CN114710786A - Unmanned aerial vehicle base station dynamic deployment method based on user trajectory prediction - Google Patents
Unmanned aerial vehicle base station dynamic deployment method based on user trajectory prediction Download PDFInfo
- Publication number
- CN114710786A CN114710786A CN202210420046.3A CN202210420046A CN114710786A CN 114710786 A CN114710786 A CN 114710786A CN 202210420046 A CN202210420046 A CN 202210420046A CN 114710786 A CN114710786 A CN 114710786A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- user
- aerial vehicle
- track
- base station
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 230000005540 biological transmission Effects 0.000 claims abstract description 24
- 239000013598 vector Substances 0.000 claims abstract description 20
- 238000004364 calculation method Methods 0.000 claims abstract description 14
- 238000005457 optimization Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 230000007787 long-term memory Effects 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 2
- 238000004891 communication Methods 0.000 abstract description 10
- 238000012549 training Methods 0.000 description 9
- 230000033001 locomotion Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000013526 transfer learning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000005059 dormancy Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013209 evaluation strategy Methods 0.000 description 1
- 230000012010 growth Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/18—Network planning tools
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/04—Wireless resource allocation
- H04W72/044—Wireless resource allocation based on the type of the allocated resource
- H04W72/0446—Resources in time domain, e.g. slots or frames
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Strategic Management (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Marketing (AREA)
- Databases & Information Systems (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Tourism & Hospitality (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Probability & Statistics with Applications (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Remote Sensing (AREA)
- Primary Health Care (AREA)
Abstract
The invention discloses an unmanned aerial vehicle base station dynamic deployment method based on user track prediction, and belongs to the technical field of communication. The invention provides a new user track prediction method, which divides users with similar tracks into a group, shares a track prediction model, measures the track similarity between every two users by using an improved longest common sub-vector, and classifies the track similarity into a group when the track similarity reaches a set threshold; and synchronizing the predicted multi-user position of the next time slot to the unmanned aerial vehicle, and dynamically deploying the base station by the unmanned aerial vehicle according to the user position so as to achieve the maximum total transmission rate of the downlink channel. The invention realizes the rapid deployment of the unmanned aerial vehicle, avoids the estimation and calculation of the channel state with large calculation amount and long time consumption, can rapidly and dynamically adjust the position under the user moving condition, realizes better downlink data service, realizes the real-time deployment of the unmanned aerial vehicle base station under the user moving condition, and meets the actual requirement.
Description
Technical Field
The invention belongs to the technical field of communication, and particularly relates to an unmanned aerial vehicle base station dynamic deployment method based on user trajectory prediction.
Background
In the information-oriented era of the continuous rise of the internet, the communication traffic is explosively increased, and rich multimedia services put higher demands on mobile communication networks, which require wider coverage and more reliable service quality. In addition, under the influence of daily life of people, the traffic demand of the mobile network presents the characteristic of obvious time-space unevenness, and the demand of different areas and different time periods on the traffic of the mobile network has great difference. The rapid growth of mobile network traffic and the characteristics of spatial-temporal non-uniformity of traffic demand pose serious challenges for the deployment of future communication network base stations.
The existing solution is to combine Heterogeneous Network (HetNet) with base station dormancy. The micro base station is deployed on the basis of the macro base station as required, the distance between the micro base station and a user is shortened, the spectrum efficiency and the throughput are improved, the problem of traffic congestion in a hot spot area is solved, but the micro base station is still statically deployed according to the traffic peak value of the area in essence, and serious resource waste is caused in a time period with smaller traffic demand; after the base station dormancy technology is introduced, the base station is selected to be dormant or awakened according to the change of the flow demand in time, although a certain operation expenditure is reduced, the deployment and maintenance cost is still huge.
An Unmanned Aerial Vehicle (UAV) has the advantages of being high in flexibility, convenient to deploy, low in expenditure, capable of achieving line-of-sight communication and the like, carries a base station to conduct auxiliary communication, provides possibility for establishing a new mobile network, and provides a new idea for solving the problems. However, unmanned aerial vehicle deployment under the condition of multiple unmanned aerial vehicle services and multiple users is a non-convex and NP-difficult problem, the traditional communication method is difficult to solve, meanwhile, as the users are in the middle of continuous movement, the static unmanned aerial vehicle base station deployment is difficult to provide better services, the unmanned aerial vehicle base station position is adjusted in real time according to the movement condition of the users, and dynamic deployment of the unmanned aerial vehicle base station is an excellent solution, but the solution has certain challenges.
