CN115297518A - Network switching method and system based on mobile user position - Google Patents

Network switching method and system based on mobile user position Download PDF

Info

Publication number
CN115297518A
CN115297518A CN202210945986.4A CN202210945986A CN115297518A CN 115297518 A CN115297518 A CN 115297518A CN 202210945986 A CN202210945986 A CN 202210945986A CN 115297518 A CN115297518 A CN 115297518A
Authority
CN
China
Prior art keywords
network
user
switching
candidate
mobile
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
Application number
CN202210945986.4A
Other languages
Chinese (zh)
Inventor
刘洋
柴进
赵鑫
刘中艳
陈泽
李东柏
贺鑫
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Inner Mongolia University
Original Assignee
Inner Mongolia University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Inner Mongolia University filed Critical Inner Mongolia University
Priority to CN202210945986.4A priority Critical patent/CN115297518A/en
Publication of CN115297518A publication Critical patent/CN115297518A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W36/00—Hand-off or reselection arrangements
    • H04W36/14—Reselecting a network or an air interface
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/08—Load balancing or load distribution
    • H04W28/09—Management thereof
    • H04W28/0925—Management thereof using policies
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W36/00—Hand-off or reselection arrangements
    • H04W36/24—Reselection being triggered by specific parameters
    • H04W36/32—Reselection being triggered by specific parameters by location or mobility data, e.g. speed data
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The application discloses a network switching method and a system thereof based on the position of a mobile user, wherein the network switching method based on the position of the mobile user specifically comprises the following steps: network scanning is carried out; predicting the position of the mobile user to predict the position of the user at the next moment; estimating the received signal strength according to the predicted position of the user at the next moment, and entering a switching trigger mechanism according to the received signal strength; performing candidate network pre-selection according to the moving speed of a user and the bandwidth required by the user request service; judging whether the candidate network is unique; and if the candidate network is not unique, performing multi-attribute decision, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete the switching of the network. The method and the device can improve the user position prediction precision, thereby effectively reducing the switching times and the switching failure times, reducing the switching time delay, and balancing the network load while ensuring that the user switches to a proper network.

