CN113163332A - Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning - Google Patents
Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning Download PDFInfo
- Publication number
- CN113163332A CN113163332A CN202110449681.XA CN202110449681A CN113163332A CN 113163332 A CN113163332 A CN 113163332A CN 202110449681 A CN202110449681 A CN 202110449681A CN 113163332 A CN113163332 A CN 113163332A
- Authority
- CN
- China
- Prior art keywords
- unmanned aerial
- aerial vehicle
- node
- information
- nodes
- 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 37
- 238000004040 coloring Methods 0.000 title claims abstract description 34
- 238000013480 data collection Methods 0.000 title claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims abstract description 33
- 230000033001 locomotion Effects 0.000 claims abstract description 24
- 238000005265 energy consumption Methods 0.000 claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000005259 measurement Methods 0.000 claims description 21
- 230000005540 biological transmission Effects 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 6
- 238000005562 fading Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 230000003321 amplification Effects 0.000 claims description 4
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 4
- 239000003086 colorant Substances 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims 1
- 230000009466 transformation Effects 0.000 claims 1
- 238000004891 communication Methods 0.000 abstract description 14
- 238000010586 diagram Methods 0.000 description 3
- 241000854291 Dianthus carthusianorum Species 0.000 description 2
- 101100172132 Mus musculus Eif3a gene Proteins 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241001331845 Equus asinus x caballus Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000021715 photosynthesis, light harvesting Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/021—Services related to particular areas, e.g. point of interest [POI] services, venue services or geofences
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/029—Location-based management or tracking services
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/06—Selective distribution of broadcast services, e.g. multimedia broadcast multicast service [MBMS]; Services to user groups; One-way selective calling services
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/30—Services specially adapted for particular environments, situations or purposes
- H04W4/40—Services specially adapted for particular environments, situations or purposes for vehicles, e.g. vehicle-to-pedestrians [V2P]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W74/00—Wireless channel access
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W84/00—Network topologies
- H04W84/18—Self-organising networks, e.g. ad-hoc networks or sensor networks
-
- 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)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning. The problem of throughput and energy consumption in the unmanned aerial vehicle auxiliary sensor network is mainly solved. The method comprises the following steps: model establishment is carried out on the unmanned aerial vehicle auxiliary sensor network, a homogeneous network model is provided, and nodes in the network have mobility. And (3) continuously optimizing the trajectory of the unmanned aerial vehicle by using a metric learning method, training a metric matrix off line, and constructing the metric matrix by using an LMNN algorithm in an off-line training stage. The network can be represented by a graph, each point in the graph represents a node in the network, and an edge connecting two points represents packet communication of the two points, so that the channel allocation problem of the multi-hop network can be converted into a graph coloring problem. In the unmanned aerial vehicle auxiliary network supporting fast movement, the flight speed of the unmanned aerial vehicle is high, and the ground node can be a mobile robot with mobility, so that a contact time prioritized channel access method is adopted.
Description
Technical Field
The invention belongs to the field of unmanned aerial vehicle auxiliary sensor networks, and particularly relates to a road sign map coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning.
Background
With the rapid development of electronics, sensors and communication counting, unmanned aerial vehicles are widely used in various military and civil fields. The unmanned aerial vehicle is applied to the wireless sensor network, so that sensor data can be effectively acquired, the service life of the network is prolonged, and the energy consumption of the network is reduced. In unmanned aerial vehicle assisted sensor networks (UWSNs), a channel access algorithm is very important, and affects not only system performance but also energy efficiency of sensor nodes. Therefore, designing an efficient channel access algorithm for UWSNs is a problem worthy of research.
In unmanned aerial vehicle auxiliary sensor network, because unmanned aerial vehicle all has other tasks usually, so unmanned aerial vehicle all flies according to fixed flight path, accomplishes data acquisition's task in the in-process of flying by the way. With the development of unmanned aerial vehicle technology, a flight base station with an unmanned aerial vehicle as a data mule is developed, and the unmanned aerial vehicle can be specially used for data collection, so that how to effectively utilize the high-speed mobility of the unmanned aerial vehicle to plan reasonable flight trajectory to collect data and improve throughput becomes a problem to be solved, the planning of reasonable flight path requires the unmanned aerial vehicle to collect information of nodes of the whole network, and the network is expanded from a single-hop star network to a multi-hop network, so that higher requirements are provided for a channel access algorithm of an unmanned aerial vehicle auxiliary network.