Most of the existing multi-unmanned aerial vehicle base station dynamic deployment methods do not consider actual conditions, user movement is modeled by using mathematics, and real data is not adopted. And a few deployment methods using actual mobile data ignore the hysteresis of synchronizing user position information to the unmanned aerial vehicle, so that the deployment scheme obtained by solving is not the current optimal solution. In order to solve this problem, it has been proposed to predict the user trajectory and then adjust the base station position using the predicted user position. However, the method adopts single-user prediction, a prediction model is trained and stored for each user, the workload is large, and the actual deployment is impractical.
Disclosure of Invention
Aiming at the problems, in order to meet the user requirements under the condition of uneven flow space-time distribution, the invention provides a deployment method of an unmanned aerial vehicle base station based on user track prediction, which is used for dynamically deploying the base station according to the position of a user so as to achieve the maximum downlink transmission rate under the constraint condition. Meanwhile, the invention provides a new user track prediction method in consideration of the mobility of the user and the lag of synchronizing the user position information to the unmanned aerial vehicle, and the base station position is calculated in one step based on the predicted position so as to realize the real-time deployment of the unmanned aerial vehicle base station under the user moving condition and meet the actual requirement.
The invention discloses a dynamic deployment method of an unmanned aerial vehicle base station based on user track prediction, which is characterized in that the user position distribution of the next time slot is obtained in advance by using the prediction of the user track, and the optimal position of the unmanned aerial vehicle base station in the deployment of the next time slot is calculated on the basis of the distribution. Specifically, the method comprises the following steps:
s1: and constructing a system model, and providing downlink service for the mobile user by using the unmanned aerial vehicle carrying the base station.
S2: and training a user track prediction model in advance, and when a user enters the network service range of the unmanned aerial vehicle base station, triggering a track prediction algorithm to predict the next time slot position of the user.
And (3) pre-training and storing the user track prediction model, and directly predicting the position of the next time slot by using the track prediction model of the group corresponding to the user when calling.
The method divides users with similar tracks into a group and shares one track prediction model. The similarity of the traces between two users is measured using VLCSS (longest common subvector). The inter-user track similarity is obtained by calculating the ratio of the longest common sub-vector of the two tracks to the number of the sub-vectors of the longer track. And grouping the users by adopting a DBSCAN clustering algorithm, and classifying the users with the track similarity reaching a set threshold into a group. And each group shares one track prediction model, and the track prediction model of each group is trained.
S3: and according to the user position distribution obtained by prediction, the optimal deployment position of the multiple unmanned aerial vehicle base stations is obtained by taking the maximum total transmission rate of the downlink channel as an optimization target.
S4: and each unmanned aerial vehicle moves to the calculated optimal deployment position.
In step S1, the unmanned aerial vehicle uses a Frequency Division Multiple Access (FDMA) mode to serve the Ground users, and an Air-to-Ground (AtG) transmission channel is formed between the Ground users, and the path loss model of the unmanned aerial vehicle uses a hybrid model.
In step S2, the trajectory prediction model uses a long-term and short-term memory network (LSTM) as a basic structure, and adopts an automl (auto Machine learning) mode to realize automation of the hyperparticipation selection process, and a high-performance LSTM can be generated for a given task without manual intervention, specifically: the LSTM is used as a student predictor, the Q-learning is used as a controller, a better framework is provided for the student LSTM, and a migration learning unit is added to accelerate the searching process of the framework.
In step S3, the method for dynamically deploying an unmanned aerial vehicle base station uses the user location as the periodic input and the deployment location of the unmanned aerial vehicle base station as the output, and specifically includes: and calculating the center of the user cluster by adopting an iterative K-means algorithm to obtain the initial position of the base station of the unmanned aerial vehicle, constructing a search space around the position, and searching for the optimal position in the space by a fixed step length to enable the total transmission rate of all the unmanned aerial vehicles to be maximum.
In step S4, after the unmanned aerial vehicle obtains the optimal position of the next time slot, the unmanned aerial vehicle moves in the manner of the minimum update distance.