Description

Network switching method and system based on mobile user position
Technical Field
The present application relates to the field of mobile communications, and in particular, to a network handover method and system based on the location of a mobile subscriber.
Background
The rapid rise of mobile internet applications has made the global wireless communication service to exhibit exponential growth. In response to the communication service demands of users, related research has produced many different types of wireless communication access technologies. Existing Wireless access technologies can be classified into Wireless Local Area Networks (WLANs), mobile cellular networks, wireless Wide Area Networks (WWANs), satellite communication networks, and the like according to coverage areas. On the other hand, mobile communication technology has progressed from the first generation communication technology to the fifth generation (5G) communication technology in short decades, and with the gradual commercialization of the 5G technology, research on the sixth generation (6G) mobile communication technology has been developed. The network environment with multiple coexisting access technologies is called a heterogeneous network, and because of the position movement of a user or the change of a user communication service, a base station or a network access point currently accessed by the user may not guarantee the communication quality at the next moment, the user needs to adjust to access a new network, and how to guarantee that the communication is not interrupted when the access network is changed and the communication quality is not affected after the access to the new network is a hot point problem to be solved in the heterogeneous network. When a user is confronted with handover in a heterogeneous network environment, it is considered when to initiate a handover request and how to select a network most suitable for the user among a plurality of candidate networks to complete network handover according to the service preference of the user, the mobility of the user and the actual network condition of each candidate network after the handover is initiated, which also becomes a main standard for judging the handover policy is good or bad. In a dense heterogeneous network environment, because network coverage areas and network performances of different access network types are greatly different, when network handover is required, research needs to be performed from when and where network handover is initiated and how to handover to a network that is satisfied by a user in consideration of aspects such as mobility of the user, network parameter requirements of current service of the user, actual conditions of an access network, and load conditions of the access network. For example, a switching algorithm based on RSS combined with hysteresis level is proposed in the prior art, and network switching is performed only when the RSS value of a candidate network is greater than the sum of the RSS value and the hysteresis level value of the current access network. This reduces the number of unnecessary handovers to a certain extent. And judging to turn on or turn off the network discovery function by comparing the predicted RSS with a threshold value, thereby reducing the energy consumption of the terminal. Since the network discovery operation for scanning the network environment around the user is a precondition for subsequent handover decision, if the predicted RSS fluctuates around the threshold, the handover cannot be triggered in time, which may cause communication interruption. In summary, RSS-based handover algorithms generally have the advantages of short algorithm operation time and easy implementation of handover procedures, but for current and future increasingly complex heterogeneous network environments, it is difficult to ensure that a handed-over network can meet the communication service requirements of users only by using RSS as a parameter for deciding handover, and thus a ping-pong effect may be caused. Therefore, the RSS based handover algorithm should be used in conjunction with other types of algorithms. Still, a handover decision algorithm based on mobile user location prediction and a fuzzy analytic hierarchy process is also provided in the prior art, and simulation results prove that the method effectively reduces handover delay and balances network load. However, the 5G network is not considered when modeling the network environment, and thus the established handover strategy may not be able to cope with the more complex 5G heterogeneous network environment. In addition, the proposed method for predicting the mobile location of the user does not obviously explain the role in the heterogeneous network environment, so that the influence of the prediction method on the handover performance needs to be further researched. And allocating subjective weight to the network parameters subjectively preferred by the user by using an analytic hierarchy process, allocating objective weight to the network parameters by using an exponential weighted product and a simple weighting method, and finally selecting a target network according to the calculated candidate network return value. Simulation results verify that the algorithm can select a proper switching network for a user according to subjective preferences of the user and network conditions of candidate networks, but the current gradually-used 5G network is not considered in network environment modeling, and the phenomena of when to initiate network switching and how to avoid frequent switching are not explicitly analyzed. Currently, most of such algorithms only study how to switch users to appropriate networks, but ignore the conditions of when the users should trigger the handover mechanism and whether all candidate networks meet the access after the handover trigger, so that the computation time complexity of the whole handover process is high, and the unnecessary handover times are increased. The prior art still discloses a switching decision algorithm based on a feedforward neural network, a data set is constructed according to collected attributes of different networks and is input into the feedforward neural network for training, and then a trained neural network model is used for providing an optimal network which a user should be connected between the networks when switching and selecting the network. From experimental simulation results, the feasibility of the neural network for the network environment perception is verified, but the performance of the proposed algorithm on handover performance is not explicitly discussed, the collection process of the data set is not explicitly described, and when handover should be triggered is not considered.
In summary, the above handover strategies in heterogeneous network environments have their advantages and disadvantages, but the following problems still remain to be solved: the whole process of the handover mainly comprises a handover triggering stage and a handover decision stage. The handover trigger phase determines when the user initiates a handover. In the handover decision stage, the most appropriate network is selected from the candidate networks for the user to complete handover, and if the selected network does not meet the user requirement after the handover decision, frequent handover requests occur, so that network access is overloaded, the user experience is influenced, and the communication cost of the base station is increased. In addition, the whole switching process needs to pay attention to the operation time of the algorithm, so that the influence on the user experience caused by too long switching time delay is avoided. And the switching strategy based on the artificial intelligence technology can realize the advance pre-knowledge of the network condition compared with other algorithms by sensing the known data, so that the base station can reserve resources in advance or carry out switching judgment in advance, and the aim of reducing switching failure is fulfilled. However, the currently proposed prediction method is more dependent on the current network environment during model training, and has no general feasibility, and the prediction accuracy of the model directly affects the handover performance, so the prediction accuracy also needs to be further optimized.
Therefore, how to provide a method for providing a high-quality handover strategy for a user in time and efficiently in a future complex communication network environment so as to ensure seamless communication of a mobile user and access to a corresponding network anytime and anywhere as required is a problem which needs to be solved urgently by a person skilled in the art.
Disclosure of Invention
The invention aims to design and realize a high-quality network switching system which is timely and efficiently provided for users in a future complex communication network environment, and ensures that the communication of the users is not interrupted in the network switching.
In order to solve the above problem, the present application provides a network handover method based on a mobile user location, which specifically includes the following steps: network scanning is carried out; in response to the completion of network scanning, predicting the position of the mobile user, and predicting the position of the user at the next moment; estimating the received signal strength according to the predicted position of the user at the next moment, and entering a switching trigger mechanism according to the received signal strength; after entering a switching trigger mechanism, pre-selecting a candidate network according to the moving speed of a user and the bandwidth required by the user request service; judging whether the candidate network is unique; and if the candidate network is not unique, performing multi-attribute decision, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete the switching of the network.
As above, if the candidate network is unique, the handover is directly performed to the candidate network, and the handover of the network is completed.
As above, in response to completing the network scanning, the location prediction of the mobile user is performed, and the location of the user at the next time is predicted, which specifically includes the following sub-steps: acquiring a mobile position coordinate and a mobile speed of a user terminal to construct a data set; inputting the constructed data set into a BOLSTM model for data preprocessing, and extracting the mobile characteristics of a user; in response to the extraction of the mobile features of the user, optimizing the LSTM model by using Bayes, determining hyper-parameters, and finishing the training and prediction of the BOLSTM model according to the determined hyper-parameters; the trained BOLSTM model is loaded to predict the location of the user at the next time.
The above, wherein the collected coordinates of the user's moving position and the user's moving speed will form a user track sequence, and the user track sequence T will be r ＝{p 1 ,p 2 ,...,p T Each moving track p of the mobile users in T Corresponding position coordinate l i And (i is more than or equal to 1 and less than or equal to T) preprocessing the data and inputting the preprocessed data into the neural network to extract the mobile features of the user.
As above, if the received signal strength is lower than the handover trigger threshold while the candidate network set is not empty, the handover trigger mechanism is activated in advance.
As above, the received signal strength is specifically expressed as:
P(dBm)＝P t -PL
P t for the transmission power of MBS or SBS, PL denotes PL (dB) MBS Or PL (dB) SBS When calculating the signal strength of the access MBS, P t For the transmit power of MBS, PL denotes PL (dB) MBS (ii) a When calculating the signal strength of the SBS access, P t For the transmission power of SBS, PL denotes PL (dB) SBS 。
As above, if the moving speed of the user is greater than the maximum moving speed supported, the networks in the candidate network set are deleted, and if the bandwidth required by the user to request the service is greater than the available bandwidth of the candidate network, the candidate network is moved out of the candidate network set.
As above, the performing the multi-attribute decision, determining the target network in the candidate networks according to the multi-attribute decision, and switching the network to the target network to complete the network switching specifically includes the following sub-steps: calculating subjective weight; calculating an objective weight in response to completing the calculation of the subjective weight; determining an optimal decision according to the subjective weight and the objective weight so as to determine a target network; and switching the network to the target network to complete the switching of the network.
As above, the network parameter weight assignment is specifically expressed as:
Figure BDA0003787482290000051
wherein w j Weight representing the jth network parameter, c ij Indicating the relative importance of the network parameter i to the network parameter j and n indicating the total number of network parameters.
A network switching system based on mobile user position comprises a network scanning unit, a user position prediction unit, a switching trigger unit, a candidate network pre-selection unit, a judgment unit and a network switching unit; a network scanning unit for performing network scanning; the user position prediction unit is used for predicting the position of the mobile user in response to the completion of network scanning and predicting the position of the user at the next moment; the switching trigger unit is used for estimating the received signal strength according to the predicted position of the user at the next moment and entering a switching trigger mechanism according to the received signal strength; the candidate network pre-selection unit is used for pre-selecting a candidate network according to the moving speed of the user and the bandwidth required by the user request service after entering a switching trigger mechanism; a judging unit, configured to judge whether the candidate network is unique; if the candidate network is unique, directly switching the network to the candidate network to complete the switching of the network; and the network switching unit is used for carrying out multi-attribute decision if the candidate network is not unique, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete network switching.
The application has the following beneficial effects:
the network switching method and the network switching system based on the mobile user position can improve the user position prediction precision through the network switching strategy, thereby effectively reducing the switching times and the switching failure times, reducing the switching time delay, and balancing the network load while ensuring that the user is switched to a proper network.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art according to the drawings.
Fig. 1 is a schematic diagram of a network handover procedure in a dense heterogeneous network scenario according to an embodiment of the present application;
fig. 2 is an internal structure diagram of a network handover system based on a mobile subscriber location according to an embodiment of the present application;
fig. 3 is a flowchart of a network handover method based on a mobile user location according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a hierarchy model provided according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application are clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments obtained by a person skilled in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
The network switching method and the system based on the mobile user position can conveniently realize the data storage of the mobile position of the user so as to complete the training of a prediction model, and can better utilize global network environment data to assist in completing the network switching.