In past research, researchers have proposed a series of wireless channel access algorithms, which can be roughly classified into two categories, contention-based channel access algorithms and assignment-based channel access algorithms. Since there is no dependency on any topology or synchronization information, the contention-based channel access algorithm (e.g., CSMA) is robust to topology dynamics, but due to collisions caused by contention, the performance of the protocol may be greatly degraded in a high contention environment. In contrast, an allocation-based channel access algorithm (e.g., TDMA) utilizes synchronicity between neighboring nodes to achieve collision-free transmission by allocating transmission time slots to each node, and although an allocation-based protocol can generally guarantee better network performance in a high-contention environment, the protocol performance is negatively affected by the inability to reuse time slots and the overhead associated with synchronization in a low-contention environment. How to dynamically balance CSMA and TDMA is therefore a considerable problem.
Unmanned Aerial Vehicle (UAV) wireless communications has received widespread attention in military and civilian applications due to its high mobility, low cost, on-demand deployment and inherent line-of-sight air-to-ground access, and integration with fourth generation mobile communications system (5G) networks. Communication between unmanned aerial vehicle and the ground equipment has constituted ground-to-air communication system, can roughly divide into two types according to the effect difference: the FANET formed by the unmanned aerial vehicle cluster transmits data to the ground base station and the unmanned aerial vehicle auxiliary network, and the unmanned aerial vehicle auxiliary network uses the unmanned aerial vehicle as the aerial base station to collect the data of the ground equipment. The channel model of the ground-air communication system can use a free space propagation model, which means that the traditional channel access algorithm can be simply applied in the ground-air communication system. Line of sight (LOS) communication components dominate air-to-ground channels in many practical scenarios, particularly in rural areas or at moderate altitude drone altitudes. Such channel characteristics allow the channel state information to be determined directly from the location of each node, facilitating the design of high-speed communication systems.
Disclosure of Invention
In order to solve the problems of low throughput performance of an auxiliary network of a fast moving unmanned aerial vehicle and insufficient flight energy consumption of the unmanned aerial vehicle, the embodiment of the invention adopts a metric learning method to plan a path and optimize the throughput, the method obtains a metric matrix through off-line training, and predicts the optimal information of the throughput and the energy consumption performance by using the metric matrix on line. The metric learning method is used for planning the flight trajectory of the unmanned aerial vehicle, the graph coloring method is used for channel access optimization, and the method comprises the following steps:
and establishing a model according to the unmanned aerial vehicle auxiliary sensor network, and applying a network model, an energy consumption model and a movement model to the unmanned aerial vehicle auxiliary sensor network model.
Specifically, ground robot groups move in a region, and unmanned aerial vehicle base stations are unique and move freely in the region. The robot in unmanned aerial vehicle basic station coverage area R is the robot that needs carry out the channel access with unmanned aerial vehicle, and the robot of the outer R of unmanned aerial vehicle coverage area transmits to the basic station through jumping, consequently need not carry out the channel access with unmanned aerial vehicle. Due to the fast mobility of the robot, and the fast mobility of the drone, nodes within the drone coverage R will typically come out of coverage, while nodes outside coverage will enter the drone coverage. The coverage area R of the drone is related to the flying height h of the drone. In the network model, the node and the base station are mobile, the base station is in a constantly moving state in the area, and the node can move in the fixed area to detect the nearby environment information. In this model, energy is required for both the nodes to transmit and receive data. The energy consumption of the transmitting end is related to the data size, the transmission distance and the energy consumption of the power amplifier, and the energy consumption of the receiving end is related to the received data and the transmission distance. If the distance between the transmitter and the receiver is less than a threshold value d0Then the free space model is used. If not, a multipath fading channel model is adopted.
Where k is the packet size in bits, d is the distance between two nodes,
and
is the energy dissipation of the transmitter and receiver circuits that each node operates individually. EpsilonfsIs the signal amplification factor, ε, of a free space channel modelmpIs the signal amplification factor of the multipath fading channel model. d0The boundary condition threshold for distinguishing the two models is:
among the movement models, the group movement model we use is a reference point group movement model. In the reference point group movement model, the network is divided into a plurality of groups. For each group, there is one target in the group, and the nodes in the group move according to their targets and maintain certain constraints. The network is divided into several groups according to the requirement, the speed of the nodes in the group is controlled between 0 and the maximum speed, and the direction of the nodes is controlled between 0 and 2 pi. This allows the nodes within the group to maintain restricted random motion due to the presence of the target point within the group. The set of reference points is moved in the model. There is a reference point within each group and each time the reference point of a group member moves to a new location, the group member also moves to a randomly selected location within a circular neighborhood of radius R around its new reference point location. At the same time, at the center of each panel coverage area is a logical guide point whose motion defines the motion of the entire panel, including velocity, direction and acceleration. The logical bootstrap point is based on a specific entity mobility model. The reference points follow around the logical guide points, each guiding one or more reference points, and maintain a constant distance and direction from the logical guide points. The velocity magnitude and direction update formula of the node is as follows:
v∈(vmin,vmax) (4)
θ∈(0,2π) (5)
assume the initial position of the base station is (x)0,y0) And the coordinates after the time t are as follows:
the throughput and the energy prediction are carried out by applying the metric learning, and the data acquisition throughput of the unmanned aerial vehicle and the charging energy consumption of the unmanned aerial vehicle can be optimized.