Compared with the prior art, the invention has the beneficial effects that:
(1) the method provides a new track similarity measurement method in the process of realizing track clustering, and solves the defect that the directionality of the moving track of a user is not considered by an LCSS (longest common subsequence) algorithm. According to the invention, the similarity among the tracks is calculated through an improved algorithm based on the longest common subsequence, the user tracks are clustered, each group shares one track prediction model, and the training quantity and the calculation quantity of different models are reduced.
(2) The method provided by the invention provides a user track prediction method, and realizes the balance of prediction precision and calculation storage resources. The method of the invention predicts the positions of multiple users in the service scene of the multiple unmanned aerial vehicle base stations in advance at each time slot, synchronizes the positions of the users to the unmanned aerial vehicles, and further calculates the optimal deployment positions of the multiple unmanned aerial vehicle base stations, thereby solving the problem that the deployment scheme obtained by solving is not the current optimal solution because the lag of synchronizing the position information of the users to the unmanned aerial vehicles is ignored in the prior art; the problem that the workload is large and the method is unrealistic due to the fact that a prediction model is trained and stored in advance for each user in the prior art is solved.
(3) The method adopts automatic parameter selection for the LSTM track prediction model, avoids introducing artificial errors and enables the prediction result to be more accurate; meanwhile, the training time is reduced by adopting transfer learning, users with similar tracks are clustered, a prediction model is shared, and the model training calculated amount and the storage amount are reduced.
(4) When the optimal deployment position of the multi-unmanned aerial vehicle base station is obtained in each time slot, the method firstly obtains the predicted position of each user, obtains the initial unmanned aerial vehicle position according to the minimum communication distance between the ground user and the unmanned aerial vehicle, then carries out fine adjustment, enables the total transmission rate of all the unmanned aerial vehicles to be maximum, and finally obtains the optimal deployment position of the multi-unmanned aerial vehicle. The method realizes the rapid deployment of the unmanned aerial vehicle by directly taking the positions of multiple users as input and the positions of the base stations of multiple unmanned aerial vehicles as output, avoids the estimation and calculation of the channel state with large calculation amount and long time consumption, can rapidly and dynamically adjust the positions under the condition of user movement, and realizes better downlink data service.
Drawings
In order to illustrate embodiments of the invention or solutions in the prior art more clearly, the drawings that are needed in the embodiments will be briefly described below, so that the features and advantages of the invention will be more clearly understood by referring to the drawings that are schematic and should not be understood as limiting the invention in any way, and other drawings may be obtained by those skilled in the art without inventive effort. Wherein:
FIG. 1 is a schematic view of a scene architecture model of an unmanned aerial vehicle base station deployment method based on user trajectory prediction according to the present invention;
FIG. 2 is a schematic diagram of calculating inter-track similarity provided by the present invention;
FIG. 3 is a schematic diagram of the neural architecture automatic search (NAS) employed by the present invention;
fig. 4 is a schematic flowchart of the method for deploying the base station of the unmanned aerial vehicle based on user trajectory prediction according to the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention, taken in conjunction with the accompanying drawings and detailed description, is set forth below. It should be noted that the embodiments of the present invention and features of the embodiments may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced otherwise than as specifically described herein and, therefore, the scope of the present invention is not limited by the specific embodiments disclosed below.
In the existing unmanned aerial vehicle base station deployment methods, dynamic deployment, particularly rapid deployment methods are few, and most of the existing methods do not consider user movement in a real scene. The invention provides an unmanned aerial vehicle base station dynamic deployment method based on user track prediction, which takes user distribution obtained by prediction as periodic input, achieves the maximum downlink transmission rate, and outputs the optimal position of an unmanned aerial vehicle base station so as to realize the dynamic rapid deployment of the unmanned aerial vehicle base station.
As shown in fig. 1, a scene architecture model of the unmanned aerial vehicle base station deployment method based on user trajectory prediction of the present invention is: the unmanned aerial vehicle carries on the basic station, for the ground user is for providing down service, the target is under the condition that satisfies every ground user demand, and developments deployment unmanned aerial vehicle promptly acquires the best position of unmanned aerial vehicle at every time slot for total transmission rate is the biggest, specifically as follows:
a plurality of unmanned aerial vehicles are deployed in a target area ([0, L ] × [0.M ]) to provide wireless communication service for ground users, N users are arranged in the area, K unmanned aerial vehicles are arranged, and the unmanned aerial vehicles adopt an orthogonal frequency division multiple access technology.
