Dynamic self-adaptive positioning method for underwater wireless sensor network
Technical Field
The invention belongs to the technical field of underwater wireless sensor network positioning, in particular relates to a self-positioning method of an underwater wireless sensor network node applied to a complex marine environment, and particularly relates to a dynamic self-adaptive positioning method of an underwater wireless sensor network.
Background
An Underwater Wireless Sensor network (undersater Wireless Sensor Networks) is a revolution in the field of marine information perception and application, and large-scale random deployment in a complex marine environment is a typical application of the Underwater Wireless Sensor network for executing tasks such as information perception, collection, processing and transmission, such as sea battlefield situation perception, marine environment monitoring, Underwater target positioning and tracking and the like. Node self-positioning is a key supporting technology of an underwater sensor network and is a precondition for the application, because the sensed information usually needs to be accompanied by the position information of the node. In the self-positioning process, the node to be positioned needs to depend on corresponding reference information, such as coordinates of the reference node, a distance from the node to be positioned to the reference node, and the like. Ideally, it is generally assumed that the reference information, such as the anchor node location and the measured distance between nodes, is accurate. However, in practical applications, the reference information required for the positioning process is inevitably affected by the complex marine environment, and the occurrence of multiple sources of uncertain noise will seriously degrade the performance of the positioning system, specifically: the position deviation of the anchor node is limited by the performance of the anchor node and the environment interference, the ranging error is limited by the ranging capability of the node and caused by the multipath effect, the multi-hop accumulated error is influenced by the network condition of the reference link, and the like.
Most of the existing positioning algorithms assume that the position error and the ranging error of an anchor node obey normal distribution according to a central limit theorem, or assume that the position error and the ranging error obey ideal zero-mean Gaussian distribution directly, and then adopt a least square method idea to perform error processing so as to reduce the influence on positioning accuracy. The practical effectiveness of such methods in complex marine environment positioning is often less than ideal. The reasons are three: firstly, for the deployment environment of the underwater sensor network, it is difficult to acquire accurate distribution rules and characteristic parameters of all noises in advance, and it is unreasonable to directly assume that errors obey normal distribution and even standard normal distribution under the condition that prior-test information is insufficient; secondly, some scholars consider the problem of insufficient prior information and adopt a Monte Carlo method to carry out statistical inference on noise characteristic parameters, but the positioning algorithm for extracting noise characteristics on line based on the traditional statistical method needs to rely on a large number of effective measurement samples, and for the underwater sensor network nodes with severely limited resources, the cost of the method for obtaining a large number of samples through repeated measurement cannot be borne; third, both methods are based on the assumption that the multi-source noise follows the gaussian distribution, but actually, the multiple noises accompanying the complex environment localization process do not strictly follow the gaussian distribution
Therefore, it is necessary to explore a new method for self-positioning of an underwater wireless sensor network node, which can overcome adverse effects caused by inaccurate reference information under the condition of a small sample and has strong online cognition and dynamic self-adaption capability to a complex positioning environment, aiming at the problems of insufficient prior information, uncertain multisource noise distribution rule, insufficient number of effective ranging samples and the like in the positioning process.
Disclosure of Invention
The invention aims to provide a dynamic self-adaptive positioning method for an underwater wireless sensor network, which solves the problem of self-positioning of underwater wireless sensor network nodes under the influence of multi-source uncertain noise and with the simultaneous existence of anchor node position error and multi-hop distance estimation deviation; under the condition of a small sample, the influence of the position error of the anchor node on the positioning performance is reduced by carrying out online evaluation and calibration on the accuracy of the declared position coordinate of the anchor node, the influence of the multi-hop distance estimation deviation of the node on the positioning performance is reduced by the multi-hop distance estimation of the deviation adaptive correction, the online cognition and dynamic adaptive capacity of the node in a complex positioning environment is enhanced, and the positioning performance of the underwater wireless sensor network is improved.