It is assumed herein that the drone cannot know global node information and therefore cannot obtain global path planning gain information in path planning. In route planning, instead of constructing a global map using specific information, a probabilistic map is used to construct a map whose local information is known and route planning is performed using the local information. The probability map is a directed-cycle graph and is represented by G ═ Gn (Ge), wherein Gn represents a node set in a space, and Ge represents a local path edge set formed between nodes.
The specific measurement matrix is a matrix structure, and the objective of measurement learning is to calculate the mahalanobis distance measurement between samples by finding a suitable measurement matrix M:
DM(di,dj)2＝(di-dj)TM(di-dj) (7)
where M is a semi-positive definite symmetric matrix, which may be denoted as M ═ LTL, this is equivalent to finding an L matrix as a mapping matrix to map the original data d to a new classification space L, so the euclidean distance may be regarded as a special case when the matrix L is a unit matrix.
And establishing a measurement matrix by using the distributed information as an input characteristic attribute, wherein the specifically considered attribute is as follows. 1. Threas is real-time throughput performance, the throughput prediction performance of the unmanned aerial vehicle at the next moment of data collection is related to the current throughput performance, 2 Pos is position information of a ground mobile robot node, the unmanned aerial vehicle data collection throughput performance is related to the distribution of the ground mobile robot, 3 Speed is the Speed of the ground mobile robot, the unmanned aerial vehicle data collection throughput performance is related to the Speed of the ground mobile robot, 4 Degree is the density of the neighbor nodes of the ground mobile robot node, the larger the density is, the more mobile robots exist around the robot, and the throughput can be improved when the unmanned aerial vehicle flies to the direction. 5. Energy is the residual Energy of the charging battery of the ground mobile robot, and the more the residual Energy is, the stronger the charging capability of the unmanned aerial vehicle is.
The label of metric learning is traction force, the traction force represents the willingness of the unmanned aerial vehicle base station to move towards the direction of a certain mobile robot, and the flight track of the unmanned aerial vehicle finally selects the direction with the maximum traction force to fly.
The LMNN algorithm is adopted to construct a metric matrix, and the core of the LMNN algorithm is to lead K input neighbors to belong to the same category in a new conversion space through learning a distance metric, while samples of different categories keep a certain distance. The LMNN algorithm generates the metric matrix as follows.
Firstly, an LMNN-based training network is operated in a test set to train a characteristic metric matrix off line. Assuming that the target sample xi has class label ci, and xl class label cl is in K adjacent points of the target sample xi, defining a noise point as cl ≠ ci for any target sample xi, and satisfying:
||L(xi-xl)||2≤||L(xi-xl)||2+1 (8)
where L is a distance metric matrix. According to the constraint, a non-equivalent constraint is first defined:
wherein D isL(xi,xj)＝||L(xi-xj)||2Representing the mapped point xiAnd xjA distance measure of (d); i, j ∈ KNN represents training sample xiFor testing sample xjK neighbors of (a), the K neighbors being a priori knowledge expressed in Kp; x is the number oflIs represented by and is at xiTraining samples within the maximum boundary but not identical to the test sample class labels; c. ClIs xlClass labels of (1); when x isiClass label c ofi＝clTime yil1 or else 0; equivalent constraints are defined as:
the final combination of non-equivalence and equivalence constraints can construct the following loss function:
ε(L)＝(1-u)εpull(L)+uεpush(L) (11)
wherein u is a weight coefficient generally 0.5.
The essence of the metric learning offline training is to generalize a set of scoring rules from a data set, each set of input can be given a definite score, and the method has a decisive effect on unmanned aerial vehicle path planning.
And updating the information of the whole network mobile robot by adopting a mode of periodically broadcasting Hello packets (H _ Pkt) hop by the unmanned aerial vehicle base station. The H _ Pkt includes the ID of the broadcast packet, the ID of the broadcast transmitting node, the IP address, the node speed, the node position, the broadcast transmission time, and the hop count information. The broadcast packet is transmitted to mobile robot nodes of the whole network in a hop-by-hop mode, the mobile robot receiving the broadcast packet regenerates the broadcast packet with the same packet ID by utilizing the information of the mobile robot and transmits the broadcast packet back to the unmanned aerial vehicle base station through the same path, and the unmanned aerial vehicle base station utilizes the information as the input of metric learning to carry out path planning.
After network initialization, periodically broadcasting Beacon packets by an unmanned aerial vehicle to collect basic information of ground sensor nodes, broadcasting information of a flight path to the sensor nodes in a coverage range by the unmanned aerial vehicle, collecting sensor information in the coverage range of the next flight path by the sensor nodes through multi-hop and sending the sensor information to the unmanned aerial vehicle, and adjusting the flight path and carrying out communication scheduling planning by the unmanned aerial vehicle by utilizing the information. The unmanned aerial vehicle path planning is carried out after the first periodic broadcasting information collection is finished, but due to the mobility of the ground sensor nodes, the planned flight path point is not the optimal flight path, so the unmanned aerial vehicle carries out the optimization adjustment of the flight path by using a metric learning method according to the position and speed information of the sensor nodes on the path point, the next adjustment path point is output periodically, and the next adjustment path point can not be effectively adjusted every time.
After the next flight path point is determined, the fact that the information of the unmanned aerial vehicle flying in the future period is known means that the unmanned aerial vehicle carries out scheduling distribution of channel access through the flight track of the unmanned aerial vehicle and the ground sensor information in the coverage area, the unmanned aerial vehicle actively establishes connection with the ground sensor nodes according to a channel access strategy to collect sensor data, and the process can be effectively achieved by a graph coloring theory.
Since the transmission of the data packet may overlap with the broadcast information collecting node, there is a multi-hop channel allocation problem, so that the network can be represented by a graph, each point in the graph represents a node in the network, and an edge connecting two points represents packet transmission communication of two points, so that the channel allocation problem of the multi-hop network can be converted into a graph coloring problem.
Given an undirected graph G ═ V, E, where V is the set of vertices and E is the set of edges. The graph coloring problem translates into dividing the set V into K independent subsets that do not contain any edges. The shading map problem is mainly used to calculate the minimum color set number K, which is represented by the number of channels. It is known that K channels utilize graph coloring theory in reverse to build a collision-free undirected graph for time slot assignment channel access. The basic theory of channel access is that when a node is in a transmit mode, it cannot receive a data packet, and in a receive mode, it can receive data packets of multiple nodes at the same time, so the color represents that the node occupies a channel to transmit data, and in the time slot represented by the color, the node cannot receive the data packet. In the shading algorithm used herein, the following parameters need to be defined: 1. and each node coloring number n, 2, a node neighbor coloring number table and 3, coloring expiration time.
The mechanism for establishing the collision-free coloring graph is that in the process of forwarding a data packet by a node i, coloring information of the node is added into the data packet, and a color-collision-free distribution mode is kept between the coloring information and neighbors in a one-hop range, namely, each node needs to ensure that the color of the node is different from the colors of all the neighbor nodes.
The unmanned aerial vehicle basic station carries out wireless charging when collecting ground mobile robot data, and ground mobile robot is equipped with wireless battery charging outfit, can long-range energy of providing give the unmanned aerial vehicle basic station, keeps the equipment energy consumption of unmanned aerial vehicle basic station sufficient.
And the unmanned aerial vehicle base station performs time division multiple access time slot allocation on the nodes in each group, and allocates time slots to the nodes in the coverage range of the unmanned aerial vehicle according to the priority sequence of the contact time to transmit data packets. And the real-time throughput result is returned to the unmanned aerial vehicle track optimization module, so that the flight track of the unmanned aerial vehicle is optimized. And reconstructing the transmission priority based on the change of the contact time caused by the position distribution and the mobility of the sensor node, wherein the transmission priority is used as the most important measurement index of the channel access priority. In a traditional ground-air communication network, data are transmitted to a ground sensor priority access channel according to the ground coverage and the flight direction of an unmanned aerial vehicle. In the unmanned aerial vehicle auxiliary network supporting fast movement, the flight speed of the unmanned aerial vehicle is high, and the ground node can be a mobile robot with mobility, so that a contact time prioritized channel access method is adopted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings 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 of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
Fig. 1 is a schematic flow chart of a traffic map coloring unmanned aerial vehicle energy-saving cruising data collection method based on metric learning according to an embodiment of the present invention;
FIG. 2 provides an overall block diagram for an embodiment of the present invention;
FIG. 3 is a schematic diagram of the broadcast Hello packet collecting map information according to the present invention;
FIG. 4 is a schematic diagram of a graph coloring method according to an embodiment of the present invention.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The technical scheme of the invention is specifically explained according to the attached drawings.
The method for collecting the energy-saving endurance data of the road map coloring unmanned aerial vehicle based on metric learning comprises the following steps:
s101, model establishment is carried out according to the unmanned aerial vehicle auxiliary sensor network, and a network model, an energy consumption model and a movement model are applied to the unmanned aerial vehicle auxiliary sensor network model.
The network model defines the data transmission process, the nodes in the cluster transmit the collected information to the cluster heads, the routing is established among the cluster heads, the collected information is transmitted to the base station, and the member nodes can also be used as relay routing to establish connection. The energy consumption model describes the energy consumption during data transmission if the distance between the transmitter and the receiver is less than a threshold value d0Then the free space model is used. If not, a multipath fading channel model is adopted. The movement model describes the unstable movement of the node and sets simulation boundaries.
And S102, constructing a global map by applying the probability road signs. It is assumed herein that the drone cannot know global node information and therefore cannot obtain global path planning gain information in path planning. In route planning, instead of constructing a global map using specific information, a probabilistic map is used to construct a map whose local information is known and route planning is performed using the local information. The probability map is a directed-cycle graph and is represented by G ═ Gn (Ge), wherein Gn represents a node set in a space, and Ge represents a local path edge set formed between nodes.
And S103, training a metric matrix off line. The specific measurement matrix is a matrix structure, and the purpose of measurement learning is to calculate the mahalanobis distance measurement between samples by finding a suitable measurement matrix M.
And S104, constructing a measurement matrix by using an LMNN algorithm. The LMNN algorithm is adopted to construct a metric matrix, and the core of the LMNN algorithm is to lead K input neighbors to belong to the same category in a new conversion space through learning a distance metric, while samples of different categories keep a certain distance.
S105, broadcasting a Hello packet to collect map information. And updating the information of the whole network mobile robot by adopting a mode of periodically broadcasting Hello packets (H _ Pkt) hop by the unmanned aerial vehicle base station. The H _ Pkt includes the ID of the broadcast packet, the ID of the broadcast transmitting node, the IP address, the node speed, the node position, the broadcast transmission time, and the hop count information. The broadcast packet is transmitted to mobile robot nodes of the whole network in a hop-by-hop mode, the mobile robot receiving the broadcast packet regenerates the broadcast packet with the same packet ID by utilizing the information of the mobile robot and transmits the broadcast packet back to the unmanned aerial vehicle base station through the same path, and the unmanned aerial vehicle base station utilizes the information as the input of metric learning to carry out path planning. See in particular fig. 3.
And S106, performing channel access based on the graph coloring method. The mechanism for establishing the collision-free coloring graph is that in the process of forwarding a data packet by a node i, coloring information of the node is added into the data packet, and a color-collision-free distribution mode is kept between the coloring information and neighbors in a one-hop range, namely, each node needs to ensure that the color of the node is different from the colors of all the neighbor nodes. See in particular fig. 4.
And S107, measuring the access priority based on the contact time. And the unmanned aerial vehicle base station performs time division multiple access time slot allocation on the nodes in each group, and allocates time slots to the nodes in the coverage range of the unmanned aerial vehicle according to the priority sequence of the contact time to transmit data packets. And the real-time throughput result is returned to the unmanned aerial vehicle track optimization module, so that the flight track of the unmanned aerial vehicle is optimized. And reconstructing the transmission priority based on the change of the contact time caused by the position distribution and the mobility of the sensor node, wherein the transmission priority is used as the most important measurement index of the channel access priority.
The invention assumes that the nodes are randomly distributed according to the group reference point group movement model, the unmanned aerial vehicle group performs group movement in the region, and the base station is unique and can freely move in the region. The unmanned aerial vehicles in the group can communicate with each other, and the communication between the unmanned aerial vehicles in the group is intermittently connected, so that the network has the intermittent connection characteristic of the opportunity network. The unmanned aerial vehicle is provided with a memory space which can cache data packets, the data packets are placed into the cache when the network connection is disconnected, and data are transmitted when the network is connected. After deployment of the sensor nodes in the field, the nodes may move within a fixed area to detect nearby environmental information. The invention considers how to ensure the effective transmission of data under the group moving scene. Considering the actual situation, the following assumptions are made:
(1) the nodes have equal initial energy and computing power and are equal in position;
(2) the nodes are randomly deployed in the region and accord with the group movement model initialization characteristics;
(3) all nodes in the network are mobile, including base stations and other nodes;
(4) the nodes know their own properties (e.g., remaining energy, speed and direction, etc.);
(5) the nodes adjust the transmission power according to the received signal strength and the communication link between the nodes is symmetrical.
(6) The drone remains flying at the same altitude.
(7) The drone may be controlled in flight.
Claims (8)
1. The method for collecting the energy-saving cruising data of the road map coloring unmanned aerial vehicle based on metric learning is characterized by comprising the following steps of:
the method comprises the steps of firstly, establishing a model according to an unmanned aerial vehicle auxiliary sensor network, and applying a network model, an energy consumption model and a movement model to the unmanned aerial vehicle auxiliary sensor network model.
And secondly, constructing a global map by using the local information by using the probability road signs, wherein the probability map does not determine the information construction map, but constructs a map with known local information by using a certain probability method, and performs path planning by using the local information. The probability map is a directed-cycle graph and is represented by G ═ Gn (Ge), wherein Gn represents a node set in a space, and Ge represents a local path edge set formed between nodes.
Thirdly, off-line training a measurement matrix, wherein the specific measurement matrix is a matrix structure, and the purpose of measurement learning is to calculate the mahalanobis distance measurement between samples by searching a suitable measurement matrix M
And fourthly, constructing a measurement matrix by using an LMNN algorithm. The core of the LMNN algorithm is to learn a distance metric such that in a new transformation space, K neighbors for an input all belong to the same class, while samples of different classes are kept at a certain large distance.
And fifthly, the broadcasting Hello packet collects map information, the broadcasting packet is transmitted to mobile robot nodes of the whole network in a hop-by-hop mode, the mobile robot receiving the broadcasting packet regenerates the broadcasting packet with the same packet ID by utilizing the information of the mobile robot, and transmits the broadcasting packet back to the unmanned aerial vehicle base station through the same path, and the unmanned aerial vehicle base station utilizes the information as input of metric learning to carry out path planning.
And sixthly, performing channel access based on a graph coloring method. The mechanism for establishing the collision-free coloring graph is that in the process of forwarding a data packet by a node i, coloring information of the node is added into the data packet, and a color-collision-free distribution mode is kept between the coloring information and neighbors in a one-hop range, namely, each node needs to ensure that the color of the node is different from the colors of all the neighbor nodes.
And seventhly, measuring access priority based on the contact time. And the unmanned aerial vehicle base station performs time division multiple access time slot allocation on the nodes in each group, and allocates time slots to the nodes in the coverage range of the unmanned aerial vehicle according to the priority sequence of the contact time to transmit data packets. And the real-time throughput result is returned to the unmanned aerial vehicle track optimization module, so that the flight track of the unmanned aerial vehicle is optimized.
2. The method of claim 1, wherein the unmanned aerial vehicle assists in building a sensor network model. Specifically, the network model defines the moving states of the node and the base station, specifies the data transmission mode of the node, and determines the channel access mechanism.
The energy consumption model defines a boundary condition threshold d0。
εfsIs the signal amplification factor, ε, of a free space channel modelmpIs the signal amplification factor of the multipath fading channel model.
If the distance between the transmitter and the receiver is less than a threshold value d0Then the free space model is used. If not, a multipath fading channel model is adopted.
The movement model describes the node's unstable motion, the node randomly selects the direction and speed of travel, and the new speed and direction are selected within a predetermined range. The selected group movement model is a reference point group movement model. In the reference point group movement model, the network is divided into a plurality of groups. For each group, there is one target in the group, and the nodes in the group move according to their targets and maintain certain constraints. The network is divided into several groups according to the requirement, the speed of the nodes in the group is controlled between 0 and the maximum speed, and the direction of the nodes is controlled between 0 and 2 pi. This allows the nodes within the group to maintain restricted random motion due to the presence of the target point within the group. The random direction and random velocity in the reference point group movement model are selected as follows:
v∈(vmin,vmax) (2)
θ∈(0,2π) (3)
3. the method of claim 1, wherein a probability map is applied to construct a global map with local information, assuming that the drone cannot know global node information and therefore cannot obtain global path planning gain information in path planning. In route planning, instead of constructing a global map using specific information, a probabilistic map is used to construct a map whose local information is known and route planning is performed using the local information. The probability map is a directed-cycle graph and is represented by G ═ Gn (Ge), wherein Gn represents a node set in a space, and Ge represents a local path edge set formed between nodes.
4. The method of claim 1, wherein the metric matrix is trained offline. And establishing a measurement matrix by using the distributed information as an input characteristic attribute, wherein the specifically considered attribute is as follows. 1. Threas is real-time throughput performance, the throughput prediction performance of the unmanned aerial vehicle at the next moment of data collection is related to the current throughput performance, 2 Pos is position information of a ground mobile robot node, the unmanned aerial vehicle data collection throughput performance is related to the distribution of the ground mobile robot, 3 Speed is the Speed of the ground mobile robot, the unmanned aerial vehicle data collection throughput performance is related to the Speed of the ground mobile robot, 4 Degree is the density of the neighbor nodes of the ground mobile robot node, the larger the density is, the more mobile robots exist around the robot, and the throughput can be improved when the unmanned aerial vehicle flies to the direction. 5. Energy is the residual Energy of the charging battery of the ground mobile robot, and the more the residual Energy is, the stronger the charging capability of the unmanned aerial vehicle is.
The label of metric learning is traction force, the traction force represents the willingness of the unmanned aerial vehicle base station to move towards the direction of a certain mobile robot, and the flight track of the unmanned aerial vehicle finally selects the direction with the maximum traction force to fly.
5. The method of claim 1, wherein the LMNN algorithm is used to construct the metric matrix, and the LMNN-based training network is first run in the test set to train the feature metric matrix offline. Assuming that the target sample xi has class label ci, and xl class label cl is in K adjacent points of the target sample xi, defining a noise point as cl ≠ ci for any target sample xi, and satisfying:
||L(xi-xl)||2≤||L(xi-xl)||2+1 (4)
where L is a distance metric matrix. According to the constraint, a non-equivalent constraint is first defined:
wherein, the distance measurement of the mapped points xi and xj is represented; representing that the training sample xi is a K neighbor of the test sample xj, wherein the K neighbor is priori knowledge and is represented by Kp; xl represents the training sample that is within the xi maximum boundary but not identical to the test sample class label; cl is a class label of xl; when ci-cl is a class label of xi, yil-1 is 0 otherwise; equivalent constraints are defined as:
the final combination of non-equivalence and equivalence constraints can construct the following loss function:
ε(L)＝(1-u)εpull(L)+uεpush(L) (7)
wherein u is a weight coefficient generally 0.5.
The essence of the metric learning offline training is to generalize a set of scoring rules from a data set, each set of input can be given a definite score, and the method has a decisive effect on unmanned aerial vehicle path planning.
6. The method for collecting the energy-saving cruising data of the unmanned aerial vehicle colored by the roadmap based on the metric learning of claim 1, wherein the map information is collected by broadcasting a Hello packet, and the information of the whole network mobile robot is updated by broadcasting the Hello packet (H _ Pkt) by the base station of the unmanned aerial vehicle hop by hop. The H _ Pkt includes the ID of the broadcast packet, the ID of the broadcast transmitting node, the IP address, the node speed, the node position, the broadcast transmission time, and the hop count information. The broadcast packet is transmitted to mobile robot nodes of the whole network in a hop-by-hop mode, the mobile robot receiving the broadcast packet regenerates the broadcast packet with the same packet ID by utilizing the information of the mobile robot and transmits the broadcast packet back to the unmanned aerial vehicle base station through the same path, and the unmanned aerial vehicle base station utilizes the information as the input of metric learning to carry out path planning.
7. The method of claim 1, wherein the graph coloring method is used for channel access, and an undirected graph G (V, E) is given, where V is a vertex set and E is an edge set. The graph coloring problem translates into dividing the set V into K independent subsets that do not contain any edges. The shading map problem is mainly used to calculate the minimum color set number K, which is represented by the number of channels. It is known that K channels utilize graph coloring theory in reverse to build a collision-free undirected graph for time slot assignment channel access. The basic theory of channel access is that when a node is in a transmit mode, it cannot receive a data packet, and in a receive mode, it can receive data packets of multiple nodes at the same time, so the color represents that the node occupies a channel to transmit data, and in the time slot represented by the color, the node cannot receive the data packet. In the shading algorithm used herein, the following parameters need to be defined: 1. and each node coloring number n, 2, a node neighbor coloring number table and 3, coloring expiration time.
8. The method of claim 1, wherein priority is accessed based on time-of-contact metric. Because the flight speed of the unmanned aerial vehicle is high, the contact time between the unmanned aerial vehicle and the ground mobile robot is short, the unmanned aerial vehicle is likely to fly out of the contact area with the sensor when the distributed access time is reached, priority measurement of the contact time is needed, the ground mobile robot with short contact time is placed in the previous time to access a channel for data transmission, and high throughput performance is guaranteed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110449681.XA CN113163332B (en) | 2021-04-25 | 2021-04-25 | Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110449681.XA CN113163332B (en) | 2021-04-25 | 2021-04-25 | Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113163332A true CN113163332A (en) | 2021-07-23 |
| CN113163332B CN113163332B (en) | 2022-07-05 |
Family
ID=76870626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110449681.XA Active CN113163332B (en) | 2021-04-25 | 2021-04-25 | Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113163332B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111709565A (en) * | 2020-06-08 | 2020-09-25 | 山东大学深圳研究院 | Energy efficiency optimization method and system based on multi-layer shuttle system |
| CN114302339A (en) * | 2021-12-25 | 2022-04-08 | 宁波凯德科技服务有限公司 | Augmented Lagrange method capable of covering robot signal for positioning unmanned aerial vehicle |
| CN115178944A (en) * | 2022-08-04 | 2022-10-14 | 广东工业大学 | Narrow space robot operation planning method for safety reinforcement learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180007615A1 (en) * | 2016-06-30 | 2018-01-04 | Fortinet, Inc. | Automatic channel selection in wireless local area network (wlan) controller based deployments |
| CN107924384A (en) * | 2015-03-11 | 2018-04-17 | 阿雅斯迪公司 | For the system and method using study model prediction result is predicted |
| CN110225582A (en) * | 2019-07-01 | 2019-09-10 | 西安电子科技大学 | Unmanned plane energy supply dispatching method based on cooperation transmission |
-
2021
- 2021-04-25 CN CN202110449681.XA patent/CN113163332B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107924384A (en) * | 2015-03-11 | 2018-04-17 | 阿雅斯迪公司 | For the system and method using study model prediction result is predicted |
| US20180007615A1 (en) * | 2016-06-30 | 2018-01-04 | Fortinet, Inc. | Automatic channel selection in wireless local area network (wlan) controller based deployments |
| CN110225582A (en) * | 2019-07-01 | 2019-09-10 | 西安电子科技大学 | Unmanned plane energy supply dispatching method based on cooperation transmission |
Non-Patent Citations (5)
| Title |
|---|
| DANIEL RICCIO: "A_New_Unsupervised_Approach_for_Segmenting_and_Counting_Cells_in_High-Throughput_Microscopy_Image_Sets", 《IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS》 * |
| ERIC P XING: "大数据的分布式机器学习的策略与原则", 《ENGINEERING》 * |
| YANG WANG: "Unsupervised_Metric_Fusion_Over_Multiview_Data_by_Graph_Random_Walk-Based_Cross-View_Diffusion", 《IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS》 * |
| 孙彦赞: "基于图着色的密集D2D网络资源分配算法", 《计算机工程》 * |
| 方宏昊: "大规模、快速移动机器人群的机器学习组网关键技术研究", 《硕士学位论文集》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111709565A (en) * | 2020-06-08 | 2020-09-25 | 山东大学深圳研究院 | Energy efficiency optimization method and system based on multi-layer shuttle system |
| CN114302339A (en) * | 2021-12-25 | 2022-04-08 | 宁波凯德科技服务有限公司 | Augmented Lagrange method capable of covering robot signal for positioning unmanned aerial vehicle |
| CN115178944A (en) * | 2022-08-04 | 2022-10-14 | 广东工业大学 | Narrow space robot operation planning method for safety reinforcement learning |
| CN115178944B (en) * | 2022-08-04 | 2024-05-24 | 广东工业大学 | Narrow space robot operation planning method for safety reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113163332B (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113163332B (en) | Road sign graph coloring unmanned aerial vehicle energy-saving endurance data collection method based on metric learning | |
| You et al. | Hybrid offline-online design for UAV-enabled data harvesting in probabilistic LoS channels | |
| CN111786713B (en) | Unmanned aerial vehicle network hovering position optimization method based on multi-agent deep reinforcement learning | |
| Zhang et al. | Energy-efficient trajectory optimization for UAV-assisted IoT networks | |
| Rahmani et al. | OLSR+: A new routing method based on fuzzy logic in flying ad-hoc networks (FANETs) | |
| Alam et al. | Topology control algorithms in multi-unmanned aerial vehicle networks: An extensive survey | |
| CN113163466B (en) | Self-adaptive fish school routing packet routing method based on fuzzy decision tree | |
| Ramli et al. | Hybrid mac protocol for uav-assisted data gathering in a wireless sensor network | |
| Perera et al. | A WPT-enabled UAV-assisted condition monitoring scheme for wireless sensor networks | |
| Park et al. | Analysis of dynamic cluster head selection for mission-oriented flying ad hoc network | |
| CN111479239A (en) | Sensor emission energy consumption optimization method of multi-antenna unmanned aerial vehicle data acquisition system | |
| Wang et al. | A-GR: A novel geographical routing protocol for AANETs | |
| CN115499921A (en) | Three-dimensional trajectory design and resource scheduling optimization method for complex unmanned aerial vehicle network | |
| Lu et al. | Relay in the sky: A UAV-aided cooperative data dissemination scheduling strategy in VANETs | |
| CN116113025A (en) | Track design and power distribution method in unmanned aerial vehicle cooperative communication network | |
| Gao et al. | Coverage-maximization and energy-efficient drone small cell deployment in aerial-ground collaborative vehicular networks | |
| Gu et al. | An aerial-computing-assisted architecture for large-scale sensor networks | |
| Liu et al. | A Q-learning based adaptive congestion control for V2V communication in VANET | |
| Syfullah et al. | Mobility-based clustering algorithm for multimedia broadcasting over IEEE 802.11 p-LTE-enabled VANET | |
| Li et al. | TaskPOI priority-based energy balanced multi-UAVs cooperative trajectory planning algorithm in 6G networks | |
| CN113776531A (en) | Multi-unmanned-aerial-vehicle autonomous navigation and task allocation algorithm of wireless self-powered communication network | |
| Liu et al. | An efficient message dissemination scheme for cooperative drivings via multi-agent hierarchical attention reinforcement learning | |
| Akin et al. | Multiagent Q-learning based UAV trajectory planning for effective situationalawareness | |
| Say et al. | Cooperative path selection framework for effective data gathering in UAV-aided wireless sensor networks | |
| CN114598721A (en) | High-energy-efficiency data collection method and system based on joint optimization of track and resources |
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 |