The position of the ground user n at time t is denoted as un(t)=[xn(t),yn(t)](ii) a The position of drone k at time t is denoted pk(t)=[xk(t),yk(t)](ii) a The communication distance between the ground user n and the drone k is then denoted asSo that the distance between each drone k and all users isWherein h is the flight altitude of unmanned aerial vehicle k.
The link between unmanned aerial vehicle and ground user is the air-to-ground channel, and the channel model is mixed model, is formed by the combination of line of sight transmission (LoS) and non-line of sight transmission (NLoS), and the formula is expressed as:
PL=PLLoS×PrLoS+PLNLoS×(1-PrLoS)
wherein PLLoSIndicating line-of-sight path loss, PLNLoSRepresenting non-line-of-sight path loss; pr (Pr) ofLoSIs the line-of-sight link probability between drone k and user n,where θ is the elevation angle between the user and the drone, and a and b are environmental parameters.
Line-of-sight path loss and non-line-of-sight path loss can be expressed according to a path loss model as:
wherein, muLoSAnd muNLoSThe path loss coefficients respectively represent a line-of-sight link and a non-line-of-sight link, alpha is a line loss index, f is a carrier frequency, c is an optical speed, and d is a path length.
To sum up, the channel gain g from the drone to the user can be calculated as:
in order to evaluate the advantages and disadvantages of unmanned aerial vehicle base station deployment, the invention takes the total transmission rate of a downlink channel as a measurement standard. If the unmanned plane k serves the user n, the transmission rate r of the channel is determined according to the Shannon formulak,nCan be expressed as:
wherein, Bk,nThe bandwidth allocated to user n for drone k,m is the number of orthogonal sub-channels, and B is the bandwidth of the unmanned plane k; p is a radical ofk,nRepresenting the transmission power of drone k for user n; gk,nRepresenting the channel gain of drone k to user n; sigma2Is gaussian white noise power.
Assuming that each timeslot drone can fly to the desired location, the deployment can be considered as a static deployment for each timeslot, and the optimization objective can be expressed as:
wherein R issumRepresenting the total transmission rate of all drones within the scene; rhok,nRepresenting the relevance of the user and the unmanned aerial vehicle, representing whether the user n is accessed to the unmanned aerial vehicle k, and if so, rhok,nTaking the value as 1, otherwise, taking 0; r is0The minimum rate that needs to be met for a single user to access the network.
In order to reduce the calculation and storage amount of a user track prediction model, the invention divides users with similar tracks into groups and shares the prediction model.
As shown in fig. 2, the present invention provides a new algorithm for measuring similarity between tracks based on the improvement of the lcs (longest common subsequence) algorithm, which includes the following steps for the directionality of the track into the calculation range:
let two users u1 and u2 have their trajectories T, respectively1And T2The track is composed of a plurality of track segments and is represented as a vector set; for the track T1And T2First, the number of its common subvectors VLCSS (T) is calculated1,T2) The following were used:
wherein v isu1(i) And vu2(j) Respectively represent the track T1I vector segments and trajectories T2J vector segments;representing an empty set; gamma is a threshold constant; sigma (v)u1(i),vu2(j) Represents two subvectors vu1(i) And vu2(j) The distance between them is calculated as follows:
σ(vu1(i),vu2(j))=σθ(vu1(i),vu2(j))·σd(vu1(i),vu2(j))
wherein σθIs the angle between the vectors, σdIs the vector sum of the distances of the two end points.
VLCSS(vu1(i),vu2(j) Is a vector segment v of a computational two-useru1(i) And vu2(j) The number of common subvectors.
VLCSS(vu1(i-1),vu2(j-1)、VLCSS(vu1(i-1),vu2(j) And VLCSS (v)u1(i),vu2(j-1)) are the common subvector numbers for the corresponding vector segments for the two users, respectively.
Wherein the content of the first and second substances,respectively represent the track T1And T2The number of vector segments.