The invention provides a dynamic self-adaptive positioning method for an underwater wireless sensor network, which specifically comprises the following steps:
firstly, an anchor node evaluates the accuracy of self-declared position coordinates on line;
(1) taking the declaration coordinates (or the initial coordinates calibrated when the nodes are deployed) of the initial moments as a reference, taking the Euclidean distance between the nth updated coordinates and the reference coordinates in the online evaluation time as a position deviation sample point, collecting n times (generally n is less than or equal to 5), and constructing an anchor node position deviation sample set;
(2) performing Bootstrap resampling on the position deviation sample set according to a nonparametric back sampling method;
(3) b Bootstrap samples with unchanged volume are extracted independently in succession; in order to meet the sampling requirement of the Bootstrap method and control the node calculation amount, 200 is taken as a common condition B;
(4) calculating the mean value of Bootstrap samples and solving the standard deviation
(5) Computing accuracy of anchor node declared location coordinates ηi:Wherein R is the communication radius of the node;
(6) calibrating reference levels of anchor nodes: calibrating the anchor node with the accuracy higher than a preset precision threshold value as a primary reference node; if the precision is lower than the preset precision threshold value, calibrating the anchor node as a common anchor node; the preset precision threshold value is usually a real number within a [0.95,1] interval;
step two, the nodes to be positioned construct a dynamic renewable positioning group;
(1) counting the number of primary reference nodes in a range of one-hop to multiple-hop threshold values by the node to be positioned, and when the number of the primary reference nodes reaches the minimum number of coordinate estimation, storing all the primary reference nodes and common intermediate nodes participating in information forwarding into a positioning group, and performing the third step; if the number is less than the minimum number of the coordinate estimation, the node is calibrated as the node which does not meet the positioning condition; the multi-hop threshold value is usually a natural number not greater than 5; the minimum number of coordinate estimates is 4 in the case of three-dimensional deployment and 3 in the case of two-dimensional deployment;
(2) when the position coordinate accuracy of the primary reference node is reduced to be lower than a preset accuracy threshold value, the primary reference node is calibrated to be a common anchor node again, and a positioning group is removed;
(3) the common node acquires position coordinates, the precision of the position coordinates is higher than a preset precision threshold value, and the common node is calibrated to be a secondary reference node; in a new positioning period, the nodes which do not meet the positioning conditions expand the statistical range to secondary reference nodes, when the sum of the number of the primary reference nodes and the number of the secondary reference nodes reaches the minimum number of coordinate estimation, a positioning group is constructed, and the third step is carried out; otherwise, continuing to wait for the next positioning period;
step three, the positioning group carries out multi-hop distance estimation of deviation self-adaptive correction;
(1) constructing an original ranging sample set between a node to be positioned and a reference node, and calculating the mean value of original ranging samples; typically the number of raw ranging samples is no greater than 5;
(2) b' Bootstrap resampling is carried out on the original ranging sample set successively and independently, and the mean value of the Bootstrap resampling is calculated; the resampling time B' is usually 200;
(3) checking for deviation of estimated values WhereinIs the mean of the jth Bootstrap sample,is the mean of the original ranging samples; if it isIndicating that the mean estimate is unbiased; if it isIndicating that the mean estimate is high; if it isIndicating that the mean estimate is low;
(4) calculating an estimate of the offset correction Will be provided withArranged from small to large to obtain
(5) Calculating the upper and lower bounds of the distance estimation after the self-adaptive correction, and solving the estimation of the multi-hop distance interval with the confidence level of 1- α, namely Q1Is composed ofThe integer part of (1), i.e.Get Q2Is composed ofExtracting to obtainAnda lower bound and an upper bound as estimates of the multi-hop distance interval;
step four, solving a compartmentalized position-tolerant space of the nodes to be positioned;
(1) extracting a class of constraint space formed by upper and lower boundaries of a distance estimation interval between a node to be positioned and an adjacent reference node;
(2) extracting a class II constraint space formed by multi-hop distance estimation between a node to be positioned and a multi-hop reference node; the upper bound of the second class of constraint space is the upper bound of the multi-hop distance estimation interval, the lower bound of the constraint space is the smaller one between the communication radius and the lower bound of the multi-hop distance estimation interval, and the extraction formula is as follows:
wherein S isja(x) For a node N to be positionedaAnd a multi-hop reference node NjR is the communication radius of the node,for the lower bound of the multi-hop distance estimation interval, | Xj-Xa||2Is NaAnd NjThe Euclidean distance between the two electrodes,estimating the upper bound of the interval for the multi-hop distance;
(3) replacing all the first class and second class constraint spaces in the positioning group with interval numbers; solving the intersection of all interval numbers, and taking the intersection as the compartmentalized position compatible space of the node to be positioned;
fifthly, estimating coordinates and evaluating precision of the node to be positioned;
(1) scanning the compartmentalized position-tolerant space of the node to be positioned, solving the central position of each interval number subset in the space, and taking the central position as a sample set of coordinates of the node to be positioned; the optimal point estimation of the coordinates of the node to be positioned is obtained, and the calculation formula is as follows:
wherein,for a node N to be positionedaThe optimal point estimate of the coordinates is estimated,for a sample in the sample set of coordinates of the node to be located, XiFor positioning groupInternal reference node NiCoordinate of d'aiEstimating median values for adaptively corrected distances ΩaIs a position-compatible space after compartmentalization;
(2) carrying out precision evaluation on the estimated coordinates of the nodes; calibrating the positioning precision higher than a preset precision threshold value to be a secondary reference node, and turning to the second step; otherwise, calibrating the node as the finished positioning node, and taking the estimated coordinate as the final coordinate; the formula for the accuracy assessment is:
wherein, ηaTo locate node NaEstimation accuracy of the estimated coordinates, ηkPositioning accuracy for positioning k reference nodes participating in positioning within a cluster, Xa=[xa,ya,za]TFor a node N to be positionedaEstimated coordinates of (2), Xi=[xi,yi,zi]TIn order to refer to the coordinates of the nodes,
the invention has the advantages that:
(1) the invention provides a dynamic self-adaptive positioning method of an underwater wireless sensor network, which reduces the adverse effect of anchor node position error on the coordinate estimation of a node to be positioned by on-line evaluation of the accuracy of anchor node statement coordinates; the Bootstrap small sample estimation method does not need to know the distribution of the sample space, and can estimate the related parameters only on the basis of a small sample number, thereby reducing the energy consumption of sample acquisition and improving the effectiveness of on-line evaluation.
(2) The invention provides a dynamic self-adaptive positioning method of an underwater wireless sensor network, which reduces the adverse effect of multi-hop distance estimation errors on positioning performance, improves the precision of coordinate estimation and enhances the dynamic self-adaptive capability of nodes in a complex positioning environment by carrying out deviation self-adaptive correction on multi-hop distance estimation.
(3) The invention provides a dynamic self-adaptive positioning method of an underwater wireless sensor network, which reduces the computational complexity and enhances the practicability of a system by performing interval processing on the position-compatible space of a node to be positioned.
Drawings
FIG. 1 is a flow chart of the implementation steps of a dynamic adaptive positioning method for an underwater wireless sensor network according to the present invention;
FIG. 2 is a diagram illustrating single-hop and multi-hop distance estimation performed by a position group according to the present invention;
FIG. 3 is a diagram of a node N to be positioned in the present inventionaThe position of (a) can accommodate a space schematic diagram;
FIG. 4 is a diagram of a counter node N in the present inventionaThe position of (2) can be used for carrying out interval processing in a space;
FIG. 5 is a schematic diagram of three-dimensional deployment of an underwater wireless sensor network according to the present invention;
fig. 6 is a comparison graph of average positioning errors of the new method and the conventional method under different network connectivity.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings. The invention discloses a dynamic self-adaptive positioning method of an underwater wireless sensor network, which is implemented by the following steps as shown in figure 1:
firstly, an anchor node evaluates the accuracy of self-declared position coordinates on line;
(1) anchor node NiWith declared coordinates (or calibrated initial coordinates at node deployment) X at the initial instant of each evaluation cyclei (0)=[xi (0),yi (0),zi (0) ,]TBased on the n-th updated coordinate X in the online evaluation time delta ti (n)With the initial coordinate Xi (0)Relative euclidean distance Xi (n)-Xi (0)||2As a positionally displaced sample pointCollecting n times, and establishing an anchor node position deviation sample setIn order to improve the timeliness of evaluation and control communication consumption, the number n of updating is as small as possible, and n is less than or equal to 5 in general; the evaluation time delta t is controlled within a short time period, which is generally equivalent to the accumulation of 3-5 positioning times; the sample set V thus obtainediIs a small sample set;
(2) for small sample set ViPerforming Bootstrap resampling: method for sampling from original sample by nonparametric back-samplingRandomly and repeatedly extracting Bootstrap samples with the capacity of n
(3) B Bootstrap samples with the capacity of n are extracted independently in succession to obtainB is 1,2, …, B; in order to meet the sampling requirement of a Bootstrap method and control the node calculation amount, the Bootstrap sample number B is usually 200;
(4) for the b-th Bootstrap sample, the mean value is calculated:b=1,2,…,B;
(5) estimating standard deviation of anchor node position deviation
(6) Computing anchor node position coordinate accuracy ηi:
(7) Calibrating reference levels of anchor nodes: if anchor node Niη (g)iIf the precision is greater than the preset precision threshold lambda, calibrating as a primary reference node; otherwise, calibrating as a common anchor node; the predetermined precision threshold λ is usually [0.95,1]]Real numbers within the interval;
step two, the nodes to be positioned construct a dynamic renewable positioning group;
(1) all anchor nodes broadcast a group of positioning packets which comprise information such as self IDs (identity) of the anchor nodes, coordinates, position coordinate accuracy, reference level identifiers and the like; all nodes broadcast a group of ranging packets, and information such as IDs (identity) and adjacent distances of all neighbor nodes in a communication range of the nodes is acquired through information interaction;
(2) node N to be positionedaCounting the number m of first-level reference nodes in the neighbor nodesa: such as maTo a minimum number of coordinate estimates, i.e. maIf the number of the nodes to be localized is more than or equal to 4, the nodes to be localized accommodate all the neighbor primary reference nodes into a localization group; e.g. m in a hop rangeaIf < 4, searching for the first-level reference nodes within the range of 2 hops to a multi-hop threshold zeta in sequence until maWhen the number of the reference nodes is more than or equal to 4, all the first-level reference nodes and the common intermediate nodes participating in information forwarding are accommodated into a positioning group; otherwise, the node is calibrated as a node which does not meet the positioning condition; the multi-hop threshold ζ is a natural number which is not more than 5 in a common case;
(3) for nodes which do not meet the positioning condition, along with the progress of the positioning process, if the common nodes in the multi-hop threshold zeta range acquire position coordinates and the precision is higher than the preset precision threshold lambda, the positioned nodes are calibrated as secondary reference nodes, and the number n of the secondary reference nodes is countedi(ii) a Such as ni+miIf the number of the primary reference nodes and the secondary reference nodes in the zeta hop number range is more than or equal to 4, all the primary reference nodes, the secondary reference nodes and the common intermediate nodes participating in information forwarding are accommodated in a positioning group;
(4) along with the proceeding of the positioning process, if the position coordinate accuracy of the first-stage reference node and the second-stage reference node in the positioning group is reduced to be lower than a preset accuracy threshold lambda, the reference node is calibrated to be a common anchor node or a common node again, and the positioning group is removed;
step three, the positioning group carries out multi-hop distance estimation of deviation self-adaptive correction;
node N to be positionedaAnd a reference node NiA schematic diagram for performing single-hop and multi-hop distance estimation is shown in fig. 2, wherein a solid square block in the schematic diagram is a node to be positioned, and a solid dot is a reference node;
(1) construction of a node N to be positionedaAnd a reference node NiOriginal ranging sample set ofAnd calculating the mean of the original ranging samplesThe number n of samples is a natural number not greater than 5;
(2) performing Bootstrap resampling on an original ranging sample set: from the original sampleExtracting Bootstrap sample with n capacity according to the method of putting back to sample
(3) M Bootstrap samples with the capacity of n are extracted independently in succession,wherein j is 1,2, …, M; for the jth Bootstrap sample, calculate its meanj ═ 1,2, …, M; in order to meet the sampling requirement of a Bootstrap method and control the node calculation amount, the Bootstrap sample number M is usually 200;
(4) checking for deviation of estimated values If it isIndicating that the mean estimate is unbiased; if it isIndicating that the mean estimate is high; if it isIndicating that the mean estimate is low;bias of multi-hop distance estimationTypically greater than 0;
(5) calculating an estimate of the offset correction Will be provided withArranged from small to large to obtain
(6) The distance after adaptive correction estimates the upper and lower bounds: get Q1Is composed ofThe integer part of (1), i.e.GetExtraction ofAs(ii) an estimate of (d);
(7) locating a node N to be located with an intra-group confidence level of 1- αaAnd a reference node NiInter-hop distance estimationIs finally determined as
Step four, solving a compartmentalized position-tolerant space of the nodes to be positioned;
(1) extracting the node N to be positionedaWith adjacent reference nodes NiThe distance between them estimates the class constraint space formed by the upper and lower bounds of the interval:
(2) extracting the node N to be positionedaAnd a multi-hop reference node NjEstimating a two-class constraint space formed by multi-hop distance between the two classes of constraint spaces; the upper bound of the second class of constraint space is the upper bound of the multi-hop distance estimation interval, the lower bound of the constraint space is the smaller one between the communication radius and the lower bound of the multi-hop distance estimation interval, and the extraction formula is as follows:
wherein R is0Is the communication radius of the node;
(3) obtaining a node N to be positionedaIntersection S of all first-class and second-class constraint spacesa(x) I.e. NaCan accommodate the space:
wherein k is the node N to be positionedaThe number of reference nodes in the positioning group as a core;
as shown in fig. 3, treatPositioning node NaAnd a reference node NiIs a neighbor node, the formed annular space is a class of constraint space, NaAnd a multi-hop reference node NjThe annular space is a two-class constraint space, and the shaded part is the intersection part of the two constraint spaces, namely NaThe position of (A) can contain space;
(4) compartmentalization of NaCan accommodate the space: node N to be positionedaA certain constraint space Sap(x) By number of intervalsInstead of, whereinIs the number of intervals thetaap IThe lower bound of (a) is,is the number of intervals thetaap IThe upper bound of (c); number of two intervals thetaap IAnd thetaaq IThe intersection of (a) and (b) is:Nathe position capacity space after the interval is as follows:
node N to be positionedaA top view of the compartmentalized, position-tolerant space is shown in fig. 4;
fifthly, estimating coordinates and evaluating precision of the node to be positioned;
(1) estimating coordinates of a node to be positioned;
the compartmentalized position can accommodate a space ΩaIs to locate all k θ within the groupai IIs geometrically a set of cubesSet, denoted as Ωa={Θa1,Θa2,…,Θan}; will each cube ΘanOf (2) centerAs a node N to be positionedaA sample of coordinates, the position of the centre being foundThe formula of (1) is:
node N to be positionedaCoordinate estimation ofThe following formula is used to obtain:
wherein, XiTo locate an intra-cluster reference node NiCoordinate of d'aiEstimating median values for adaptively corrected distances Solved node N to be positionedaOptimal point estimation of coordinatesComprises the following steps:
(2) to node NaIs estimated byAnd (3) line precision evaluation: node N to be positionedaIs estimated as Xa=[xa,ya,za]TThe coordinates of k primary reference nodes participating in the positioning process are Xi=[xi,yi,zi]T, i=1,2,…,k;NaPositioning accuracy ηaThe following formula is used to obtain:
wherein:
positioning accuracy η if there is a secondary reference node in the positioning group to participate in positioningaThe following formula is used to obtain:
η thereinkPositioning accuracy of k secondary reference nodes participating in positioning in a positioning group is obtained;
if the positioning accuracy is ηaIf the precision is larger than the preset precision threshold lambda, the positioned node N is determinedaMarking as a secondary reference node; otherwise, marking the node as a positioning completed node, node NaThe final estimated coordinates are
Examples
As shown in FIG. 5, 200 sensor nodes are randomly deployed in a three-dimensional space region of 300m × 300m × 100m, wherein the anchor node proportion is 10%, and the anchor node proportion is represented by a five-pointed starID is 1-20; the number of the nodes to be positioned is 180, the nodes are represented by solid circles, and the IDs of the nodes to be positioned are 21-200 respectively. All nodes have the distance measuring function, a dotted line between the nodes represents that the two nodes can carry out direct communication, and the length of the dotted line represents the Euclidean distance between the two nodes. The maximum distance measurement times n among the nodes is 5, and the position of the anchor node deviates from a sample ViThe Bootstrap resampling time B is 200, and the original ranging sample D between nodesiaThe number of resampling times M of (2) is 200. The preset node position evaluation accuracy threshold lambda is 0.97, and the confidence level of the multi-hop distance estimation is 0.95. The multi-hop threshold ζ for ranging and information forwarding within the positioning group is 4. The position error of the anchor node and the estimation error of the multi-hop distance are non-Gaussian noises which obey Rayleigh distribution and have standard deviation of 0.02 times of the actual distance. By adjusting the communication radius of the nodes, the network connectivity is sequentially increased from 4 to 13.
Under the network environment, node self-positioning is performed by respectively using the dynamic self-adaptive positioning method of the underwater wireless sensor network and the traditional DV-distance algorithm provided by the invention, and the average positioning error change condition of the sensor network under different network connectivity degrees is obtained as shown in FIG. 6. The solid line marked with the hollow square is the average positioning error of the dynamic self-adaptive positioning method, and the dotted line marked with the hollow circle is the average positioning error of the traditional positioning method. With the increase of the network connectivity, the average positioning error is firstly increased and then rapidly decreased, the variation trends of the two algorithms are basically the same, but the positioning error of the DV-distance algorithm is higher than that of the positioning method provided by the invention. The reason why the network connectivity is 4 is that only a few nodes can be positioned, and as the connectivity increases, part of multi-hop nodes participate in positioning, the number of nodes which can complete positioning increases, but the average positioning error increases in a small range due to the large multi-hop estimation error. With the increase of the network connectivity, the estimation error of the multi-hop distance becomes smaller, and the average positioning error gradually decreases. The average positioning error of the method is lower than that of a DV-distance algorithm by more than 20%, which shows that the inhibition capability of the method on the position error of the anchor node and the estimation error of the multi-hop distance is better than that of the traditional method.
It should be noted that the embodiment is only used for illustrating the technical solution of the present invention and not for limiting, and although the present invention is described in detail with reference to the preferred embodiment, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced equivalently without departing from the spirit and scope of the technical solution of the present invention.