CN102711158A - Method and device for directional diagnosis of wireless sensor network - Google Patents
Method and device for directional diagnosis of wireless sensor network Download PDFInfo
- Publication number
- CN102711158A CN102711158A CN2012101271035A CN201210127103A CN102711158A CN 102711158 A CN102711158 A CN 102711158A CN 2012101271035 A CN2012101271035 A CN 2012101271035A CN 201210127103 A CN201210127103 A CN 201210127103A CN 102711158 A CN102711158 A CN 102711158A
- Authority
- CN
- China
- Prior art keywords
- node
- module
- network
- wireless sensor
- sequence number
- 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.)
- Pending
Links
Images
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a method and a device for directional diagnosis of a wireless sensor network. The method includes recording conditions of each transmitted packet by a wireless sensor node and forming identifications at last; starting a node in a potential problem area to collect identifications by a base station node finding out some network abnormalities; building an initial inference model by obtained local topological graph; selecting the optimum detection scheme to further detect the network according to the built reference model, detailing the reference model further by each detection result; and generating a diagnosis report after increment detection is completed. Accuracy of reference is increased, diagnosis area is limited in the potential problem area, and accordingly, consumption of network diagnosis is effectively reduced.
Description
Technical field
The present invention relates to the wireless sensor network technology field, relate in particular to a kind of directed diagnostic method and device of wireless sensor network.
Background technology
Though have many research work to make the connectivity and functional of wireless sensor network all be improved at present, still have frequent faults such as node failure, link instability and external environmental interference in the wireless sensor network.Wireless sensor network is in case after disposing; Because the characteristics of its self-organizing mode; Network internal ruuning situation is transparent to the keeper; This has caused difficulty to network relevant abnormality detection and location, and therefore designing effective inline diagnosis instrument, to help that network manager's monitor network running status and maintenance sensor system move be very significant.
Diagnosis algorithm in the existing wireless sensor network has all comprised two independent parts: collect running state information and abnormal network cause and infer.For example the someone has proposed a kind of active diagnostic techniques, and base-station node is periodically collected running state information from each sensor node and utilized decision-tree model to carry out cause inferred then.The somebody has proposed a kind of energy information that utilizes the external energy survey tool to collect all nodes, uses grader to judge those not the inside health status of responsive node and possible reasons then.Can find out that from above these methods have a common defective, their infonnation collection process is static predefined process, does not utilize any operating information.Clearly, predefined infonnation collection process may cause two bad results, and too much information gathering has brought extra communication-cost to network, and very few information gathering makes cause inferred produce too much false judgment.
Detection is a kind of diagnostic method that obtains the network internal state very commonly used in Internet and enterprise network.Many different detection methods in many different application scenes, but their great majority all need depend on expertise or information network topology for example in advance, and also they suppose that also network is static when operation.Compare the characteristics of wireless sensor network some self: the computational resource and the finite energy of (1) sensor node with Internet and enterprise network; (2) because external factor such as environment are disturbed and the unsteadiness of wireless transmission, the network topology dynamic change.And because self-organizing, can't obtain about the prior information of topology.As shown in Figure 1, the upper left corner is that the physics of GreenOrbs forest detection system is disposed structure chart, and other three width of cloth figure are different network topology snapshots constantly, can see that the different connection status of network constantly are different; (3) because sensor node is unusual fallibility; The interstitial content that lost efficacy simultaneously is difficult to prediction in advance; But in enterprise network, usually suppose the constant k that maximum wrong node numbers that occur simultaneously are a predefined; Therefore can find out that the diagnosis algorithm in more existing Internet and the enterprise network can not directly apply to wireless sensor network, because there is not the dynamic change of topology in prior information and the network operation process about the network internal operation.
Summary of the invention
The objective of the invention is to propose a kind of directed diagnostic method and device of wireless sensor network; The process of information gathering is under the guidance of inference pattern, to carry out; Along with information gathering is constantly perfect, inference pattern also constantly obtains refinement, thereby improves the accuracy rate of reasoning.
For reaching this purpose, the present invention adopts following technical scheme:
A kind of directed diagnostic method of wireless sensor network may further comprise the steps:
The nodal information of the forwarding bag of A, record node, the forwarding list that generates each node is tabulated with following the trail of;
The forwarding list of B, collection network abnormal area sensor node is set up local topology;
C, set up the inference pattern to show the network dependence based on local topology;
D, model carries out increment and surveys by inference, upgrades inference pattern according to result of detection, up to obtaining diagnosis report.
The nodal information that writes down in the steps A comprises
The node ID of previous dive and the number of transmitting bag are recorded in the forwarding list;
Transmit the source node ID of bag and the sequence number of said source node, be recorded in and follow the trail of in the tabulation.
The forwarding list that generates each sensor node in the steps A further may further comprise the steps with following the trail of when tabulating:
A1, sensor node are received and are transmitted bag, whether have the list item of previous dive node ID in the inspection forwarding list; If do not exist, then increase the list item of said previous dive node ID, and transmits the number that wraps and be set to 1; If exist, then the forwarding bag number in the said previous dive node ID list item is increased by 1;
The list item that whether has the source node ID that transmits bag in A2, the audit trail tabulation; If do not exist, then set up the list item of said source node ID, and the sequence number of source node is set to the sequence number of said forwarding bag; If exist; Then more said source node sequence number and the size of transmitting packet number when transmitting packet number greater than source node sequence number, are updated to said forwarding packet number with said source node sequence number; When transmitting packet number, do not do any operation less than source node sequence number.
Step B further may further comprise the steps:
B1, base-station node take out the list item of node under a cloud from follow the trail of tabulation, be packaged into a detection packet and be broadcast to network;
After B2, sensor node receive detection packet; Whether comprise source node ID under a cloud in the detection packet in the audit trail tabulation; If comprise, the size of the sequence number of then more said source node ID in following the trail of tabulation and in detection packet is when the sequence number in the detection packet during greater than the sequence number followed the trail of in the tabulation; Said sensor node sends to base-station node with forwarding list, and detection packet is sent to neighbor node;
B3, base-station node are set up local topology according to the forwarding list of receiving, said local topology is made amendment, and the leaf node in the amended local topology includes only base-station node and node under a cloud.
Among the step C, generate inference pattern, and confirm the probability distribution on each summit, instruct subsequent probe route selection and cause inferred with wherein the antenna vertex and the posterior probability on perception data summit based on bayesian network structure according to local topology.
Step D further may further comprise the steps:
D1, will have symptom input inference pattern now, produce the set of candidate's mistake according to the posterior probability of reason node;
The confidence level of D2, calculated candidate mistake set when confidence level during greater than predetermined threshold value, is diagnosis report output with the cooperation of current candidate's Error Set; When confidence level during, get into step D3 less than predetermined threshold value;
D3, will cover with said candidate's mistake set associative do not survey the maximum path of symptom number as detective path; Utilization comprises the detection packet of minimum suspection interstitial content to be surveyed, and feeds back to inference pattern and returns step D1 surveying the network information of collecting.
A kind of directed diagnostic device of wireless sensor network comprises the node tracing module, the mark collection module; Model reasoning module and increment detecting module, said node tracing module, mark collection module; The model reasoning module is connected with the increment detecting module successively, wherein
The node tracing module is used to write down the nodal information of the forwarding bag of node, the forwarding list that generates each node with follow the trail of tabulation;
The mark collection module is used for the forwarding list of collection network abnormal area sensor node, sets up local topology;
The model reasoning module is used for setting up the inference pattern that can show the network dependence based on local topology;
The increment detecting module is used for by inference model and carries out increment and survey, and upgrades inference pattern according to result of detection, up to obtaining diagnosis report.
The nodal information of said node tracing module record comprises:
The node ID of previous dive and the number of transmitting bag are recorded in the forwarding list;
Transmit the source node ID of bag and the sequence number of said source node, be recorded in and follow the trail of in the tabulation.
Said model reasoning module; Generate inference pattern according to local topology based on bayesian network structure; And confirm the probability distribution on each summit, instruct subsequent probe route selection and cause inferred with wherein the antenna vertex and the posterior probability on perception data summit.
Said increment detecting module further comprises the false inference module, and mistake set evaluation module is chosen module with the detection route, wherein,
The false inference module is used for existing symptom input inference pattern, according to the posterior probability generation candidate mistake set of reason node;
Mistake set evaluation module is used for the confidence level of calculated candidate mistake set, when confidence level during greater than predetermined threshold value, is diagnosis report output with the cooperation of current candidate's Error Set;
Survey route and choose module; Be used for when confidence level during less than predetermined threshold value; With covered with said candidate's mistake set associative do not survey the maximum path of symptom number as detective path; Utilization comprises the detection packet of minimum suspection interstitial content to be surveyed, and feeds back to the model reasoning module with surveying the network information of collecting.
Adopt technical scheme of the present invention, the process of information gathering is under the guidance of inference pattern, to carry out, and along with information gathering is constantly perfect, inference pattern also constantly obtains refinement, thereby improves the accuracy rate of reasoning.The diagnostic region of this method is confined to potential problem area, thereby has reduced the consumption of network diagnosis effectively, a kind of very effective online wireless sensor network diagnosis method.
Description of drawings
Fig. 1 is that the physics of existing forest detection system is disposed structural representation.
Fig. 2 is the schematic flow sheet of the directed diagnostic method of the wireless sensor network that provides of the specific embodiment of the invention.
Fig. 3 is a forwarding list and the structural representation of following the trail of tabulation in the specific embodiment of the invention.
Fig. 4 is the construction process sketch map of local topology in the specific embodiment of the invention.
Fig. 5 is the bayesian network structure sketch map that uses in the specific embodiment of the invention.
Fig. 6 is the structural representation of the directed diagnostic device of the wireless sensor network that provides of the specific embodiment of the invention.
Embodiment
Further specify technical scheme of the present invention below in conjunction with accompanying drawing and through embodiment.
The main thought of the specific embodiment of the invention is, the node tracing module of each wireless sensor node is used for writing down the situation of the bag of each forwarding, forms the mark of oneself at last.At the base-station node end, when finding that some network is unusual, base-station node will start the mark collection module and the node of potential problem area is carried out mark collect.After mark is collected; Can obtain local topology figure and be used for setting up preliminary inference pattern, according to the inference pattern of setting up, each detecting strategy of optimum of selecting is further surveyed network; Each result of detection is used for further refinement inference pattern; When increment is surveyed end, can generate a diagnosis report, wherein comprise the posteriority error probability of network element.As shown in Figure 1, this orientation diagnostic method may further comprise the steps:
Step S201, the nodal information of the forwarding bag of record node, the forwarding list that generates each node is tabulated with following the trail of.
Node tracing algorithm in this step has been safeguarded two tabulations at each sensor node, and is as shown in Figure 3, and one is forwarding list, is used for writing down last ID and the number of the bag that sends; Another one is to follow the trail of tabulation, is used for writing down the source ID of bag and up-to-date sequence number thereof, and is as shown in Figure 3.In this tracing scheme, every row of forwarding list and tracking tabulation only is 6 bytes, and maximum line numbers of forwarding list are its neighbours' number, are all interstitial contents and follow the trail of the maximum line numbers of tabulation.When the network operation, each wraps from the source node to the base-station node, and each intermediate node all can write down the situation of transmitting bag.The number of assumed wireless sensor network is N; Forwarding list can not surpass the 6N byte respectively with following the trail of to tabulate so; So maximum memory spaces of a node can not surpass the 12N byte, suppose N=1024, the maximum storage consumption of each node can not surpass 12KB so.And in fact forwarding list is far smaller than N with the line number of following the trail of tabulation, so the scheme of this data structure should be acceptable for wireless senser.
Following mask body place of matchmakers states the particular content of node tracing algorithm:
When node is received 1 bag, will at first check forwarding list.If there is not the corresponding list item of previous dive ID, will set up a new row so, and its number of transmitting bag is set to 1.If exist, so just its number of transmitting bag is increased by 1.Continue audit trail tabulation then,, will set up the list item of a new source ID so, and the sequence number that said bag is set is its sequence number if there is not the corresponding list item of source ID.If there is corresponding list item, so with the size of comparative sequences number.If the sequence number of said bag will upgrade sequence number so, otherwise will not do any operation greater than the sequence number in the list item of source ID correspondence.In this process, base-station node is also participated in this tracing process.But distinguish be, what base-station node write down is minimum sequence number, and other nodes records is maximum sequence number.When once diagnosis finishes, thereby the sequence number of those wrong nodes of resetting is begun a new time interocclusal record.
Step S202, the forwarding list of collection network abnormal area sensor node is set up local topology.
When network takes place unusually, base-station node will start the mark collection module potential problem area will be carried out information gathering.Forwarding list through the node of potential problem area is replied just can obtain local topology.Because the process of information gathering is carried out as required; Therefore can in time reflect the dynamic change of network; Like this be designed with two advantages: one is diagnostic region only in the part, is when network is not unusual when taking place more in addition, can not produce any extra diagnostic traffic expense.
At base-station node, the mark collection process is triggered by some networks is unusual.Standard that is widely used is in certain time period, when the perception data of collecting from node does not reach the number of expection, thinks that having some network errors takes place.During this time, base-station node will take out the list item of those nodes under a cloud from follow the trail of tabulation, then it is packaged into a detection packet and is broadcast to network.This flooding algorithm is different from common flooding algorithm, and this is because when node is received detection packet after, if in the forwarding list of oneself for to comprise any suspection node, go off the air so, otherwise, just can detection packet be proceeded to broadcast.Use the mode of this limited broadcast, drop to its extra communication-cost that brings as far as possible little.
Following mask body place of matchmakers states flooding algorithm: when receiving a detection packet, each node need check oneself whether to be to suspect that node transmitted bag.At first whether comprise in the tracking tabulation of inspection oneself and suspect ID,, will compare its sequence number size so if having.If the sequence number in the detection packet is bigger, node need be submitted the forwarding list of oneself to base-station node so, continues then according to forwarding list said detection packet to be issued own neighbours.If the sequence number in the detection packet is smaller, can only explains before diagnosis time started period and transmit data, so needn't be included in this diagnostic procedure for node under a cloud.The forwarding list that is collected into will be used for setting up inference pattern, because wherein comprised the degree of dependence relation between the node.
After having collected needed information, base-station node will begin to construct local topology and diagnose.Process of reconstruction begins from base-station node, and the connection in all neighbor lists of collecting is all combined.Clearly, having some irrelevant nodes has like this been included with being connected.Therefore, also need carry out the process of a beta pruning.According to the needs of diagnosis, can obtain such conclusion: having only base-station node and node under a cloud just possibly be leaf node.Therefore the beta pruning process is exactly that constantly to remove those degree be 1 all have been removed up to all underproof leaf nodes point and related connection the thereof in circulation.
The local topology reconstruction process is as shown in Figure 4.Among Fig. 4 (a); Node d and node k are nodes under a cloud; Base-station node beginning label collection process then, the neighbor list of replying from the potential problems node obtains local topology figure such as Fig. 4 (b), and to remove those degree be 1 and be not the node and the connection thereof of node under a cloud in circulation then.Can arrive local topology figure such as Fig. 4 (c) at last.
Step S203 sets up the inference pattern that can show the network dependence based on local topology.
The dependence of network internal is an internal correlation.The breaking down suddenly of certain father node will have to seek new path with the child node that causes it and transmit the data of oneself, if seek less than new path, the data of this node can't successfully be sent to base-station node so.The another one example is if there is mulitpath all to have higher packet loss simultaneously, possibly exist certain shared link to take place unusually by this mulitpath so.Therefore based on the local topology model, can set up an inference pattern and can the dependence of network be expressed, node state for example, Link State and perceptional function etc.Other similar data acquisition amounts wait the input of outside symptom as model inadequately, when observing some symptom, just can carry out reasoning to model and find out most possible explanation reasons then.
Because the height dynamic change of topology when not having priori and operation for network topology, the inference pattern in most of Internet and the static enterprise network is also inapplicable in wireless sensor network.Here, used a kind of multi-level inference pattern to express the internal relations of wireless sensor network.
A Bayesian network is a directed acyclic graph, vertex representation stochastic variable wherein, and the causality between the variable is represented on the limit.Directed edge from A to B representes that B is the result of A, and A is a reason.All there is its probability distribution on each summit, and difference is that the point that does not have father node has only prior probability, become the reason summit to these points, and other intermediate node existence condition probability probability distribution is called the symptom summit.Given certain evidence has wherein comprised the state of some symptoms, and Bayesian network just can be derived the posterior probability on all reason summits so.
The foundation of Bayesian network generates according to local topology.A typical example is as shown in Figure 5.In the node topology structure chart of Fig. 5 (a), represented that a base-station node 0 has connected two ordinary nodes.Directed edge is represented wireless transmission link.Have six kinds of summits in the Bayesian network shown in Fig. 5 (b), wherein R is an antenna vertex, comprises R0, R1 and R2; S is the perception data summit, comprises S1 and S2; D is the data report summit, comprises D1 and D2; C comprises C1 and C2 for connecting the summit; P is a path vertices, comprises P
20And P
210L is a link vertex, comprises L
20, L
10And L
21Each summit exists UP and DOWN two states.The implication on these summits below makes introductions all round.Antenna vertex is represented the antenna function of sensor node, mainly influences the state of link.The perceptional function of perception data vertex representation sensor node connects the connectivity of certain sensor node of vertex representation to base-station node.Path vertices is represented the particular course of certain bar to base-station node.Link vertex is represented two communications status between the node.The data report vertex representation data cases of certain node of receiving of base-station node.Give an example, if the perception data amount that base-station node is observed certain node certain period does not reach expection or has very large delay, data report is named a person for a particular job and is set to DOWN so, triggers a diagnostic procedure then.The data report summit depends on the perception data summit and is connected the summit.The state of link vertex then depends on the antenna vertex state of two nodes.And the connection summit, the relation between path vertices and the link vertex is then obtained from network topology structure.According to above rule, can obtain the structure chart of a Bayesian network, shown in Fig. 5 (b).In six kinds of stochastic variables, antenna vertex and perception data summit are the reason summits, just need infer during their state.And the connection summit, the state of link vertex and path vertices is the symptom summit, their state is that the result according to detection packet carries out assignment.The state on data report summit is the actual conditions assignment of base-station node according to data.Thereby the structure of this Bayesian network can be upgraded the dynamic change of reflection network timely according to the carrying out of node tracking module.
When the Bayesian network structure finished, next step important process was confirmed probability distribution for each summit exactly.The prior probability distribution on antenna vertex and perception data summit rule of thumb is worth definite.The conditional probability distribution on all the other symptom summits can be based on noise-OR door or Select door.In a noise-OR door, represent that any one father node is in the state of DOWN, the state of child node just will be in DOWN.In inference pattern, connect the influence of summit and perception data summit for the data report summit, link vertex be connected the summit and can represent with the noise-OR door for the influence of path vertices.Relation between the summit of mulitpath and the connectivity summit can represent that because any paths is in the UP state, the connectivity summit all will be in the UP state with the Select door.The output of inference pattern is exactly that the posterior probability on antenna vertex and perception data summit is used for instructing follow-up detection route selection and cause inferred.
Step S204, model carries out the increment detection by inference, upgrades inference pattern according to result of detection, up to obtaining diagnosis report.
There is not the process of a feedback in present most diagnosis policy for the collection of diagnostic message.In the method, inference pattern begins to start when network takes place unusually, but owing to the external symptom that has just begun to observe is not enough, obtains those might produce the maximum symptom that helps to reasoning state with regard to the mode of having taked increment type to survey.Each survey choose all by inference that model obtains, the result who surveys then is used for upgrading inference pattern. this method can be located a plurality of mistakes in the wireless sensor network effectively simultaneously.
In case inference pattern is set up and to be finished, and just can begin the increment type detection process, this process is under the guidance of model by inference, to carry out.Mainly being of this method if the symptom node of existing observation also is not enough to explain present problem, with select the optimum detection packet of next bar come to those not cardinal symptom survey.Mainly comprise three main modular: the false inference module, mistake set evaluation module and detection route are chosen module.As shown in Figure 6, the false inference module produces candidate's mistake set according to up-to-date posterior probability then with existing symptom input model.The error evaluation module is calculated candidate collection and is judged whether it meets the requirements afterwards; If satisfy; So present candidate's mistake is gathered just as final result, otherwise those are to survey symptom with regard to the input as detection route selection module for candidate collection contribution maximum.Select the detection route of an optimum that network is carried out information gathering at last, result of detection begins the reasoning of a new round again as the input of false inference module, produces or all symptom summits are all finished by detection up to the set of candidate's mistake of meeting the demands.Introduce false inference below respectively, the mistake set is estimated and is surveyed route and chooses three steps:
(1) false inference
The purpose of false inference is to find out the reason of observing symptom, and the input of false inference mainly comprises the state of the network element that the data report state that obtains from base-station node and result of detection obtain.When the state that gets access to new symptom, the false inference module will be upgraded the posterior probability of reason node then the state of all known evidences as input.Use a probability threshold value whether to become a mistake then as passing judgment on certain reason.After up-to-date posterior probability surpasses this threshold value, just put it into the set of candidate's mistake, this be a simplification be used for seek those for the method for observing the maximum reason of symptom contribution.Here it should be noted that if initial threshold is provided with too highly, may become candidate's mistake in neither one reason summit.In this case, can simply choose the maximum reason of posterior probability as initial error.A bit it should be explicitly made clear at this point in addition, do not exist between two posterior probability of reason node and contact directly, because the obtaining to reduce also and possibly improve this error probability of new symptom state.Along with network state is constantly surveyed, more symptom summit is determined just representes to approach more truth.
(2) the mistake set is estimated
The candidate's mistake set that is produced by the false inference module possibly be inaccurate, because the state on symptom summit is incomplete, the error evaluation module is estimated the confidence level of candidate's mistake set just.
Defined a confidence values that the confidence level measure equation comes the calculated candidate mistake to gather.Suppose that candidate's mistake set has comprised some mistakes, is expressed as C
f={ f
1, f
2... }, and f
iRelevant mistake set is expressed as
Therefore can be expressed as with a relevant symptom set of candidate's mistake set
Use S
0The symptom set of representing to have observed.So, can define the computing formula of the confidence values of candidate's mistake set:
In inference pattern; Even all symptoms have all been observed; The confidence values of candidate's mistake set still maybe be less than 1; Even because the set of candidate's mistake is exactly a truth, the state of some symptom is impossible, therefore needs to set a confidence level threshold value and judge whether the set of candidate's mistake is acceptable.Being provided with of this value needs the long-term observation and the accumulation of experience, if but threshold value is too high, and so correct mistake set possibly not satisfied yet; And threshold value is low excessively; This can comprise too much false error mistake, and this formula is used for the evaluate candidate set, if the result surpasses threshold value; Detection process will finish so, and the set of candidate's mistake will be exported as diagnosis report.
(3) surveying route chooses
If the confidence values of candidate's mistake set is less than threshold value, the detection route that will choose an optimum so obtains the state information of network.Verified in many previous work, the problem of choosing that minimizes the detection route of surveying expense is equal to the minimal set covering problem, is a np complete problem.Therefore, propose a heuristic greedy algorithm and solve this problem.Choose those and cover detection route with maximum numbers of not observing symptom of candidate's mistake set associative as optimal route.In the identical detection packet of this condition, select those minimal symptoms numbers again, reduce to survey expense with this.
The increment exploration policy has two significant advantages, the first, diagnostic message obtain the inference pattern that is based on continual renovation; The second, the choosing of detection packet is based on the inference pattern after the renewal, so communication-cost is minimized, because can not produce useless detection packet.Node pursive strategy can be in the less expense of assurance with in time of delay simultaneously, successful recovery local topology.
Adopt the specific embodiment of the invention to dispose the tentative sensing net system of about 300 sensor nodes.Information such as this system can collecting temperature, humidity, intensity of illumination are for the forest environment monitoring provides important information.Directed diagnosis algorithm is specifically realized in this forest monitoring system.We have carried out the experiment greater than 3 months, and having chosen an interval censored data of 6 days then is benchmark, has carried out the experiment based on trace.This interval censored data of 6 days comprises 42228 packets and 309 sensor nodes altogether.Trace experiment mainly comprises 3 parts, at first, in the stage very first time perception data is accepted situation and counts, if some node is not submitted the perception data of q.s in this stage, classifies them as the suspection node.Start then and survey diagnostic procedure, this is second time phase.Utilize the neighbor list situation of collecting to obtain the local topology information that those have participated in stage very first time suspection node.So detection process just is converted into inquiry and whether has corresponding path at second time phase.According to each result of detection, last inference pattern obtains final diagnosis report.In the form below, listed the diagnosis report of 6 days certain two minizones in the interval, in diagnosis report 1,, located 12 perception failure nodes and 7 transmission failure nodes through 21 detection packet.In diagnosis report 2,, 15 perception failure nodes and 9 transmission failure nodes have been located through 25 detection packet.
The structural representation of the directed diagnostic device of the wireless sensor network that Fig. 6 provides for the specific embodiment of the invention; This device comprises node tracing module 601, mark collection module 602, and model reasoning module 603 and increment detecting module 604, said node tracing module 601, mark collection module 602, model reasoning module 603 is connected with increment detecting module 604 successively, wherein,
The nodal information of said node tracing module 601 records comprises:
The node ID of previous dive and the number of transmitting bag are recorded in the forwarding list;
Transmit the source node ID of bag and the sequence number of said source node, be recorded in and follow the trail of in the tabulation.
Said model reasoning module 603; Generate inference pattern according to local topology based on bayesian network structure; And confirm the probability distribution on each summit, instruct subsequent probe route selection and cause inferred with wherein the antenna vertex and the posterior probability on perception data summit.
Said increment detecting module 604 further comprises the false inference module, and mistake set evaluation module is chosen module with the detection route, wherein,
The false inference module is used for existing symptom input inference pattern, according to the posterior probability generation candidate mistake set of reason node;
Mistake set evaluation module is used for the confidence level of calculated candidate mistake set, when confidence level during greater than predetermined threshold value, is diagnosis report output with the cooperation of current candidate's Error Set;
Survey route and choose module; Be used for when confidence level during less than predetermined threshold value; With covered with said candidate's mistake set associative do not survey the maximum path of symptom number as detective path; Utilization comprises the detection packet of minimum suspection interstitial content to be surveyed, and feeds back to model reasoning module 603 with surveying the network information of collecting.
Adopt technical scheme of the present invention, the process of information gathering is under the guidance of inference pattern, to carry out, and along with information gathering is constantly perfect, inference pattern also constantly obtains refinement, thereby improves the accuracy rate of reasoning.The diagnostic region of this method is confined to potential problem area, thereby has reduced the consumption of network diagnosis effectively, a kind of very effective online wireless sensor network diagnosis method.
The above; Be merely the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, anyly is familiar with this technological people in the technical scope that the present invention disclosed; The variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection range of claim.
Claims (10)
1. the directed diagnostic method of a wireless sensor network is characterized in that, may further comprise the steps:
A, record node are transmitted the nodal information of bag, generate the forwarding list and tracking tabulation of each node;
The forwarding list of B, collection network abnormal area sensor node is set up local topology;
C, set up the inference pattern to show the network dependence based on local topology;
D, model carries out increment and surveys by inference, upgrades inference pattern according to result of detection, up to obtaining diagnosis report.
2. the directed diagnostic method of wireless sensor network according to claim 1 is characterized in that, the nodal information that writes down in the steps A comprises
The node ID of previous dive and the number of transmitting bag are recorded in the forwarding list;
Transmit the source node ID of bag and the sequence number of said source node, be recorded in and follow the trail of in the tabulation.
3. the directed diagnostic method of wireless sensor network according to claim 1 and 2 is characterized in that, the forwarding list that generates each sensor node in the steps A further may further comprise the steps with following the trail of when tabulating:
A1, sensor node are received and are transmitted bag, whether have the list item of previous dive node ID in the inspection forwarding list; If do not exist, then increase the list item of said previous dive node ID, and transmits the number that wraps and be set to 1; If exist, then the forwarding bag number in the said previous dive node ID list item is increased by 1;
The list item that whether has the source node ID that transmits bag in A2, the audit trail tabulation; If do not exist, then set up the list item of said source node ID, and the sequence number of source node is set to the sequence number of said forwarding bag; If exist; Then more said source node sequence number and the size of transmitting packet number when transmitting packet number greater than source node sequence number, are updated to said forwarding packet number with said source node sequence number; When transmitting packet number, do not do any operation less than source node sequence number.
4. the directed diagnostic method of wireless sensor network according to claim 1 is characterized in that step B further may further comprise the steps:
B1, base-station node take out the list item of node under a cloud from follow the trail of tabulation, be packaged into a detection packet and be broadcast to network;
After B2, sensor node receive detection packet; Whether comprise source node ID under a cloud in the detection packet in the audit trail tabulation; If comprise, the size of the sequence number of then more said source node ID in following the trail of tabulation and in detection packet is when the sequence number in the detection packet during greater than the sequence number followed the trail of in the tabulation; Said sensor node sends to base-station node with forwarding list, and detection packet is sent to neighbor node;
B3, base-station node are set up local topology according to the forwarding list of receiving, said local topology is made amendment, and the leaf node in the amended local topology includes only base-station node and node under a cloud.
5. the directed diagnostic method of wireless sensor network according to claim 1; It is characterized in that; Among the step C; Generate inference pattern according to local topology, and confirm the probability distribution on each summit, instruct subsequent probe route selection and cause inferred with wherein the antenna vertex and the posterior probability on perception data summit based on bayesian network structure.
6. the directed diagnostic method of wireless sensor network according to claim 1 is characterized in that step D further may further comprise the steps:
D1, will have symptom input inference pattern now, produce the set of candidate's mistake according to the posterior probability of reason node;
The confidence level of D2, calculated candidate mistake set when confidence level during greater than predetermined threshold value, is diagnosis report output with the cooperation of current candidate's Error Set; When confidence level during, get into step D3 less than predetermined threshold value;
D3, will cover with said candidate's mistake set associative do not survey the maximum path of symptom number as detective path; Utilization comprises the detection packet of minimum suspection interstitial content to be surveyed, and feeds back to inference pattern and returns step D1 surveying the network information of collecting.
7. the directed diagnostic device of a wireless sensor network is characterized in that, comprises the node tracing module, the mark collection module; Model reasoning module and increment detecting module, said node tracing module, mark collection module; The model reasoning module is connected with the increment detecting module successively, wherein
The node tracing module is used to write down the nodal information of the forwarding bag of node, the forwarding list that generates each node with follow the trail of tabulation;
The mark collection module is used for the forwarding list of collection network abnormal area sensor node, sets up local topology;
The model reasoning module is used for setting up the inference pattern that can show the network dependence based on local topology;
The increment detecting module is used for by inference model and carries out increment and survey, and upgrades inference pattern according to result of detection, up to obtaining diagnosis report.
8. the directed diagnostic device of wireless sensor network according to claim 7 is characterized in that, the nodal information of said node tracing module record comprises:
The node ID of previous dive and the number of transmitting bag are recorded in the forwarding list;
Transmit the source node ID of bag and the sequence number of said source node, be recorded in and follow the trail of in the tabulation.
9. the directed diagnostic device of wireless sensor network according to claim 7; It is characterized in that; Said model reasoning module; Generate inference pattern according to local topology, and confirm the probability distribution on each summit, instruct subsequent probe route selection and cause inferred with wherein the antenna vertex and the posterior probability on perception data summit based on bayesian network structure.
10. the directed diagnostic device of wireless sensor network according to claim 7 is characterized in that, said increment detecting module further comprises the false inference module, and mistake set evaluation module is chosen module with the detection route, wherein,
The false inference module is used for existing symptom input inference pattern, according to the posterior probability generation candidate mistake set of reason node;
Mistake set evaluation module is used for the confidence level of calculated candidate mistake set, when confidence level during greater than predetermined threshold value, is diagnosis report output with the cooperation of current candidate's Error Set;
Survey route and choose module; Be used for when confidence level during less than predetermined threshold value; With covered with said candidate's mistake set associative do not survey the maximum path of symptom number as detective path; Utilization comprises the detection packet of minimum suspection interstitial content to be surveyed, and feeds back to the model reasoning module with surveying the network information of collecting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101271035A CN102711158A (en) | 2012-04-26 | 2012-04-26 | Method and device for directional diagnosis of wireless sensor network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101271035A CN102711158A (en) | 2012-04-26 | 2012-04-26 | Method and device for directional diagnosis of wireless sensor network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102711158A true CN102711158A (en) | 2012-10-03 |
Family
ID=46903677
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012101271035A Pending CN102711158A (en) | 2012-04-26 | 2012-04-26 | Method and device for directional diagnosis of wireless sensor network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102711158A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104955059A (en) * | 2015-05-07 | 2015-09-30 | 西安电子科技大学 | Cellular network base station state time-varying model establishing method based on Bayesian network |
| CN105530122A (en) * | 2015-12-03 | 2016-04-27 | 国网江西省电力公司信息通信分公司 | Network failure diagnosis method based on selective hidden Naive Bayesian classifier |
| CN106603418A (en) * | 2016-12-09 | 2017-04-26 | 中国联合网络通信集团有限公司 | Network topology updating method and flow analysis equipment |
| CN110474818A (en) * | 2019-07-11 | 2019-11-19 | 福州博泉网络科技有限公司 | Block chain network sniffer, Network Sniffing method and optimization method |
| CN111695583A (en) * | 2019-07-18 | 2020-09-22 | 广东电网有限责任公司信息中心 | Feature selection method based on causal network |
| CN111769976A (en) * | 2020-06-11 | 2020-10-13 | 北京邮电大学 | Self-adaptive recovery method and system for power internet of things terminal fault |
-
2012
- 2012-04-26 CN CN2012101271035A patent/CN102711158A/en active Pending
Non-Patent Citations (1)
| Title |
|---|
| WEI GONG等: "Directional Diagnosis for Wireless Sensor Networks", 《IEEE》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104955059A (en) * | 2015-05-07 | 2015-09-30 | 西安电子科技大学 | Cellular network base station state time-varying model establishing method based on Bayesian network |
| CN104955059B (en) * | 2015-05-07 | 2018-11-16 | 西安电子科技大学 | Cellular network base stations state time-varying model method for building up based on Bayesian network |
| CN105530122A (en) * | 2015-12-03 | 2016-04-27 | 国网江西省电力公司信息通信分公司 | Network failure diagnosis method based on selective hidden Naive Bayesian classifier |
| CN105530122B (en) * | 2015-12-03 | 2018-09-25 | 国网江西省电力公司信息通信分公司 | A kind of network fault diagnosis method based on the hidden Naive Bayes Classifier of selectivity |
| CN106603418A (en) * | 2016-12-09 | 2017-04-26 | 中国联合网络通信集团有限公司 | Network topology updating method and flow analysis equipment |
| CN110474818A (en) * | 2019-07-11 | 2019-11-19 | 福州博泉网络科技有限公司 | Block chain network sniffer, Network Sniffing method and optimization method |
| CN110474818B (en) * | 2019-07-11 | 2023-02-28 | 福州博泉网络科技有限公司 | Block chain network sniffer, network sniffing method and optimization method |
| CN111695583A (en) * | 2019-07-18 | 2020-09-22 | 广东电网有限责任公司信息中心 | Feature selection method based on causal network |
| CN111769976A (en) * | 2020-06-11 | 2020-10-13 | 北京邮电大学 | Self-adaptive recovery method and system for power internet of things terminal fault |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Miao et al. | Agnostic diagnosis: Discovering silent failures in wireless sensor networks | |
| CN102711158A (en) | Method and device for directional diagnosis of wireless sensor network | |
| US8577663B2 (en) | System and methods for fault-isolation and fault-mitigation based on network modeling | |
| Memarzadeh et al. | Integrated inspection scheduling and maintenance planning for infrastructure systems | |
| Jamshidi et al. | Machine learning meets quantitative planning: Enabling self-adaptation in autonomous robots | |
| CN1665205B (en) | Fault detection and diagnosis | |
| Titouna et al. | Outlier detection approach using bayes classifiers in wireless sensor networks | |
| Nesa et al. | Outlier detection in sensed data using statistical learning models for IoT | |
| EP2083335B1 (en) | Automatic controller relationship resolution | |
| US20110004589A1 (en) | Diagnostics in a distributed directory system | |
| Gaddam et al. | Anomaly detection models for detecting sensor faults and outliers in the IoT-a survey | |
| Appice et al. | Dealing with temporal and spatial correlations to classify outliers in geophysical data streams | |
| CN102540054B (en) | Multiple sectioned Bayesian network-based electronic circuit fault diagnosis method | |
| Liu et al. | Using trust in distributed consensus with adversaries in sensor and other networks | |
| CN115278741A (en) | Fault diagnosis method and device based on multi-mode data dependency relationship | |
| CN102883359B (en) | A kind of method of measurement of the key node for wireless sensor network, device and system | |
| Priya et al. | Data fault detection in wireless sensor networks using machine learning techniques | |
| Zhang et al. | A real-time anomaly detection algorithm/or water quality data using dual time-moving windows | |
| Mohapatra et al. | Artificial immune system based fault diagnosis in large wireless sensor network topology | |
| Ferreira et al. | Fault detection and diagnosis for solar-powered wireless mesh networks using machine learning | |
| Scheffel et al. | Increasing sensor reliability through confidence attribution | |
| Ma et al. | BOND: Exploring hidden bottleneck nodes in large-scale wireless sensor networks | |
| US20180216960A1 (en) | Monitoring a sensor array | |
| CN115550144B (en) | Distributed fault node prediction method and device, electronic equipment and storage medium | |
| Arjannikov et al. | Using markov chains to model sensor network reliability |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20121003 |
|
| RJ01 | Rejection of invention patent application after publication |