In the embodiment of the invention, the user track Clustering adopts a Density-Based Clustering algorithm-DBSCAN (Density-Based Spatial Clustering of Applications with Noise) Clustering algorithm, and the algorithm derives the sample sets with the maximum Density connection according to the Density reachable relation to form a Clustering cluster. The density reachable relation in the invention is represented by DVLCSSGreater than a certain threshold.
As shown in FIG. 3, the present invention employs a Neural Architecture Search (NAS) framework to automatically determine network parameters for an LSTM network. The NAS framework specifically includes:
and searching a network structure in the constructed search space by using a certain search strategy, triggering a performance evaluation strategy to evaluate along with the network performance, and finally feeding back an evaluation result to the search strategy. So iterate, search for the most network architecture automatically.
As shown in fig. 4, a method for predicting dynamic deployment of drone base station based on user trajectory according to the present invention aims to dynamically deploy UAV (obtain the best position of UAV in each timeslot) to maximize RsumWhile meeting the requirements of each ground user, comprising the following steps:
s1: the user trajectory prediction comprises the following sub-steps S1-1 to S1-5. The user has a history track within the region.
S1-1: and intercepting the user track. The method and the device have the advantages that the calculation complexity of the similarity between the tracks is high, the calculation time is consumed, the user tracks are intercepted according to the time periods, the tracks with the time length T are intercepted, and the calculation complexity is reduced.
S1-2: and calculating the similarity between the tracks. The invention adopts the improved VLCSS, improves on the basis of LCSS (longest common subsequence), considers the directionality of the track, replaces the distance between point points by the distance between vector segments, and calculates the ratio of the longest common subvector to the longer track subvector, thereby obtaining the similarity between the tracks.
S1-3: and (6) clustering tracks. And adopting a DBSCAN clustering algorithm, and regarding the condition that the similarity among the tracks reaches a certain threshold value as that the density is reachable, thereby completing track clustering and realizing the division of user groups. Inputting: the system comprises a database containing N user track objects, a similarity threshold parameter epsilon and a minimum number MinPts of users in each track similarity group; and (3) outputting: all users meeting the density requirement have similar tracks.
The group division can be realized through the steps S1-1 to S1-3, users with similar tracks are regarded as a group, and a track prediction model is shared, so that the model training calculation amount and the storage resource are saved. The track prediction model takes a long-term memory network (LSTM) as a basic structure, the LSTM can solve the problem of medium-term and long-term dependence of a neural network, and is widely used for track prediction, and the next time slot position is predicted by taking a previous track point of a user as input.
S1-4: and realizing super-parameter optimization by using automatic machine learning (AutoML). By adopting the NAS architecture, the LSTM is used as a student predictor, the Q-learning is used as a controller, parameters are automatically selected for the LSTM, a better network architecture is searched, and manual errors are avoided. A migration learning (TL) unit is added to accelerate the search process of the architecture.
And constructing an action space and a parameter space, and taking the accuracy of the LSTM network as an evaluation feedback signal. Setting an early termination mechanism, stopping searching when the LSTM reaches the prediction precision of 90% (which can be flexibly set), otherwise, terminating after the whole space is searched; the parameter space includes: the number of hidden layers, the number of neurons in each hidden layer, the exit rate and the like; the ReLU function is used as the activation function.
S1-5: and the framework searching process is accelerated by adopting transfer learning. The two LSTMs have similar system structures in the aspects of layers and connectivity, and transfer learning is utilized to transfer the training knowledge at the iteration t-1 moment to the iteration t moment, so that the training time is shortened, and rapid convergence is achieved.
The training of the user trajectory prediction model can be completed from step S1-4 to step S1-5, and the next slot position can be output through historical trajectory prediction by the model, so as to obtain the user position distribution.
S2: the dynamic deployment of the unmanned aerial vehicle base station comprises the following substeps S2-1-S2-3.
S2-1: k-means clustering algorithm. By K unmanned aerial vehicles in the region, each time slot gathers users into K clusters, and each user cluster adopts one unmanned aerial vehicle to serve. By using K-means algorithm, K user clusters can be obtained and cluster center positions can be obtained, namely D is obtainedsumAnd when the minimum value is smaller, the position of the unmanned aerial vehicle is taken as the initial position of the unmanned aerial vehicle.