Specifically, since the quality of communication services provided by multiple radio access technologies to users varies greatly, users are faced with not only the problem of when to switch but also the problem of switching to a suitable network. In the switching triggering stage, a switching triggering strategy for predicting the mobile position of the user by using a Bayesian optimization long-short term memory network is provided. The neural network can learn the user movement behavior more efficiently by optimizing the super-parameter setting of the long-term and short-term memory network through Bayes, so that a user movement model is constructed. And then, a prediction model is loaded to predict the change of the mobile position estimation network condition of the user at the next moment so as to realize the purpose of triggering the switching in time. In the stage of switching decision, a switching decision strategy of multi-attribute decision is provided. The method comprises the steps of screening and removing candidate networks which do not meet requirements according to a candidate network pre-selection mechanism, and selecting a proper network for a user according to subjective preferences of the current service of the user on five network parameters including bandwidth, time delay jitter, packet loss rate and cost and actual conditions of the candidate networks if the candidate networks are not unique.
Fig. 1 shows a schematic diagram of a handover process in a dense heterogeneous network scenario studied by the present application, where a 4G macro base station cellular network provides wide area communication coverage, a 5G small cellular base station provides low-latency and high-capacity access service, a WLAN access point provides network access service with low access cost and high data transmission rate, and a user performs multiple network handovers due to location movement and a service quality requirement of a request service in the dense heterogeneous network scenario composed of the 5G heterogeneous cellular network and multiple WLAN access points.
The method for calculating the received signal strength of the user terminal accessing the WLAN access point comprises the following steps:
P rW ＝P tW -35.2-20lgf W -20lgd W -u (1)
wherein: p tW Representing the transmit power of the WLAN access point. f. of W Representing the carrier frequency in MHz. u is gaussian white noise subject to a mean of 0 and a variance of σ = 2.
In the application, a macro base station of 4G selects a COST-231Walfisch-Ikegami path loss transmission model, a small base station of 5G selects a UMI path loss model, and the path loss PL (dB) of the COST-231Walfisch-Ikegami path loss transmission model MBS And path loss PL (dB) of the UMI path loss model SBS See formula (1.2) and formula (1.3), respectively.
PL(dB) MBS ＝42.6+26log 10 (d i )+20log 10 (f m ) (2)
PL(dB) SBS ＝32.4+21log 10 (d j )+20log 10 (f s ) (3)
Wherein: d i Respectively, the distances between the user and an MBS (Macro base station), d j Denotes the distance of the user from the SBS (Small base station), f m And f s The operating frequencies of MBS and SBS are indicated separately.
Example one
Referring to fig. 2, the system for network handover based on the location of a mobile subscriber provided in this embodiment may specifically be a central controller, where the central controller specifically includes a network scanning unit 210, a subscriber location prediction unit 220, a handover trigger unit 230, a candidate network pre-selection unit 240, a determination unit 250, and a network handover unit 260.
The network scanning unit 210 is used for network scanning.
The user location prediction unit 220 is connected to the network scanning unit 210, and is configured to perform mobile user location prediction in response to completion of network scanning, so as to predict a location of the user at the next time.
The handover triggering unit 230 is connected to the user location predicting unit 220, and is configured to estimate the received signal strength according to the predicted location of the user at the next time, and enter a handover triggering mechanism according to the received signal strength.
The candidate network pre-selection unit 240 is connected to the handover triggering unit 230, and is configured to perform candidate network pre-selection according to the moving speed of the user and the bandwidth required by the user request service after entering the handover triggering mechanism.
The judging unit 250 is connected to the candidate network pre-selecting unit 240 for judging whether the candidate network is unique. And if the candidate network is unique, directly switching the network to the candidate network to complete the switching of the network.
The network switching unit 260 is connected to the determining unit 250, and configured to perform a multi-attribute decision if the candidate network is not unique, determine a target network in the candidate network according to the multi-attribute decision, and switch the network to the target network, thereby completing the network switching.
Example two
Referring to fig. 3, a network handover method based on a mobile subscriber location provided in this embodiment specifically includes the following steps:
step S310: network scanning is performed.
Step S320: and responding to the completion of network scanning, predicting the position of the mobile user, and predicting the position of the user at the next moment.
The LSTM is subject to the problems of slow training convergence and low prediction precision due to the setting of hyper-parameters during training and prediction, and the hyper-parameter combination which is most suitable for the current user is automatically searched through the Bayesian algorithm, so that the training convergence speed and the prediction precision can be effectively improved. In the step, the LSTM model after Bayesian optimization is used as a BOLSTM model (Bayesian Optimized LSTM), the movement information of the user is mined through the BOLSTM model, and the LSTM loss function is Optimized by means of a Bayesian optimization algorithm to find the hyperparameter which is most suitable for learning and predicting the position of the mobile user, so that the convergence of the LSTM model is accelerated, and the error of predicting the mobile position of the user is reduced.
The step S320 specifically includes the following sub-steps:
step S3201: and collecting the mobile position coordinates and the mobile speed of the user terminal to form a data set.
Specifically, because there is a certain correlation between adjacent track points moved by the user and the movement rule of the user is learned by the prediction model for better learning, the movement position coordinates and the movement speed of the user are sampled at regular time according to the period T.
Step S3202: and inputting the constructed data set into a BOLSTM model for data preprocessing, and extracting the movement characteristics of the user.
The collected coordinates of the moving position of the user and the moving speed of the user form a user track sequence, and the user track sequence T r ＝{p 1 ,p 2 ,...,p T Each moving track p of the mobile users in (1) } T Corresponding position coordinate l i (i is more than or equal to 1 and less than or equal to T) and then input into a BOLSTM model for extraction after data preprocessingA movement characteristic of the user.
Specifically, in this step, a maximum and minimum scaler is used to perform data preprocessing of the position coordinates, where the preprocessing process is specifically expressed as:
Figure BDA0003787482290000091
wherein: x = (X) 1 ,x 2 ,x t ...,x n )，x t Is X = (X) 1 ,x 2 ,x t ...,x n ) Of (1) a value to be transformed, x j Is the transformed value. In particular x t As a mobile position coordinate, x, of the user j The transformed mobile position coordinates of the user. min (X) denotes the sequence X = (X) 1 ,x 2 ,x t ...,x n ) Max (X) represents the sequence X = (X) 1 ,x 2 ,x t ...,x n ) Of (2) is calculated.
Step S3203: and in response to the extraction of the mobile features of the user, optimizing the LSTM model by using Bayes, determining the hyperparameter, and finishing the training and prediction of the BOLSTM model according to the determined hyperparameter.
The penalty function used in the LSTM model is set using conventional experience. The loss function uses the root mean square error as an index, and the corresponding calculation formula is:
Figure BDA0003787482290000092
wherein: l i+1 For the actual value of the mobile user's position coordinates,
Figure BDA0003787482290000093
is a predicted value of the mobile user's position coordinates.
The specific reference of the hyper-parameters to be optimized of the LSTM is shown in Table 1.
TABLE 1 LSTM to-be-optimized hyper-parameter List
Figure BDA0003787482290000101
Step S3203 specifically includes the following substeps:
step S32031: and establishing the Gaussian distribution condition of the objective function of Bayesian optimization according to the hyper-parameters to be optimized.
According to the analysis of the super-parameter setting of the LSTM model in Table 1, the initial value of the super-parameter to be optimized is set to be the minimum value in the range of Table 1. In the process of training the LSTM model, a loss function value of the LSTM model training is used as an objective function in the Bayesian optimization process, and a corresponding hyper-parameter to be optimized is used as a sampling point. Therefore, the Gaussian distribution condition of the objective function of Bayesian optimization is established by sampling hyper-parameter combinations. The hyper-parameter combinations consist of the parameters in 4 in table 1. Wherein the value in the hyper-parameter combination at this time is the above value range. For example, the value of droupout is [0.1,0.8].
Step S32032: and calculating and evaluating the Gaussian distribution condition according to a sampling function in Bayesian optimization, and determining the value range of the hyper-parameter combination.
Specifically, the gaussian distribution condition is calculated and evaluated according to a sampling function in bayesian optimization, that is, a mean value and a variance corresponding to each hyper-parameter are calculated according to the hyper-parameters sampled in the front, the value range in the hyper-parameter combination is updated according to the mean value and the variance, and the value range is specifically reduced.
A larger mean value indicates a larger desired result of the model training for the hyperparameter, and a larger variance indicates a larger uncertainty of the hyperparameter in maximizing the objective function value, which is more worthwhile to be explored.
Determining the range corresponding to the hyper-parameter with the mean value and the variance larger than the specified threshold, for example, if the mean value and the variance corresponding to the parameter dropout with the value range of [0.2,0.6] are larger than the specified threshold, updating the value range of the parameter dropout, determining the hyper-parameter combination after updating the value ranges of all the parameters in the hyper-parameter combination, and executing the step S32033 by using the updated value ranges of all the parameters in the hyper-parameter combination.
Step S32033: and (4) carrying out LSTM model training by using the determined hyper-parameter combination, and outputting the obtained loss function value as a new objective function value.
In the process of one round of model training, a value is randomly selected from the updated value range of each parameter for the training of the LSTM model, for example, the parameter droupout randomly takes the value of 0.2, the value of alpha takes the value of 0.05 and the like, and the loss function value obtained by training is output as a new target function value.
Step S32034: and judging whether the new objective function meets the requirements or not.
And if the target function meets the requirements, outputting the hyper-parameters corresponding to the new target function, and performing the next round of training. If not, the process returns to step S32032-S22033.
In the next round of training, the value of the hyper-parameter different from the previous round is taken in the updated value range, for example, the parameter droupout is randomly taken to be 0.3, the alpha value is taken to be 0.1, and the like, and the next round of training of the LSTM model is carried out according to the re-selected hyper-parameter value until the value of each hyper-parameter corresponding to the minimum objective function is obtained. And performs step S32035 according to the plurality of hyper-parameter values.
And if the objective function value obtained by the sampling point meets the requirement, listing the objective function value as a secondary advantage and performing model training again until the minimum corresponding hyperparameter combination of the objective function is selected for neural network training and prediction.
Step S32035: and training the LSTM model according to the output hyper-parameters to obtain the trained BOLSTM model.
And carrying out LSTM model training by using the value of each hyper-parameter corresponding to the minimum objective function to obtain a BOLSTM model after training.
Step S3204: and loading the trained BOLSTM model to predict the position of the user at the next moment.
According to the method, an LSTM model based on Bayesian optimization is trained in an off-line mode, so that the movement rule of a user is captured, a corresponding user movement model is generated, and then a user position prediction unit can be loaded on line in a heterogeneous network environment, and the position prediction of the user is realized.
Wherein the user movement model is a model existing in the prior art.
Step S330: and estimating the received signal strength according to the predicted position of the user at the next moment, and entering a switching trigger mechanism according to the received signal strength.
The method for calculating the signal strength P (dBm) of the user accessing the MBS or SBS comprises the following steps:
P(dBm)＝P t -PL (6)
wherein: p t For the transmission power of MBS or SBS, PL denotes PL (dB) MBS Or PL (dB) SBS When calculating the signal strength of the access MBS, P t For the transmit power of MBS, PL denotes PL (dB) MBS . When calculating the signal strength of the SBS, P t For the transmission power of SBS, PL represents PL (dB) SBS 。
If the received signal strength is lower than the handover trigger threshold and the candidate network set is not empty, a handover trigger mechanism is activated in advance, so that more processing time is reserved for a subsequent handover decision stage, and the purpose of shortening the handover delay is achieved.
If the received signal strength is higher than the handover trigger threshold and at the same time, the candidate network set is not empty, the step S310 is executed.
In particular, if the user actively initiates a service type change, the trigger mechanism is directly switched.
Step S340: and after entering a switching triggering mechanism, pre-selecting the candidate network according to the moving speed of the user and the bandwidth required by the user request service.
According to the moving speed of user and the bandwidth required by user request service, the candidate network preselection is carried out to reduce unnecessary switching times and increase user access success rate, and in addition, the number of candidate networks can be reduced after preselection, thereby reducing the calculation time cost of subsequent switching judgment and realizing the reduction of switching judgmentThe purpose of the delay. And if the moving speed of the user is greater than the supported maximum moving speed, deleting the WLAN networks in the candidate network set. Then, if the bandwidth required by the user request service is larger than the available bandwidth of the candidate network I, the candidate network I is moved out of the candidate network set L J And deleting the candidate networks which do not meet the conditions to finish the pre-selection of the candidate networks.
Specifically, for the above candidate network set L J Whether the ith network will be removed from the candidate network set may be determined using a function f (I):
f(I)＝F BJ *F mJ (7)
wherein: f BJ And F mJ And respectively representing the utility functions of the bandwidth and the moving speed required by the user J, and the value is 0 or 1.
Finally, if the result of f (I) is 0, the candidate network I is selected from the candidate network set L of the user J J Otherwise, the network is reserved.
Step S350: and judging whether the candidate network is unique.
If the candidate network is not unique, the following handover decision process is performed, and step S360 is executed. Otherwise, the network is directly switched to the candidate network to complete the network switching.
Step S360: and performing multi-attribute decision, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete the switching of the network.
In the embodiment, five network parameters, namely bandwidth, time delay jitter, packet loss rate and cost, are selected to represent characteristics of different service types, and the multi-attribute decision specifically includes performing utility value calculation by comprehensively considering network parameter preference and actual network conditions when a user requests different service types, and selecting a network with the maximum utility value as a target network to complete switching. Firstly, according to the description of the communication service characteristics in the 3GPP standard document, the communication service types are divided into four communication service types of conversation type, stream type, interaction type and background type.
The step S360 specifically includes the following sub-steps:
step S3601: subjective weights are calculated.
The step determines the weight distribution of each parameter according to the subjective preference of the user to the network parameters of different service types. The method specifically comprises the following substeps:
step S36011: and establishing a hierarchical structure model.
The schematic structural diagram of the hierarchical structure model determined in this embodiment is shown in fig. 4, where the scheme layer represents listed candidate networks, the criterion layer characterizes network parameters affecting the quality of service of the candidate networks, and the target layer represents the candidate network closest to the ideal scheme.
Step S36012: and responding to the established hierarchical structure, and determining the relative importance degree of each network parameter when different service types construct a fuzzy decision matrix.
After the hierarchical structure model is established, the mutual importance degrees of the five network parameters representing the service quality selected from the text are sorted according to the different requirements of the session class, the stream class, the interaction class and the background class services summarized in the table 2 on the service quality, so that the relative importance degree between every two decision attributes is determined to construct a fuzzy decision matrix.
Table 2 characteristics of four traffic types
Figure BDA0003787482290000141
According to the analysis, a fuzzy decision matrix is established according to the relative importance degree evaluation rule among the attributes given by the fuzzy similarity relation table, wherein the fuzzy decision matrix comprises the following steps:
Figure BDA0003787482290000142
wherein: c. C ij Indicating the relative importance of the network parameter i to the network parameter j and n indicating the total number of network parameters. In particular having ii =0.5 and c ij +c ji ＝1。
Step S36013: and calculating the weight distribution of the network parameters under different service types in response to the completion of the establishment of the fuzzy matrix.
After the fuzzy decision matrix is established, if the evaluation decision matrix meets the consistency evaluation standard, the network parameter weight value is calculated, wherein the evaluation decision matrix meets the consistency evaluation standard by referring to a consistency evaluation standard method for verifying an analytic hierarchy process in statistics in the prior art, and the evaluation standard is not repeated.
The calculation mode of the network parameter weight value is specifically expressed as follows:
Figure BDA0003787482290000143
wherein: w is a j Representing the weight of the jth network parameter.
The weight distribution of the network parameters under each service type obtained by the above formula is shown in table 3.
Table 3 corresponding network parameter weight table for different services
Figure BDA0003787482290000151
Step S3602: in response to completing the calculation of the subjective weight, an objective weight is calculated.
After subjective weights of different services on network parameters are evaluated, in order to avoid the influence of subjective evaluation on the final selected target network as much as possible, the objective weights of the network attributes of the candidate networks are solved by using an entropy weight method. Since the network parameters characterizing the service quality are not simply better if the values are larger, for example, the values of delay, packet loss rate and cost are smaller, the service quality of the network is higher. In order to unify the variation standard, the network parameters are normalized according to the cost type and the benefit type.
In this step, the cost-type parameters include time delay, time delay jitter, packet loss rate and cost, and the normalization process performed on the cost-type parameters is specifically represented as:
Figure BDA0003787482290000152
wherein r' Ij The final parameter value n of the jth network parameter of the ith candidate network after normalization processing in the cost type parameters 1 Number of cost-effective network parameters, r Ij And the parameter value of the jth network parameter of the ith candidate network is represented, and m is the number of the candidate networks.
The benefit type parameter is available bandwidth, and the normalization processing on the utility type parameter is specifically represented as:
Figure BDA0003787482290000153
wherein n is 2 For the number of benefit type network parameters, r Ij A parameter value representing a jth network parameter of the ith candidate network,
Figure BDA0003787482290000154
the method is the final parameter value of the jth network parameter of the ith candidate network after normalization processing in the benefit type parameter.
After the normalization processing of the network parameters, according to the calculation flow of the entropy weight method, the information entropy E of each network parameter is obtained j The calculation formula is as follows:
Figure BDA0003787482290000161
Figure BDA0003787482290000162
or
Figure BDA0003787482290000163
Wherein N is the number of candidate networks, M is the number of network parameters corresponding to the candidate networks, p Ij Is the ratio of the jth network parameter of the candidate network I in the candidate network parameters, when the network parameters are effectiveType parameter, then p is performed using equation 13 Ij When the network parameter is a cost-type parameter, p is calculated by using equation 14 Ij Calculation of (E), E j Is the information entropy value of the jth network parameter.
Finally, solving the objective weight s of each network parameter j The calculation formula is as follows:
Figure BDA0003787482290000164
Figure BDA0003787482290000165
wherein: p is a radical of j Coefficient of difference, s, representing the j-th network parameter j Is the weight of the jth network parameter.
Step S3603: and determining an optimal decision according to the subjective weight and the objective weight so as to determine the target network.
Wherein step S3603 specifically includes the following substeps:
step S36031: and determining the combination weight of the network parameters in the candidate network according to the subjective weight and the objective weight.
The formula W for the combined weight calculation of the jth network parameter is:
W j ＝αw j +(1-α)s j (17)
wherein: and alpha is 0.5 to represent that the subjective weight and the objective weight corresponding to the network parameters are equally important.
Step S36032: and determining the utility value of the candidate network according to the combined weight, and determining the target network according to the utility value.
Wherein the utility value C of the candidate network I i The concrete expression is as follows: :
Figure BDA0003787482290000171
wherein p is Ij Is the jth network of the candidate network IThe proportion of network parameters among the candidate network parameters,
and sequencing the utility value of each candidate network, selecting the network with the highest utility value as a target network, and switching the network to the target network.
The application has the following beneficial effects:
the network switching method and the system based on the mobile user position can improve the user position prediction precision through the network switching strategy, thereby effectively reducing the switching times and the switching failure times, reducing the switching time delay, and balancing the network load while ensuring that the user switches to a proper network.
Although the present application has been described with reference to examples, which are intended to be illustrative only and not to be limiting of the application, changes, additions and/or deletions may be made to the embodiments without departing from the scope of the application.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A network switching method based on mobile user position is characterized in that the method specifically comprises the following steps:
network scanning is carried out;
in response to the completion of network scanning, predicting the position of the mobile user, and predicting the position of the user at the next moment;
estimating the received signal strength according to the predicted position of the user at the next moment, and entering a switching trigger mechanism according to the received signal strength;
after entering a switching trigger mechanism, pre-selecting a candidate network according to the moving speed of a user and the bandwidth required by the user request service;
judging whether the candidate network is unique;
and if the candidate network is not unique, performing multi-attribute decision, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete the switching of the network.
2. The method as claimed in claim 1, wherein if the candidate network is unique, the handover is directly performed to the candidate network to complete the handover.
3. The method as claimed in claim 1, wherein the step of predicting the location of the mobile subscriber in response to the completion of the network scan to predict the location of the subscriber at the next time comprises the following sub-steps:
collecting a mobile position coordinate and a mobile speed of a user terminal to form a data set;
inputting the constructed data set into a BOLSTM model for data preprocessing, and extracting the mobile characteristics of a user;
in response to the extraction of the mobile features of the user, optimizing the LSTM model by using Bayes, determining hyper-parameters, and finishing the training and prediction of the BOLSTM model according to the determined hyper-parameters;
the trained BOLSTM model is loaded to predict the location of the user at the next time.
4. The method of claim 2, wherein the collected coordinates of the user's moving position and the user's moving speed form a user trajectory sequence, and the user trajectory sequence is T r ＝{p 1 ,p 2 ,...,p T Each moving track p of the mobile users in T Corresponding position coordinate l i And (i is more than or equal to 1 and less than or equal to T) preprocessing the data and inputting the preprocessed data into the neural network to extract the mobile features of the user.
5. The method of claim 3, wherein the handover trigger mechanism is activated in advance if the received signal strength is lower than the handover trigger threshold while the candidate network set is not empty.
6. The method for mobile subscriber location based network handover as claimed in claim 1, wherein the received signal strength is specifically expressed as:
P(dBm)＝P t -PL
P t for the transmission power of MBS or SBS, PL denotes PL (dB) MBS Or PL (dB) SBS When calculating the signal strength of the access MBS, P t For the transmit power of MBS, PL denotes PL (dB) MBS (ii) a When calculating the signal strength of the SBS, P t For the transmission power of SBS, PL represents PL (dB) SBS 。
7. The method of claim 1, wherein if the user's mobility is greater than the maximum mobility supported, the network in the candidate network set is removed, and if the bandwidth required by the user to request the service is greater than the available bandwidth of the candidate network, the candidate network is removed from the candidate network set.
8. The method of claim 1, wherein the performing of the multi-attribute decision to determine the target network among the candidate networks according to the multi-attribute decision, and the handover of the network to the target network to complete the handover specifically comprises the sub-steps of:
calculating subjective weight;
calculating an objective weight in response to completing the calculation of the subjective weight;
determining an optimal decision according to the subjective weight and the objective weight so as to determine a target network;
and switching the network to the target network to complete the switching of the network.
9. The method of claim 1, wherein the network parameter weight assignment is specifically expressed as:
Figure FDA0003787482280000021
wherein w j Weight representing the jth network parameter, c ij Indicating the relative importance of the network parameter i to the network parameter j, and n indicates the total number of network parameters.
10. A network switching system based on mobile user position is characterized in that the system specifically comprises a network scanning unit, a user position prediction unit, a switching trigger unit, a candidate network pre-selection unit, a judgment unit and a network switching unit;
a network scanning unit for performing network scanning;
the user position prediction unit is used for predicting the position of the mobile user in response to the completion of network scanning and predicting the position of the user at the next moment;
the switching trigger unit is used for estimating the received signal strength according to the predicted position of the user at the next moment and entering a switching trigger mechanism according to the received signal strength;
the candidate network pre-selection unit is used for pre-selecting a candidate network according to the moving speed of the user and the bandwidth required by the user request service after entering a switching trigger mechanism;
a judging unit, configured to judge whether the candidate network is unique; if the candidate network is unique, directly switching the network to the candidate network to complete the switching of the network;
and the network switching unit is used for carrying out multi-attribute decision if the candidate network is not unique, determining a target network in the candidate network according to the multi-attribute decision, and switching the network to the target network so as to complete network switching.
CN202210945986.4A 2022-08-08 2022-08-08 Network switching method and system based on mobile user position Pending CN115297518A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210945986.4A CN115297518A (en) 2022-08-08 2022-08-08 Network switching method and system based on mobile user position

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210945986.4A CN115297518A (en) 2022-08-08 2022-08-08 Network switching method and system based on mobile user position

Publications (1)

Publication Number Publication Date
CN115297518A true CN115297518A (en) 2022-11-04

Family

ID=83828100

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210945986.4A Pending CN115297518A (en) 2022-08-08 2022-08-08 Network switching method and system based on mobile user position

Country Status (1)

Country Link
CN (1) CN115297518A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117580132A (en) * 2024-01-16 2024-02-20 杭州海康威视数字技术股份有限公司 Heterogeneous network access method, device and equipment for mobile equipment based on reinforcement learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117580132A (en) * 2024-01-16 2024-02-20 杭州海康威视数字技术股份有限公司 Heterogeneous network access method, device and equipment for mobile equipment based on reinforcement learning
CN117580132B (en) * 2024-01-16 2024-04-12 杭州海康威视数字技术股份有限公司 Heterogeneous network access method, device and equipment for mobile equipment based on reinforcement learning

Similar Documents

Publication Publication Date Title
Singhrova et al. Vertical handoff decision algorithm for improved quality of service in heterogeneous wireless networks
CN108235390B (en) Vertical switching method based on Bayesian decision in heterogeneous wireless network
Goudarzi et al. MDP-based network selection scheme by genetic algorithm and simulated annealing for vertical-handover in heterogeneous wireless networks
Chen et al. An MDP-based vertical handoff decision algorithm for heterogeneous wireless networks
CN110809306A (en) Terminal access selection method based on deep reinforcement learning
Alkhawlani et al. Multi-criteria vertical handover by TOPSIS and fuzzy logic
Khan et al. Fuzzy based multi-criteria vertical handover decision modeling in heterogeneous wireless networks
Huang et al. Self-adapting handover parameters optimization for SDN-enabled UDN
Kaleem VHITS: vertical handoff initiation and target selection in a heterogeneous wireless network
CN105491611A (en) Wireless network switching method and system
Kunarak et al. Algorithmic vertical handoff decision and merit network selection across heterogeneous wireless networks
CN110677861A (en) Network selection method facing 5G heterogeneous network
Wang et al. A flexible and generalized framework for access network selection in heterogeneous wireless networks
CN115297518A (en) Network switching method and system based on mobile user position
CN113993168B (en) Collaborative caching method based on multi-agent reinforcement learning in fog wireless access network
CN105430693B (en) Multi-user's switching method based on service feature in a kind of 5G network
CN112492656B (en) Wireless network access point switching method based on reinforcement learning
CN103476080A (en) Self-adaption vertical handoff method based on residence time
Yang et al. A novel fuzzy logic based vertical handoff decision algorithm for heterogeneous wireless networks
Bhute et al. Vertical handover decision strategies in heterogeneous wireless networks
Sun et al. Group vertical handover in heterogeneous radio access networks
CN112672393B (en) Vertical switching method for self-adaptive terminal requirements in ultra-dense heterogeneous wireless network
He et al. A vertical handoff decision algorithm based on fuzzy control in WiMAX and TD-SCDMA heterogeneous wireless networks
Aymen et al. A fuzzy logic algorithm for RATs selection procedures
CN115278792A (en) Switching judgment method for sensing access preference of terminal in ultra-dense heterogeneous wireless network

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination