each unknown node dynamic cutting in calculating process goes out two class time windows
T1With window T2, time window is each unknown node according to time series RiSituation of change independently divides ground, and each unknown node is mutual
Between do not influence, wherein T1For zone time window, T2For boundary time window.
The invention has the advantages that and advantage：
1. the method for the present invention makes full use of the distribution computing capability of wireless sensor network, by RSSI in track following problem
Location information Probability Characteristics are considered as a whole with spatiotemporal data structure, found in a manner of space time information statistical inference boundary across
Jump event and area information realize track following.
2. track following model of the method for the present invention based on location information space-time relationship, track following algorithm is converted to
Classification, feature discovery procedure during Spatial-temporal Information Fusion propose to correspond to the location information within the scope of certain space one
The regional determination algorithm SDA (Subarea Determination Algorithm) handled in a time window, algorithm use two
The statistic of dimension reconstruct judges that region corresponding with the time window, the influence of effectively overcoming noise greatly improve track following
Accuracy rate.
3. the method for the present invention proposes to carry out one-dimensional reconstruct from the data comprising abundant space time information to find boundary crossover
The detection algorithm of event SBE (Spanning Boundary Event), illustrates boundary crossover event time sequence in a model
Including abundant information, algorithm provides key message for track following, and field experiment is with simulation results show based on statistics
The correctness of the space time correlation model track following algorithm of deduction, the validity of algorithm is demonstrated from level of practice.
4. in addition to boundary is jumped other than the corresponding key position information of event (such as position C, D), the track between key point is used
Straight line is approximate, and in the case where region length of side q (field experiment takes q=6m to obtain very high precision) is smaller, this is approximately to close
Reason.
Description of the drawings
Fig. 1 is the regional determination method SDA diagrams involved in the method for the present invention；
Fig. 2 is that the boundary time window involved in the method for the present invention is illustrated with SBE detections；
Fig. 3 is that unknown node and the relativeness in region illustrate in the method for the present invention；
Fig. 4 is unknown node and region W (S, T in the method for the present invention1) relationship diagram；
Fig. 5 A are Y in the method for the present inventionSScatter chart；
Fig. 5 B are W (S, T in the method for the present invention1) scatter chart；
Fig. 6 is R in the method for the present inventionbBoundary time sequence RbVariation tendency diagram；
Fig. 7 is that sequence of events 1 illustrates in the method for the present invention；
Fig. 8 is that sequence of events 2 illustrates in the method for the present invention.
Specific implementation mode
The present invention is further elaborated with reference to the accompanying drawings of the specification.
The present invention is based on the distributed space time correlation model trace tracking methods of statistical inference, include the following steps：
According to the equidistant latticed deployment beaconing nodes in longitudinal and transverse direction in located space, by beaconing nodes information
{NodeID, (x, y) } and it is stored at each unknown node, in unknown node forming region information and boundary information, wherein NodeID
For beaconing nodes Id, (x, y) is its coordinate；
It receives unknown node and sends the beaconing nodes of notification information with fixed frequency transmitting positioning signal, unknown node connects
It receives and is respectively formed multiple time series R according to beaconing nodesi；
Each unknown node builds boundary time sequence respectively, detects boundary crossover event and determines corresponding time point；Respectively not
Know node disjoint dynamic cutting time window, build zone time window statistic, is inferred according to zone time window statistic
Current region；
Infer track and boundary intersection position；Track is formed, aggregation node is as a result uploaded.
The method of the present invention needs to arrange for data structure storage area information and boundary information at each unknown node, respectively not
Dynamic divides time window and forms track with knowing node disjoint, and structure is finally uploaded aggregation node, is a kind of distributed
Computational methods.
Unknown node receives and is respectively formed multiple time series R according to beaconing nodesiFor：Unknown node will receive multiple
The positioning signal that independent beaconing nodes are sent, be divided into according to beaconing nodes it is multigroup, every group according to time order and function formation sequence, i.e.,
For time series.
Space time correlation model based on statistical inference is defined as follows：
1) region：The plane subdivision that two dimensional surface is divided by four adjacent in length and breadth beaconing nodes, each lattice
For a region；
2) boundary：The plane line of demarcation that two dimensional surface is made of the beaconing nodes of horizontal, every horizontal line/
Ordinate (Bh/Bv) it is a line circle；
3) unknown node receives the positioning signal from each beaconing nodes and constitutes one according to the priority of arrival time
RSSI time serieses are：
Ri=(ri1,ri2,ri3...rik...ril) (1)
In a period of time, location information that each beaconing nodes are sent to current unknown node according to arrival time priority
Form a Ri, i expression beaconing nodes ID numbers；Lower target component k is implicit to indicate relative time information, and l is length of time series；
4) space time correlation：Infer area information corresponding with time window according to area of space forming region statistic, press
Boundary time Sequence Detection boundary crossover event is formed according to boundary；The included beaconing nodes number m in boundary is true according to field condition
Fixed, selection can most reflect that the beaconing nodes with unknown node relative position relation are calculated；
5) distributed computing method：Each unknown node in the location information of locally storage beaconing nodes, simultaneously examine by selected boundary
Boundary crossover event is surveyed, divides time window according to boundary crossover event time click and sweep, in local computing range statistics amount and counts
Infer corresponding region information in time window, and then trace information is calculated, the above each unknown node of calculating process is independent
It carries out, and result of calculation is uploaded into aggregation node；
6) time window：Distributed computing method each unknown node dynamic cutting in calculating process goes out two class time windows
T1With window T2, time window is each unknown node according to time series RiSituation of change independently divides ground, and each unknown node is mutual
Between do not influence, wherein T1For zone time window, T2For boundary time window.
The method of the present invention takes full advantage of the local spatial information that the topological structure of wireless network is included, and in time window
Statistical inference is carried out to positioning time sequence in mouthful, reduces the influence of noise and exceptional value to result of calculation as far as possible, relative to
Other methods improve path accuracy, method fully excavated include in location information when, empty information.
In order to make full use of the location information probability distribution statistical characteristic of mobile unknown node, for indoor application (letter
Mark node can be disposed artificially), propose the space time correlation model definition based on statistical inference, below (such as Fig. 1 by taking 2 dimensional planes as an example
It is shown) it is described.
Define 1. regions (Subarea)
The plane subdivision that 2 dimensional planes are divided by four adjacent in length and breadth beaconing nodes, each lattice is one in Fig. 1
A region.
Define 2. boundaries (Boundary)
The plane line of demarcation that 2 dimensional planes are made of the beaconing nodes of horizontal, every horizontal line/ordinate in Fig. 1
(Bh/Bv) it is a line circle.
Unknown node sends notification signal in model, and receive notification signal closes on beaconing nodes with identical frequency independence
Ground emits positioning signal, and unknown node receives the positioning signal from each beaconing nodes and constituted according to the priority of arrival time
One RSSI time series (Time Series)
Ri=(ri1,ri2,ri3...rik...ril)。 (1)
In a period of time, location information that each beaconing nodes are sent to current unknown node according to arrival time priority
Form a Ri, i expression beaconing nodes ID numbers, RiIt is stored at unknown node and is handled.In smaller range (relative to
Node communication radius), it is assumed that it closes on the emitted positioning signal unknown node of beaconing nodes and all can receive, then at the same time
The equal length of different time sequence in window, the same lower target component of different time sequence correspond to almost same time point (see
Fig. 1), in other words, it is equivalent in some position, unknown node receives it and closes on the positioning signal of beaconing nodes transmitting, when not
After knowing node motion to the next position, the positioning signal of these beaconing nodes transmission is received again, is analogized with secondary.Lower target point
The implicit expression relative time informations of k are measured, i.e., are k* (1/f) relative to the time of time window starting point, f is that beaconing nodes transmitting is fixed
Position signal frequency (secondary/second).
In wireless sensor network, since the RSSI value of positioning signal can be obtained without specialized hardware, therefore it is based on RSSI
Ranging localization be research hot spot.On the other hand, larger range error is brought with noisy RSSI value, causes to be positioned to
Power is relatively low, or even more outlier occurs.
Time series RiValue be with noisy RSSI value, the influence of noise of varying environment is different；Same ring
Under border, the influence of noise is also time-varying.Although there are many attenuation models of RSSI and distance, for practical problem, it is difficult to
Accurately select model and the parameter of these models.
After 2 dimensional planes define region and boundary, the normal movement of unknown node is exactly from a region to another
A adjacent area crosses over the process on a certain boundary therebetween.Without loss of generality, it is assumed that unknown node is done approximate even in smaller range
Speed linear motion.In testing at the scene, as long as velocity variations are not violent, the direction of motion generally maintains straight line side in adjacent area
To, so that it may to obtain good track following effect.
For track following problem, unknown node is continuous moving, after introduce region and boundary, definition region statistics
Amount and boundary time sequence, carry out track following in a manner of statistical inference.Region and the topology that boundary is according to wireless network
The beaconing nodes combination that the local spatial information that structure is included is formed, considers region and boundary in the form of positioning signal reconstructs
Involved beaconing nodes, the positioning signal after reconstruct contain abundant space time information, these statistics are analyzed in time window
Amount, can be effectively reduced influence of noise, avoid outlier.
Define 3. region RSSI time window statistics
Infer that current region is realized by following algorithm according to zone time window statistic：
argmax(W(S,T1)) (23)
Wherein, W (S, T1) it is time window T1For interior beaconing nodes using region as unit RSSI sizes, RSSI is to receive signal
Intensity indicates that n is that region S includes beaconing nodes number, and i is beaconing nodes, and j is component number in time series, rijFor correspondence
RSSI values, l is length of time series, l and time window size T1In direct ratio, wherein f is positioning signal tranmitting frequency,
L=T1*f (3)
W(S,T1) concentrate illustrate time window T1Interior beaconing nodes are to original positioning using region as unit RSSI sizes
The Two-Dimensional Reconstruction of information.
argmax(W(S,T1)) indicate to find out statistic W (S, T1) maximum region.
Define 4. boundary RSSI time serieses
Building boundary time sequence is：Select the beaconing nodes on time upper adjacent interregional ordinate or horizontal line
Group builds boundary RSSI time serieses as follows,
Rb=(rb1,rb2,rb3...rbk...rbs), (4)
Wherein
M is that boundary includes beaconing nodes number, rikFor k-th of RSSI value of i-th of beaconing nodes, b is boundary number, rbk
For k-th of value of the boundary time sequence, k is the serial number in time series.
As shown in Fig. 2, it is time-varying that boundary, which corresponds to beaconing nodes, with time window T1Corresponding region difference without
Together, algorithm makes every effort to selection and the beaconing nodes participation statistical inference on the immediate boundary line of motion track, and s is boundary time sequence
Row length, RbEach component be that the m beaconing nodes selected from boundary correspond to RSSI mean values, boundary RSSI time serieses pair
Answer time window T2Necessary crossing the boundary, RbReflect time window T2The interior RSSI situations of change as unit of boundary are to original
The one-dimensional reconstruct of beginning location information.
Define 5. boundary crossover events
Detecting boundary crossover event is：Judge boundary time sequence in T2Inside whether occur under primary more apparent rising
Drop process, if it is, boundary crossover event occurs；Maximum point in ascending and descending process indicates unknown node by boundary
Time point.
Unknown node reaches adjacent area from region, must therebetween some time point jump boundary (Bh/Bv),
Boundary crossover event can be expressed as：
espan(T2,t,B) (6)
Wherein T2For the time window where the event, t is the time point that event occurs, and B numbers for the boundary jumped.
The physical meaning of boundary crossover event is in T2Interior, unknown node gradually approaches, reaches B, later again far from B, therebetween in t and B
Intersection.
Region RSSI time windows statistic, boundary RSSI time serieses and boundary crossover event espanAll include abundant
Time, spatial information.
It is realized by following formula with boundary intersection position the deduction track：
Wherein, A is track and top horizontal boundary intersection point, and B is beaconing nodes, and C is track and longitudinal boundary intersection point, and D is
Track and following horizontal boundary intersection point, E are lower-left angle point, t1Event time point, t are corresponded to for A2Event time, t are corresponded to for C3For D
Corresponding event time point, v are the movement speed of unknown node, and BC+CE=1 indicates not considering the actual area length of side, solving equations
C points can be obtained and account for length of side percentage, to determine C points position according to region length of side q, can similarly acquire D points position；
Acquire line behind the position C, D (a small range hypothesis moves along a straight line), obtain the personnel for wearing unknown node or
The movement locus of machinery.
When unknown node is moved along the direction parallel with boundary, the intersection point with certain direction adjacent boundary can only be determined,
Track is parallel with boundary from last time point at this time.
In the present invention, W (S, T1) reflect time window T1The interior RSSI overall distribution situations as unit of region, below
W (S, T are analyzed by taking Lognormal shadowing model (Shadowing models) as an example1) distribution, and regional determination algorithm is managed
By analysis.
W(S,T1) probability distribution
Shadowing models are the path loss models most generally used in wireless sensor network research at present, comprehensive
Conjunction considers signal and decays in transmission process and relationship and anisotropic propagation feature, the theoretical model of RSSI between
It is as follows：
P (d) indicates the average energy value (unit mW) at distance signal launch point d, P (d0) indicate distance d0The energy at place is equal
Value, d0It is taken as 1m, β is the path attenuation factor, and environment is different, and β values are different.RSSI units are dBm, to：
RSSI (d)=RSSI (d0)-10βlg10(d/d0), (8)
RSSI (d) indicates that distance is the RSSI, RSSI (d at d0) indicate that distance is d0Reference point at RSSI.From theoretical mould
As it can be seen that when β values are in a certain range, RSSI reduces type with the increase of d, and the speed reduced is first quick and back slow.
The positioning signal RSSI that unknown node receives meets formula
RSSI (d)=RSSI (d0)-10βlg10(d/d0)+X, (9)
Wherein X indicates RSSI noises caused by multipath, diffraction, many factors such as barrier, and X is counted as one with 0 as mean value, σ
For the white Gaussian noise of variance, i.e.,
X~N (μ, σ2), (10)
The probability density function of X is
With environment difference, σ values are different.
Assuming that：
1) unknown node is in time window T1In always situated in some region inside, such as always situated in region SA, next with
SAThe region S of surroundingB、SC(as shown in Figure 3) as a comparison, illustrates different zones W (S, T1) statistical property difference.
2) in the contrast range much smaller than node communication radius, it is ensured that beaconing nodes emit and are unknown node
The positioning signal number received is identical and is l；For number be less than l region, in range needed to be considered it
Outside.
3) in the trace model of track, parameter setting (transmission power size) phase of all beaconing nodes transmitting positioning signals
Identical with, institute's configuration antenna, 2 dimensional planes are divided into multiple regions, in region and adjacent area that regional determination algorithm is related to
(it is much smaller than node communication radius) in smaller range, it is assumed that β and σ2Value is identical.
Define 6. region RSSI vectors
VS=(rul,rur,rlr,rll), (12)
VSIndicate that certain time point unknown node received from certain region clockwise four from the upper left corner to the lower left corner
A beaconing nodes RSSI composition of vector.
Define the sum of 7. region RSSI vectors
YSWith the form of matrix multiple indicate the cumulative of four components and, on the one hand explanation is in the power with each component in value
Heavy phase is same, on the other hand illustrates YSIndicate the relativeness of the time point unknown node and region.
W(S,T1) distribution and geometric meaning
RSSI reduces with the increase of d, the speed of reduction first quick and back slow, W (S, T1) by l position piCorresponding YSIt calculates
It is worth to, for each YS, distribution mean value be solely dependent upon piAt a distance from corresponding four beaconing nodes in each region.With
SA、SBFor (as shown in Figure 4), l is removed3、l4Two common edges are outer, l1>l5、l2>l4,To W (SA,T1) distribution
Mean value is more than W (SB,T1) distribution mean value.
W(S,T1) by l position piLocate YSCalculating is worth to, and it is equal to be equivalent to certain multiple survey calculation of fixed point in the S of region
The effect of value (from the point of view of reducing variance).
Based on W (S, T1) regional determination algorithm
For mobile unknown node, in time window T1Interior W (S, T1) distribution as shown in Fig. 5 A, 5B), Fig. 5 A
For YSDistribution, Fig. 5 B be W (S, T1) distribution, it is seen then that W (S, T1) adjacent area distribution cross section area contract significantly
Small, which reduce the possibilities of generation area erroneous judgement.
In pattern-recognition, it is known that the conditional probability density of overall probability distribution and class per class things uses Bayes
(Bayes) decision theory realizes that the method for classification is following (assuming that there are two classifications altogetherWith),
It is knownAnd
It indicatesIn observe the probability density of x,It indicatesIn observe that the probability of x is close
Degree, then
Expression observes which classification x belongs toProbability.According to minimal error rate Bayes decision rule, ifThen
As shown in Fig. 5 A, 5B, Fig. 5 A indicate YsDistribution, 5B be W (S, T1) distribution, it is seen that the possibility judged by accident
It greatly reduces.
In regional determination algorithm, the judgement to unknown node region, when be placed on studied in time window when, largely
Field experiment confirms, it is only necessary to consider SA、SBWith SCThree classes region, that is,：
p(SA|W(S,T1))+p(SB|W(S,T1))+p(SC|W(S,T1))=1 (22)
Due to every class region W (S, T1) distribution mean value be not fixed the movement locus of unknown node (depend on), therefore it is selected
Relative quantity is classified.
Based on W (S, T1) regional determination algorithm, in T1It is interior that W (S, T are calculated separately as unit of region1), with
argmax(W(S,T1)) (23)
As unknown node in T1Interior region.
Regional determination algorithm accuracy, in T1W that the interior influence due to noise makes non-unknown node be currently located region (S,
T1) it is more than that region erroneous judgement has occurred, the influence factor of erroneous judgement possibility occurrence is discussed below when being currently located region.
If regional determination is correct, that is, meets
(W(SA, T) and ＞ W (SB,T))∩(W(SA, T) and ＞ W (SC,T)) (24)
It judges by accident namely in region
(W(SA, T) and ＜ W (SB,T))∪(W(SA, T) and ＜ W (SC,T)) (25)
The size of dash area indicates that the possibility that region erroneous judgement occurs, size depend on distribution mean value in Fig. 5
Difference, the variance of distributionAnd varianceSlow down movement speed depending on l (σ is determined by environment) again (to increase in area
Residence time in domain), increase positioning signal tranmitting frequency can improve regional determination accuracy.
The SBE detections of boundary crossover event
According to the regional determination that the time window of dynamic cutting carries out, when the time window that dynamic divides and boundary coincide
When, highest accuracy can be reached.It can be concluded that it therebetween must be at some time point in the motion process of unknown node
Jump boundary (Bh/Bv), incident Detection Algorithm is according to boundary RSSI time serieses RBVariation tendency find boundary crossover event
espan(T2,t,B)。
Time window T2Adjacent area is taken to correspond to the middle section (see Fig. 2) of time window, according to the opposite of adjacent area
It can most reflect the beaconing nodes (beaconing nodes nearest apart from unknown node) of RSSI variation tendencies on position relationship selection boundary
Build RB, to avoid influence of the catastrophe point to judgement, to RBLeast square fitting processing is carried out to reflect RBGeneral morphologictrend,
Work as RBWhen there is apparent ascending and descending process and non-pseudo- peak, Decision boundaries cross over event, process of fitting treatment curve obtained
Peak point be event occur time point.
Boundary crossover event SBE detecting steps are as follows：
1)T2The interior one-dimensional reconstruct of time series obtains boundary time sequence Rb；
2) to RbCarry out least square fitting processing；
3) judge T2Inside whether there is primary more apparent ascending and descending process, if it is event occurs, finds
Maximum point；
If 4) do not had, event occurs, expands T2, return and 1) detect event again；
5) maximum point is boundary crossover event, indicates time point of the unknown node by boundary.
RBVariation tendency as shown in fig. 6, in the track following model of space time correlation, unknown node is moved from some region
Its adjacent area is moved, must jump boundary at some time point.When selecting several letters closest on boundary according to region
Mark node simultaneously builds boundary RSSI time serieses RBAfterwards, you can the variation tendency of RSSI is analyzed as unit of boundary.Do not considering
In the case of noise, unknown node obtains boundary time sequence R on boundaryBMiddle maximum value.
The determination of track and boundary intersection：2 dimensional planes being made of region are examined since each region is relatively small
Consider the track that wireless sensor network is used to track construction personnel or construction machinery, without loss of generality, it will be assumed that in adjacent region
Unknown node does near-sighted linear uniform motion in domain, seeks the intersection point on track and boundary.As event time point " t1t2t3t4" with " horizontal-
It is vertical-horizontal-vertical " order is when occurring (as shown in Figure 7), satisfaction
V indicates that the movement speed of unknown node, BC+CE=1 expressions do not consider that the actual area length of side, solving equations can in formula
It obtains C points and accounts for length of side percentage, to determine C points position according to region length of side q, can similarly acquire D points position.It can from equation group
See, solution procedure is unrelated with v, as long as can be solved using boundary crossover event time point.
Direction of motion inclination angle theta is
As event time point " t1t2t3t4" when sequentially occurring with " horizontal-vertical-vertical-horizontal " (as shown in Figure 8), satisfaction
Solving equations can obtain C points position, similar to acquire D points position.
At this point, direction of motion inclination angle theta is
Formula (30) and the meaning of formula (32) Unified Expression are, adjacent to cross over longitudinal boundary event time point difference and adjacent leap
Horizontal boundary event time point difference can determine the direction of motion.
Other boundary crossover event time point order correspond to solve intersection position method with it is upper identical.
Other than boundary crossover event time point " transverse and longitudinal " is alternately present situation, if unknown node is along parallel with boundary
Direction is moved, then continuous " transverse cross " event time point sequence or " vertical " event time point sequence occurs, at this time accurately
Intersection position is needed by traditional localization method determination, and boundary crossover event may determine that course bearing.
In the present invention, the meaning of Distributed Calculation is for centralization, and each unknown node builds boundary respectively
Time series detects boundary crossover event and determines corresponding time point；Each unknown node independence dynamic cutting time window, structure
Zone time window statistic infers current region according to zone time window statistic；Infer track and boundary intersection position；
Track is formed, aggregation node is as a result uploaded.
The method of the present invention builds space time correlation model realization track following for the latticed deployment beaconing nodes of indoor environment
Algorithm, regional determination and boundary crossover event are in time window according to the region RSSI statistics and boundary time sequence of reconstruct
Statistical inference obtains.Algorithm integrally judges statistics distribution and boundary time sequence variation trend in time window, can be effective
It reduces influence of noise and reduces the appearance of outlier using the local topology of wireless network, improve path accuracy, it is distributed
Algorithm takes full advantage of the computing capability of each unknown node.
The method of the present invention takes full advantage of the local spatial information that the topological structure of wireless sensor network is included, and will
Statistical inference is placed in time window to be carried out based on positioning time sequence, reduces noise and exceptional value as far as possible to result of calculation
It influences, path accuracy is improved relative to other methods；Distributed algorithm makes full use of the computing capability of unknown node, reduces
The traffic particularly avoids influencing each other between node, there is high robustness.
Claims (7)
1. a kind of distributed space time correlation model trace tracking method based on statistical inference, it is characterised in that including following step
Suddenly：
According to the equidistant latticed deployment beaconing nodes in longitudinal and transverse direction in located space, by beaconing nodes information { NodeID,
(x, y) } it is stored at each unknown node, in unknown node forming region information and boundary information, wherein NodeIDFor beacon section
Point ID number, (x, y) are its coordinate；
It receives unknown node and sends the beaconing nodes of notification information with fixed frequency transmitting positioning signal, unknown node receives simultaneously
It is respectively formed multiple time series R according to beaconing nodesi；
Each unknown node builds boundary time sequence respectively, detects boundary crossover event and determines corresponding time point；Each unknown section
The independent dynamic cutting time window of point, builds zone time window statistic, is inferred according to zone time window statistic current
Region；
Infer track and boundary intersection position；Track is formed, aggregation node is as a result uploaded.
2. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
In：The unknown node receives and is respectively formed multiple time series R according to beaconing nodesiFor：Unknown node will receive multiple
The positioning signal that independent beaconing nodes are sent, be divided into according to beaconing nodes it is multigroup, every group according to time order and function formation sequence, i.e.,
For time series.
3. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
In：The structure boundary time sequence is：As unit of beaconing nodes group, boundary RSSI time serieses are built as follows,
Rb=(rb1,rb2,rb3...rbk...rbs) (4)
Wherein, it includes beaconing nodes number, r that m, which is boundary,ikFor k-th of RSSI value of i-th of beaconing nodes, b is boundary number, rbk
For k-th of value of the boundary time sequence, k is the serial number in time series, and s is boundary time sequence length.
4. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
It is in the detection boundary crossover event：
Judge boundary time sequence in T2Inside whether there is primary more apparent ascending and descending process, if it is, boundary occurs
Across event；Maximum point in ascending and descending process indicates time point of the unknown node by boundary；Corresponding region time window
Mouth T1The period between event time point is crossed over for adjacent boundary.
5. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
It is realized by following algorithm according to zone time window statistic deduction current region：
argmax(W(S,T1)) (23)
Wherein, W (S, T1) it is time window T1For interior beaconing nodes using region as unit RSSI sizes, RSSI is received signal strength
Instruction, W (S, T1) it is a statistic, n is that region S includes beaconing nodes number, and i is beaconing nodes, and j is component in time series
Number, rijFor corresponding RSSI values, l is length of time series, l and time window size T1It is in direct ratio：
L=T1*f (3)
Wherein f is positioning signal tranmitting frequency；
arg max(W(S,T1)) indicate to find out statistic W (S, T1) maximum region.
6. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
It is realized by following formula with boundary intersection position in the deduction track：
Wherein, A is track and top horizontal boundary intersection point, and B is beaconing nodes, and C is track and longitudinal boundary intersection point, and D is track
With following horizontal boundary intersection point, E is lower left corner beaconing nodes, t1Event time point, t are corresponded to for A2Event time point is corresponded to for C,
t3Event time point is corresponded to for D, v is the movement speed of unknown node, and BC+CE=1 is indicated not considering actual area length of side q, be solved
Equation group can obtain C points and account for length of side percentage, to determine C points position according to region length of side q, can similarly acquire D points position；
Line behind the position C, D is acquired, the movement locus of the personnel for wearing unknown node or machinery is obtained.
7. the distributed space time correlation model trace tracking method as described in claim 1 based on statistical inference, feature exist
In：Space time correlation model based on statistical inference is defined as follows：
1) region：The plane subdivision that two dimensional surface is divided by four adjacent in length and breadth beaconing nodes, each lattice are one
A region, the lattice length of side are q；
2) boundary：The plane line of demarcation that two dimensional surface is made of the beaconing nodes of horizontal, every horizontal line/ordinate
For a line circle；
3) unknown node receives the positioning signal from each beaconing nodes and constitutes a RSSI according to the priority of arrival time
Time series is：
Ri=(ri1,ri2,ri3...rik...ril) (1)
In a period of time, the location information that each beaconing nodes are sent to current unknown node is formed according to the priority of arrival time
One Ri, i expression beaconing nodes ID numbers；Lower target component k is implicit to indicate relative time information, and l is length of time series；
4) space time correlation：Infer area information corresponding with time window, according to side according to area of space forming region statistic
Boundary forms boundary time Sequence Detection boundary crossover event；The included beaconing nodes number m in boundary is determined according to field condition, is selected
Selecting can most reflect that the beaconing nodes with unknown node relative position relation are calculated；
5) distributed computing method：For each unknown node in the location information of locally storage beaconing nodes, side is simultaneously detected in selected boundary
Event is crossed on boundary, divides time window according to boundary crossover event time click and sweep, in local computing range statistics amount and statistical inference
Corresponding region information in time window, and then trace information is calculated, the above each unknown node of calculating process independently carries out,
And result of calculation is uploaded into aggregation node；
6) time window：Distributed computing method each unknown node dynamic cutting in calculating process goes out two class time window T1With
Window T2, time window is each unknown node according to time series RiSituation of change independently divides ground, and each unknown node does not have each other
Have an impact, wherein T1For zone time window, T2For boundary time window.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610315568.1A CN106257301B (en) | 2016-05-12 | 2016-05-12 | Distributed space time correlation model trace tracking method based on statistical inference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610315568.1A CN106257301B (en) | 2016-05-12 | 2016-05-12 | Distributed space time correlation model trace tracking method based on statistical inference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106257301A CN106257301A (en) | 2016-12-28 |
| CN106257301B true CN106257301B (en) | 2018-08-28 |
Family
ID=57714216
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610315568.1A Expired - Fee Related CN106257301B (en) | 2016-05-12 | 2016-05-12 | Distributed space time correlation model trace tracking method based on statistical inference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106257301B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403477A (en) * | 2017-08-11 | 2017-11-28 | 北京启迪时代科技有限公司 | The method that discrepancy judgement is carried out using wireless signal Time-distribution |
| CN108924763B (en) * | 2018-07-16 | 2020-07-03 | 浙江大学 | Indoor area access amount continuous monitoring method facing probability type position stream data |
| CN111797174A (en) * | 2019-04-08 | 2020-10-20 | 华为技术有限公司 | Method and apparatus for managing spatiotemporal data |
| CN110263633B (en) * | 2019-05-13 | 2023-08-04 | 广州烽火众智数字技术有限公司 | Method, system and storage medium for detecting and early warning of toxic person based on space-time correlation |
| US10460581B1 (en) * | 2019-05-15 | 2019-10-29 | Cognitive Systems Corp. | Determining a confidence for a motion zone identified as a location of motion for motion detected by wireless signals |
| CN111159247B (en) * | 2019-11-20 | 2024-01-12 | 北京交通大学 | Regional traffic jam propagation track mining method |
| CN111784730B (en) * | 2020-07-01 | 2024-05-03 | 杭州海康威视数字技术股份有限公司 | Object tracking method and device, electronic equipment and storage medium |
| CN114530018B (en) * | 2022-04-24 | 2022-08-16 | 浙江华眼视觉科技有限公司 | Voice prompt method and device based on pickup trajectory analysis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103491627A (en) * | 2013-08-03 | 2014-01-01 | 东北大学 | Close range real-time accurate positioning method integrating multiple algorithms |
| CN104093204A (en) * | 2014-07-07 | 2014-10-08 | 无锡中科西北星科技有限公司 | RSSI area location method based on wireless sensor network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR100946984B1 (en) * | 2007-12-10 | 2010-03-10 | 한국전자통신연구원 | System and method for chasing location |
-
2016
- 2016-05-12 CN CN201610315568.1A patent/CN106257301B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103491627A (en) * | 2013-08-03 | 2014-01-01 | 东北大学 | Close range real-time accurate positioning method integrating multiple algorithms |
| CN104093204A (en) * | 2014-07-07 | 2014-10-08 | 无锡中科西北星科技有限公司 | RSSI area location method based on wireless sensor network |
Non-Patent Citations (2)
| Title |
|---|
| "Robust trajectory estimation for crowdsourcing based mobile application";zhagn Xinglin et al.;《IEEE Trans on Parallel and Distributed Systems》;20140731;第25卷(第7期);第1876-1885页 * |
| "基于无线传感器网络的移动机器人定位导航系统";杨秀萍 等;《电子器件》;20071231;第30卷(第6期);第2265-2268页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106257301A (en) | 2016-12-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106257301B (en) | Distributed space time correlation model trace tracking method based on statistical inference | |
| Park et al. | Self-corrective knowledge-based hybrid tracking system using BIM and multimodal sensors | |
| CN103605110B (en) | Based on the indoor passive object localization method of received signal strength | |
| CN110244715B (en) | Multi-mobile-robot high-precision cooperative tracking method based on ultra wide band technology | |
| Zhang et al. | A comprehensive study of bluetooth fingerprinting-based algorithms for localization | |
| CN103298156B (en) | Based on the passive multi-target detection tracking of wireless sensor network | |
| CN107168342B (en) | Pedestrian trajectory prediction method for robot path planning | |
| CN101179707A (en) | Wireless network video image multi-visual angle cooperation target tracking measurement method | |
| CN103281779B (en) | Based on the radio frequency tomography method base of Background learning | |
| Papaioannou et al. | Tracking people in highly dynamic industrial environments | |
| Wainer et al. | Machine learning-based indoor localization and occupancy estimation using 5G ultra-dense networks | |
| Jadidi et al. | Gaussian processes online observation classification for RSSI-based low-cost indoor positioning systems | |
| Zheng et al. | An adaptive framework coping with dynamic target speed for device-free passive localization | |
| CN104093204A (en) | RSSI area location method based on wireless sensor network | |
| Song et al. | Fingerprinting localization method based on toa and particle filtering for mines | |
| Aravinda et al. | Optimization of RSSI based indoor localization and tracking to monitor workers in a hazardous working zone using Machine Learning techniques | |
| CN104063884B (en) | The images steganalysis method being combined based on motion prediction with multi-template matching | |
| Yun et al. | Robust location tracking using a dual layer particle filter | |
| Song et al. | On the time to search for an intermittent signal source under a limited sensing range | |
| Boulanouar et al. | Pmt 2: a predictive mobile target tracking algorithm in wireless multimedia sensor networks | |
| Xu et al. | An improved MDS localisation algorithm for a WSN in a sub-surface mine | |
| Kanrar et al. | Pedestrian localisation in the typical indoor environments | |
| CN106658643A (en) | RSSI based valid anchor node selecting method | |
| Tarkanyi et al. | Roadside sensor optimization in infrastructure-based teleoperation of vehicles | |
| Dumitru et al. | Pedestrian analysis for crowd monitoring: The Milan case study (Italy) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180828 |