And predicting the position of the next time slot for each user in the service area by using the track prediction model of the corresponding group, and then synchronizing the position to the unmanned aerial vehicle. And the unmanned aerial vehicle carries out clustering by using a K-means algorithm based on the obtained multi-user position distribution to obtain the initial positions of the K unmanned aerial vehicles.
S2-2: local regulation and optimization. The transmission rate is not only related to the distance, and on the basis of the initial position of the unmanned aerial vehicle obtained in S2-1, a search space is established around the initial position as the center, searching is carried out with a fixed step length, and when the transmission rate reaches the maximum, the unmanned aerial vehicle is the optimal deployment position.
S2-3: the minimum update distance moves. And the unmanned aerial vehicle moves from the t-1 time slot position to the t time slot position by adopting a method of minimum updating distance.
The method for minimum updating distance essentially comprises the steps of pairing unmanned aerial vehicle position sets at two moments, recursively calling, and solving
Wherein (x)t,yt) Representing the deployment position of the t-slot drone, (x)t-1,yt-1) Indicating the deployment position of the t-1 slot drone, Dt-1,tRepresents the movement distance of the front slot unmanned aerial vehicle and the rear slot unmanned aerial vehicle, (x)t-1,yt-1),(xt,yt) And e (L, M) represents that the unmanned aerial vehicle is in the target area before and after moving.
Through the steps, the user trajectory prediction can be completed to obtain the user position distribution, and the unmanned aerial vehicle base station deployment is maximized based on the transmission rate completed by the user distribution.
In addition to the technical features described in the specification, the technology is known to those skilled in the art. Descriptions of well-known components and techniques are omitted so as to not unnecessarily obscure the present invention. The embodiments described in the above embodiments do not represent all embodiments consistent with the present application, and various modifications or variations which may be made by those skilled in the art without inventive efforts based on the technical solution of the present invention are still within the protective scope of the present invention.
Claims (8)
1. A dynamic deployment method of an unmanned aerial vehicle base station based on user track prediction is applied to a scene that an unmanned aerial vehicle carries a base station to provide downlink service for a mobile user, and is characterized by comprising the following steps:
(1) calculating the similarity of the tracks among the users, and dividing the users with similar tracks into a group; the inter-user track similarity is obtained by calculating the ratio of the longest common sub-vector of the two tracks to the number of the sub-vectors of the longer track;
(2) each group shares one track prediction model, and the track prediction model of each group is trained;
(3) calling a track prediction model of a corresponding group for each user in the network service area of the unmanned aerial vehicle base station, and predicting the position of the next time slot;
(4) synchronizing the position distribution of multiple users in the service area to the unmanned aerial vehicle, and maximizing the total transmission rate of the downlink channel by the unmanned aerial vehicle as an optimization target to obtain the optimal deployment position of the unmanned aerial vehicle base station in the next time slot;
(5) and each unmanned aerial vehicle moves to the optimal deployment position.
2. The method according to claim 1, wherein in step (1), the similarity between the tracks is calculated by using an improved longest common subsequence-based algorithm, which takes the directionality of the tracks into a calculation range, and the method comprises:
let the trajectories of users u1 and u2 be T respectively1And T2Calculating the trajectory T according to1And T2Number of common subvectors of (a):
wherein v isu1(i) Representing a track T1I vector segments, vu2(j) Representing a track T2J vector segments;representing an empty set; gamma is a threshold constant; sigma (v)u1(i),vu2(j) Represents two subvectors v)u1(i) And vu2(j) The distance between the two is calculated as follows:
σ(vu1(i),vu2(j))=σθ(vu1(i),vu2(j))·σd(vu1(i),vu2(j))
wherein σθIs the angle between the vectors, σdIs the sum of the distances of two end points of the vector; VLCSS (v)u1(i),vu2(j) Is a calculated vector segment vu1(i) And vu2(j) The number of common subvectors of (a);
3. The method according to claim 1 or 2, wherein in step (1), the user tracks are clustered by a density-based clustering algorithm DBSCAN, the clustering algorithm derives the maximum density connected sample sets from a density reachable relation, and forms a cluster, wherein the density reachable relation is the similarity D between the user tracksVLCSSGreater than a preset threshold.
4. The method according to claim 1 or 2, wherein in the step (1), the step of grouping the users comprises:
s1-1: acquiring a historical track of a user in a service area, and intercepting a track with required duration from the historical track;
s1-2: calculating the similarity between tracks by adopting an improved longest common subsequence-based algorithm;
s1-3: clustering user tracks by adopting a density-based clustering algorithm DBSCAN, regarding the condition that the similarity among the tracks reaches a set threshold as that the density is reachable, and setting the minimum number of users of each track similar group; and outputting all user track similar groups meeting the density requirement through clustering.
5. The method as claimed in claim 1 or 2, wherein in the step (2), the trajectory prediction model takes a long-term memory network LSTM as a basic structure, and adopts an automatic machine learning AutoML mode to automatically select the hyper-parameters, specifically, the LSTM is taken as a student predictor, the Q-learning is taken as a controller, parameters are automatically selected for the student LSTM, a better network architecture is searched, and a migration learning unit is used to accelerate the network architecture searching process.
6. The method of claim 1 or 2, wherein the step (4) comprises:
s2-1: setting K unmanned aerial vehicles in a service area, after predicting the position of a user, clustering the user into K clusters by using a K-means algorithm, wherein each cluster is served by one unmanned aerial vehicle; the obtained center position of each cluster is the distance D between each unmanned aerial vehicle and all userssumA position of the drone at a minimum;
s2-2: searching with fixed step around the unmanned aerial vehicle position obtained by S2-1 as the center to obtain the total transmission rate R of all unmanned aerial vehiclessumAnd the position of the unmanned aerial vehicle when the maximum position is reached is the optimal deployment position of the unmanned aerial vehicle.
7. The method according to claim 1 or 2, wherein in the step (5), the drone moves with a minimum update distance after obtaining the optimal deployment position of the next time slot.
8. The method according to claim 1 or 2, wherein in the step (5), there are N users in the service area, K drones, and the drones adopt orthogonal frequency division multiple access;
the optimization target of maximizing the total transmission rate of the downlink channel is represented as:
wherein R issumRepresenting the total transmission rate of all drones within the scene; rhok,nIndicating whether the user n is accessed to the unmanned aerial vehicle k or not, and if so, rhok,nTaking the value as 1, otherwise, taking 0; r isk,nChannel transmission rate when serving user n for drone k; r is0Minimum rate required to be met for a single user to access the network;
channel transmission rate rk,nThe calculation is as follows:
wherein, Bk,nThe bandwidth allocated to user n for drone k,m is the number of orthogonal sub-channels, and B is the bandwidth of the unmanned plane k; p is a radical ofk,nRepresenting the transmission power of drone k for user n; gk,nRepresenting the channel gain of drone k to user n; sigma2Is gaussian white noise power.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210420046.3A CN114710786B (en) | 2022-04-20 | 2022-04-20 | Unmanned aerial vehicle base station dynamic deployment method based on user track prediction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210420046.3A CN114710786B (en) | 2022-04-20 | 2022-04-20 | Unmanned aerial vehicle base station dynamic deployment method based on user track prediction |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114710786A true CN114710786A (en) | 2022-07-05 |
CN114710786B CN114710786B (en) | 2023-05-30 |
Family
ID=82173711
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210420046.3A Active CN114710786B (en) | 2022-04-20 | 2022-04-20 | Unmanned aerial vehicle base station dynamic deployment method based on user track prediction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114710786B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115802362A (en) * | 2022-08-18 | 2023-03-14 | 电子科技大学 | Unmanned aerial vehicle-assisted wireless network deployment method based on autonomous learning |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107205225A (en) * | 2017-08-03 | 2017-09-26 | 北京邮电大学 | The switching method and apparatus for the unmanned aerial vehicle onboard base station predicted based on user trajectory |
US20190057169A1 (en) * | 2017-02-22 | 2019-02-21 | Middle Chart, LLC | Building Vital Conditions Monitoring |
CN111193536A (en) * | 2019-12-11 | 2020-05-22 | 西北工业大学 | Multi-unmanned aerial vehicle base station track optimization and power distribution method |
US20200260404A1 (en) * | 2019-02-12 | 2020-08-13 | Samsung Electronics Co., Ltd. | Method and system for positioning low altitude platform station (laps) drone cells |
CN114339842A (en) * | 2022-01-06 | 2022-04-12 | 北京邮电大学 | Method and device for designing dynamic trajectory of unmanned aerial vehicle cluster under time-varying scene based on deep reinforcement learning |
-
2022
- 2022-04-20 CN CN202210420046.3A patent/CN114710786B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190057169A1 (en) * | 2017-02-22 | 2019-02-21 | Middle Chart, LLC | Building Vital Conditions Monitoring |
CN107205225A (en) * | 2017-08-03 | 2017-09-26 | 北京邮电大学 | The switching method and apparatus for the unmanned aerial vehicle onboard base station predicted based on user trajectory |
US20200260404A1 (en) * | 2019-02-12 | 2020-08-13 | Samsung Electronics Co., Ltd. | Method and system for positioning low altitude platform station (laps) drone cells |
CN111193536A (en) * | 2019-12-11 | 2020-05-22 | 西北工业大学 | Multi-unmanned aerial vehicle base station track optimization and power distribution method |
CN114339842A (en) * | 2022-01-06 | 2022-04-12 | 北京邮电大学 | Method and device for designing dynamic trajectory of unmanned aerial vehicle cluster under time-varying scene based on deep reinforcement learning |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115802362A (en) * | 2022-08-18 | 2023-03-14 | 电子科技大学 | Unmanned aerial vehicle-assisted wireless network deployment method based on autonomous learning |
CN115802362B (en) * | 2022-08-18 | 2024-05-24 | 电子科技大学 | Unmanned aerial vehicle auxiliary wireless network deployment method based on autonomous learning |
Also Published As
Publication number | Publication date |
---|---|
CN114710786B (en) | 2023-05-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111786713B (en) | Unmanned aerial vehicle network hovering position optimization method based on multi-agent deep reinforcement learning | |
Liu et al. | Trajectory design and power control for multi-UAV assisted wireless networks: A machine learning approach | |
Wu et al. | Optimal UAV caching and trajectory in aerial-assisted vehicular networks: A learning-based approach | |
Shi et al. | Drone-cell trajectory planning and resource allocation for highly mobile networks: A hierarchical DRL approach | |
Saxena et al. | Optimal UAV base station trajectories using flow-level models for reinforcement learning | |
CN109447275B (en) | Switching method based on machine learning in UDN | |
Thrane et al. | Deep learning-based signal strength prediction using geographical images and expert knowledge | |
Liu et al. | Artificial intelligence aided next-generation networks relying on UAVs | |
Qiu et al. | Placement optimization of aerial base stations with deep reinforcement learning | |
CN112511197B (en) | Unmanned aerial vehicle auxiliary elastic video multicast method based on deep reinforcement learning | |
Peng et al. | A predictive on-demand placement of UAV base stations using echo state network | |
Lin et al. | Anticipatory mobility management by big data analytics for ultra-low latency mobile networking | |
Zhang et al. | Cooperative trajectory design of multiple uav base stations with heterogeneous graph neural networks | |
Ji et al. | Trajectory and communication design for cache-enabled UAVs in cellular networks: A deep reinforcement learning approach | |
CN114980169A (en) | Unmanned aerial vehicle auxiliary ground communication method based on combined optimization of track and phase | |
CN114374981B (en) | Energy-saving on-demand pre-deployment method of communication unmanned aerial vehicle | |
Hu et al. | An intelligent UAV deployment scheme for load balance in small cell networks using machine learning | |
Guo et al. | Exploiting future radio resources with end-to-end prediction by deep learning | |
Tarekegn et al. | Deep-reinforcement-learning-based drone base station deployment for wireless communication services | |
CN114710786A (en) | Unmanned aerial vehicle base station dynamic deployment method based on user trajectory prediction | |
CN115499921A (en) | Three-dimensional trajectory design and resource scheduling optimization method for complex unmanned aerial vehicle network | |
Nasr-Azadani et al. | Single-and multiagent actor–critic for initial UAV’s deployment and 3-D trajectory design | |
Chaalal et al. | A new framework for multi-hop ABS-assisted 5G-networks with users’ mobility prediction | |
Parvaresh et al. | A continuous actor–critic deep Q-learning-enabled deployment of UAV base stations: Toward 6G small cells in the skies of smart cities | |
CN114885388A (en) | Multi-service type self-adaptive switching judgment method combined with RSS prediction |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |