CN103079168A - Distributed motion node positioning method based on hidden Markov model - Google Patents

Abstract

A method for locating a distributed motion node based on a hidden Markov model, comprising the following steps: first, setting a certain time interval, counting location information, generating probability matrices for the respective location transitions of different users, and counting encounter information, Generate the probability of encountering other user nodes at each location; then, according to the current existing location, divide several sub-sections, in each time interval of this sub-section, there is only encounter information, no known location, the sub-section Both the beginning and the end of the segment have known positions; then, using the obtained known positions of the fixed head and tail of each sub-segment, using the hidden Markov chain model, using the Viterbi forward-backward algorithm combined with the dynamic programming algorithm to determine the fixed head and tail Maximum probability estimation of unknown paths inside a location. The present invention achieves higher positioning accuracy, is suitable for large-area mobile networks with strong user mobility, and is especially suitable for network scenes with relatively sparse distribution in real life.

Description

Locating Method of Distributed Motion Nodes Based on Hidden Markov Model

technical field

The invention relates to a positioning method of a user node in a mobile communication network, in particular to a positioning method of a distributed moving node based on a hidden Markov model, and belongs to the technical field of network communication. the

Background technique

The positioning method of user nodes in mobile communication networks is a technology that has been extensively researched. It has a wide range of application scenarios and is a very important application technology. Existing positioning technologies can be roughly divided into two categories: distance-based methods and connectivity-based methods. the

The positioning method based on distance measurement needs to measure the relative distance between each node, and the measurement accuracy is high. The main measurement method is based on RSS (that is, the received signal strength), the angle of arrival signal, the difference of signal arrival time, etc. etc.; on the contrary, the connectivity-based positioning method utilizes the connectivity between nodes, thereby avoiding the costly high-precision measurement requirements, so this type of positioning method can adapt to various scenarios. the

The currently used connectivity-based positioning method requires a dense node environment. However, in many mobile network environments today, due to the large network area, the mobility of user nodes is too strong, and the network topology is very easy to change. Therefore, in real scenarios, only have relatively low connectivity. This brings great difficulties to the positioning of nodes in the mobile network. The most constructive positioning method in the existing mobile network is MCL (Monte Carlo positioning algorithm), but MCL is very dependent on high-density Fixed nodes, so it is difficult to adapt to practical applications in mobile network scenarios. the

Through the analysis of historical data sets, we found such a phenomenon that the mobility of user nodes shows a strong regularity related to time and space, and more importantly, the mobility of user nodes is related to the relationship between There is a strong correlation. Based on this rule, the present invention proposes a positioning method suitable for sparse mobility networks. the

Contents of the invention

The purpose of the present invention is to overcome the shortcomings of the existing technology that relies too much on the high-density fixed node environment, and provide a distributed moving node positioning method based on the hidden Markov model. The user node density is relatively sparse, and it is only in the In the scenario where other user nodes within the communication radius can only communicate, the hidden Markov chain model is used to realize the positioning of the user node. the

The present invention adopts following technical scheme to achieve the above-mentioned purpose:

A method for locating a distributed motion node based on a hidden Markov model is used for locating a user node in a mobile network, and is characterized in that it comprises the following steps:

(1) System initialization, setting communication radius and time interval, counting location information and generating respective location transition probability matrices for different user nodes, counting encounter information and generating encounter probability with other user nodes at each location;

(2) The user node accesses the AP through WiFi to obtain information about its encounter with other user nodes, and the encounter of the user node with other user nodes refers to: the user node and other user nodes visit the AP within the same time interval or APs within the communication radius of the AP;

(3) The user node obtains its own location by visiting a fixed AP or encountering other fixed user nodes, and the fixed AP and other fixed user nodes have known positions;

(4) Divide several sub-segments according to the currently existing known positions, in each time interval of the sub-segments, there is only encounter information but no known positions, and the beginning and end of each sub-segment have fixed known positions;

(5) With the known positions of the beginning and the end of each sub-section obtained in step (4), use the hidden Markov chain model, utilize the Viterbi forward-backward algorithm in conjunction with the dynamic programming algorithm to determine that the beginning and end positions are fixed and the interior Maximum probability estimation of unknown paths. the

The steps for determining the maximum probability estimation of the fixed start and end positions but the internal unknown path are as follows:

(1) Initially set the location transition probability matrix A and the encounter probability B with other user nodes through historical information statistics;

(2) Starting from the initial position g, the encounter information between the user node and other user nodes in the next time interval is recorded as e _t , and the probability of moving to all i at this moment is δ _i , using the position transfer matrix A and Known encounter information calculated ${\mathrm{\δ}}_i=A_{\mathrm{gi}}*B_{{\mathrm{ie}}_t};$

(3) Let t=t+1, the encounter information between the user node and other user nodes in the next time interval is denoted as e _t , and the probability of moving to all j in the next time interval is δ _j , using ${\mathrm{\δ}}_j=\underset{i}{\max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t},$ and record $\underset{i}{\arg \max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t};$

(4) Repeat step (3) until t exceeds the tail time;

(5) Let h be the known position of the tail, and the maximum probability δ _j of the unknown path at position j before reaching the tail is known from the first 4 steps, and starting from position j, use ${\mathrm{\δ}}_h=\underset{j}{\max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{he}}_t},$ and record ${\mathrm{\δ}}_h=\underset{j}{\max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{je}}_t};$

(6) According to the optimal path searched and recorded at each step, reversely deduce the path T with the highest probability. the

The present invention utilizes the encounter information that can be directly observed, uses the hidden Markov chain model to calculate the most probable path position, and thus realizes the location of the user node. the

The advantages of the present invention are:

(1) The positioning method of the present invention can be adapted to mobile networks in large areas, and is suitable for strong user node mobility. Even if the network topology changes, the positioning error of the present invention will not change greatly, and the present invention is applicable Compared with the network scene with relatively sparse distribution in real life, it has stronger adaptability to the strong assumptions made by the MCL method on the network environment. the

(2) The present invention is a distributed positioning method, rather than a routing method controlled by the control center. The user node only needs to count a small amount of communication with other nodes that it meets within its own communication radius, instead of Communication between each pair of user nodes is required, which reduces energy consumption and communication expenses. the

(3) The present invention utilizes the encounter information that can be directly observed, uses the hidden Markov model to calculate the most probable path position, and can achieve higher positioning accuracy. the

Detailed ways

The positioning method of the distributed motion node based on the hidden Markov model in the present invention is based on the connectivity mode, fully utilizes the mobility and the mutual relationship between the encountered user nodes, and obtains the information recorded by the user node and other users within the communication range Based on the historical encounter information of nodes, a hidden Markov model is established to calculate the most likely path position to achieve the purpose of user node location. The method is particularly suitable for mobile network scenarios where user nodes are sparse. the

Each step of the present invention is described in further detail below:

The locating method of the described distributed motion node based on Hidden Markov Model is used for the locating of user node in mobile network, and it comprises the following steps:

(1) System initialization, setting the communication radius and time interval, counting location information, and generating separate location transition probability matrices for different user nodes, that is, the state transition of the location of the user node at the previous moment to the location of the next moment Matrix, statistical encounter information, generates the probability of encounter with other user nodes at each position, that is, the probability of encounter with other user node sets at this position at the same time at this moment. the

(2) The user node accesses the AP through WiFi, and there are other user nodes visiting the AP within the same time interval, that is, the user node meets other user nodes; if there is access and the AP is within the communication distance The other user nodes of the AP also meet the user node for these other user nodes. In other words, the user node accesses the AP through WiFi to obtain information about its encounter with other user nodes. APs within the communication radius. the

(3) the location of some APs, so when the user node accesses the AP through WiFi, it can obtain the current location, or some user nodes are fixed and its location is known, then the user node When a node visits the fixed AP or meets a fixed user node, it knows where it is at that moment. In other words, the user node obtains its own location by visiting a fixed AP or meeting other fixed user nodes, and the fixed AP and other fixed user nodes have known locations. the

(4) Divide several sub-segments according to the currently existing known positions. In each time interval of this sub-segment, there is only encounter information, and there is no known position, that is, there is no fixed node in this sub-segment. Encountered, the beginning and end of each sub-segment have a fixed known position, that is, each sub-segment meets a known fixed node at the beginning and end of each sub-segment. the

(5) with the known position of each subsection of step (4) gained fixed beginning and end, use Hidden Markov chain model, utilize Viterbi forward and backward algorithm in conjunction with the relevant steps of dynamic programming algorithm to determine pair beginning and end Maximum probability estimation of paths with fixed positions and unknown interiors. the

The steps for determining the maximum probability estimation of the fixed start and end positions but the internal unknown path are as follows:

(1) Initially set the position transition probability matrix A. This state transition matrix is calculated through historical information. Set the probability matrix B that encounters other user node sets at this position at the same time. This matrix is also calculated through historical information;

(2) Starting from the initial position g, the encounter information between the user node and other user nodes in the next time interval calculated by historical information is denoted as e _t , and the probability of moving to all i at this moment is δ _i , Calculated using the position transfer matrix A and known encounter information

其中i表示前一时刻在位置i而在下一时间间隔时移动到j的概率，在求出概率最大的δ_j的同时记录下  $\underset{i}{\arg \max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t};$ (3) Let t=t+1, the user node encounters other user nodes in the next time interval is denoted as e _t , and the probability of moving to all j in the next time interval is δ _j , using

Among them, i represents the probability of moving to j in the next time interval at the position i at the previous moment, and record it while finding the δ _j with the highest probability $\underset{i}{\arg \max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t};$

(4) Step (3) is repeated until t exceeds the tail time, that is, the calculation to the next time interval is the end of the subsection;

(5) Let h be the known position of the tail, and the maximum probability δ _j of the unknown path at position j before reaching the tail is known from the first 4 steps, and starting from position j, use ${\mathrm{\δ}}_h=\underset{j}{\max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{he}}_t},$ and record ${\mathrm{\δ}}_h=\underset{j}{\max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{je}}_t};$

(6) According to the optimal path searched and recorded at each step, that is, the previous position where the probability is maximized, along the parameters that maximize the probability of each previous step, reversely deduce the path T with the maximum probability. the

Following is the pseudocode implementation of the present invention, i.e. source program:

The present invention analyzes the historical data set, applies the strong correlation between the mobility of the user node and other user nodes encountered and the regularity of time-space correlation, and utilizes the information recorded by the user node and within the communication range For the encounter information of other user nodes, the hidden Markov chain model is used to calculate the most likely path position to realize the positioning of user nodes. The present invention achieves higher positioning accuracy and is suitable for large-area mobile networks with strong user mobility. the

Claims (2)

1. the localization method based on the distributed movement node of hidden Ma Shi model is used for the mobile network to the location of user node, it is characterized in that, may further comprise the steps:

(1) system initialization is set communication radius and the time interval, and the statistics positional information also generates minute other position transfer probability matrix for the different user node, meet information and be created on each position collision probability with other user nodes of statistics;

(2) described user node obtains the information that itself and other user node meets by WiFi access AP, and described user node and other user nodes meet and refer to: this user node and other user nodes are accessed this AP or are in the interior AP of this AP communication radius in the interval at one time;

(3) described user node obtains itself present position by accessing fixing AP or meeting with fixing other user nodes, and described fixing AP and other fixing user nodes have known location;

(4) mark off some subsegments according to current existing known location, within each time interval of this subsegment, only meet information and without known position, the beginning of each subsegment and end have fixing known location;

(5) with the beginning of each subsegment of step (4) gained and the known location at end, use the hidden Markov chain model, utilize the Viterbi forward backward algorithm to determine beginning and end position are fixed and the maximum probability estimation in inner unknown path in conjunction with dynamic programming algorithm.

2. the localization method of the distributed movement node based on hidden Ma Shi model according to claim 1 is characterized in that, described beginning and end position is fixed and definite step of the maximum probability estimation in inner unknown path is as follows:

(1) by historical information statistics, initial set up position transition probability matrix A and with the collision probability B of other user nodes;

(2) from initial position g, be designated as e in the information of meeting of described user node of next time interval and other user nodes _t, and be δ at the probability that this moment may move to all i _i, utilize the position transfer matrix A and knownly meet information calculations out ${\mathrm{\δ}}_i=A_{\mathrm{gi}}*B_{{\mathrm{ie}}_t};$

(3) make t=t+1, the information of meeting of this user node and other user nodes is designated as e within next time interval _t, and the probability that moves to all j within next time interval is δ _j, utilize ${\mathrm{\δ}}_j=\underset{i}{\max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t},$ And record $\underset{i}{\arg \max }{\mathrm{\δ}}_i*A_{\mathrm{ij}}*B_{{\mathrm{ie}}_t};$

(4) repeating step (3) is until t surpasses the afterbody time;

(5) establishing h is the known present position of afterbody, by the maximum probability δ that is in the unknown path of position j before the known arrival afterbody of front 4 steps _j, establish from position j, utilize ${\mathrm{\δ}}_h=\underset{j}{\max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{he}}_t},$ And record $\underset{j}{\arg \max }{\mathrm{\δ}}_j*A_{\mathrm{jh}}*B_{{\mathrm{je}}_t};$

(6) optimal path of searching for and recording according to per step, backstepping goes out to have the path T of maximum probability.