CN116249202A - Combined positioning and computing support method for Internet of things equipment - Google Patents
Combined positioning and computing support method for Internet of things equipment Download PDFInfo
- Publication number
- CN116249202A CN116249202A CN202310237384.8A CN202310237384A CN116249202A CN 116249202 A CN116249202 A CN 116249202A CN 202310237384 A CN202310237384 A CN 202310237384A CN 116249202 A CN116249202 A CN 116249202A
- Authority
- CN
- China
- Prior art keywords
- iot device
- uav
- positioning
- computing
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 77
- 238000005265 energy consumption Methods 0.000 claims abstract description 43
- 238000004364 calculation method Methods 0.000 claims abstract description 28
- 238000013468 resource allocation Methods 0.000 claims abstract description 17
- 230000002787 reinforcement Effects 0.000 claims abstract description 15
- 238000005457 optimization Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims description 34
- 239000003795 chemical substances by application Substances 0.000 claims description 31
- 230000005540 biological transmission Effects 0.000 claims description 28
- 230000009471 action Effects 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 15
- 238000004891 communication Methods 0.000 claims description 13
- 238000009826 distribution Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 6
- 230000007613 environmental effect Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 2
- 230000001934 delay Effects 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 description 5
- 230000033001 locomotion Effects 0.000 description 5
- 230000007423 decrease Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 description 1
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
- H04W28/0967—Quality of Service [QoS] parameters
- H04W28/0975—Quality of Service [QoS] parameters for reducing delays
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
- H04W28/0967—Quality of Service [QoS] parameters
- H04W28/0983—Quality of Service [QoS] parameters for optimizing bandwidth or throughput
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W52/00—Power management, e.g. TPC [Transmission Power Control], power saving or power classes
- H04W52/04—TPC
- H04W52/06—TPC algorithms
- H04W52/14—Separate analysis of uplink or downlink
- H04W52/143—Downlink power control
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Probability & Statistics with Applications (AREA)
- Operations Research (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Algebra (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The combined positioning and computing support method of the internet of things equipment comprises the steps of collecting distance information and residual electric quantity information of the internet of things equipment and anchor nodes, and obtaining an available anchor node set of the internet of things equipment; when the number of available anchor nodes is less than 3, collecting distance information of the IoT devices and other IoT devices; the UAV builds a positioning model according to the collected information, converts the positioning problem into a semi-positive planning problem through weighted least square and semi-positive relaxation, and solves the coordinates of the IoT device; the UAV takes the minimum system total energy consumption and delay as optimization targets, and adopts a deep reinforcement learning method to obtain a combined decision of calculation task unloading, UAV track planning and UAV calculation resource allocation; obtaining an optimal downlink power allocation decision by a successive approximation method, wherein the objective is to maximize the minimum downlink throughput; the UAV performs optimal joint decisions to support offloading requests of IoT devices, performs optimal downlink power allocation decisions to promote downlink throughput of the system.
Description
Technical Field
The invention belongs to the technical field of the Internet of things, and relates to a combined positioning and computing support method of Internet of things equipment.
Background
The rapid evolution of internet of things (Internet of Things, ioT) technology has driven an exponential increase in the number of IoT devices, resulting in a large number of computationally intensive and delay sensitive applications. In some scenarios, ioT devices are widely deployed in challenging or sparse ground base station areas, such as forests, mountainous areas, deserts, and underwater locations, requiring constant movement and performance of some computationally intensive tasks, including disaster pre-warning, long pipeline infrastructure detection, underwater infrastructure detection, and military operations, which mostly require tracking of monitored targets, involving positioning functions. However, due to the high cost, ioT devices cannot continuously and stably obtain real-time locations by hosting satellite positioning modules. The mobility, privacy, limited power, and complexity of the environment in which IoT devices are located (e.g., out of line of sight) make IoT device positioning extremely challenging. In order to reduce the cost of IoT devices, protect the privacy of IoT devices, and at the same Time provide stable real-Time computing support for IoT devices, a multi-device collaborative Arrival Time (TOA) positioning scheme based on semi-positive relaxation (SemiDefinite Relaxation, SDR) needs to be constructed.
On the other hand, ioT devices have very limited battery power, computing resources, and storage resources to handle massive computing-intensive and delay-sensitive tasks. Due to hardware conditions and environmental limitations, ioT devices have limited battery life and it is difficult to obtain real-time power. Furthermore, the computing and storage resources owned by IoT devices may not be sufficient to handle certain large tasks, requiring devices with greater processing power to provide computing support. In response to the above problems, researchers have proposed an emerging computing paradigm, namely mobile edge computing (Mobile Edge Computing, MEC), to extend the computing power of IoT devices. In the challenging environment under consideration, the ground infrastructure is sparsely distributed, fails to provide stable and reliable computing support for IoT devices, and requires research into computing offloading of unmanned aerial vehicles (Unmanned Aerial Vehicle, UAV) based on flexible deployment.
The Chinese patent CN114745389A discloses a calculation unloading method of a mobile edge calculation system, which designs a calculation unloading scheme facing the mobile edge calculation system. In order to reduce the average information age in the Internet of things system, firstly, analyzing a state update task at the equipment end of the Internet of things, acquiring the computing resource requirement and the space requirement of a computing task on edge equipment, computing the time delay at the local and edge server ends, and further acquiring the information age of each equipment and the information age of the whole system; selecting an information age optimal computing and unloading strategy according to the task demand and the environmental condition of each device; and (3) providing a calculation unloading scheme based on optimal information age of game theory, and iterating opportunities for updating the unloading strategies of all devices in each round until the unloading strategies of all devices reach convergence, so as to obtain a final calculation unloading method, and effectively reducing the average information age of the system to meet the information freshness requirements of different types of Internet of things devices. However, when the technical scheme faces a large amount of computationally intensive and delay sensitive tasks in areas with challenges or sparse ground base stations, such as forests, mountainous areas and underwater positions, the scheme tends to locally process the tasks, and cannot provide enough computational support for relevant IoT devices through sparsely deployed edge servers, which increases the energy consumption of IoT devices with limited electric quantity; furthermore, this approach also fails to provide stable real-time computing support in view of the mobility of IoT devices.
The Chinese patent CN114124955A designs a computing and unloading method based on the two-stage multi-agent game. Aiming at a 5G hybrid dual-network mode, a network channel model is established, the channel transmission rate of a user is calculated according to the shannon theorem, and the time delay and the energy consumption for task unloading of the user are calculated according to the channel transmission rate; establishing an edge unloading model according to a Stackelberg game, taking the private network user as a leader, taking the public network user as a follower, taking the unloading amount as a strategy, and setting a utility function of the private network user and a utility function of the public network user; under the condition of complete information game, verifying the existence of the balance of the Stackelberg, and solving a first optimal unloading strategy of the private network user and the public network user; under the condition of incomplete information game, a TSDRL algorithm is adopted to obtain a second optimal unloading strategy, and the first optimal unloading strategy is utilized to evaluate the convergence of the second optimal unloading strategy. According to the technical scheme, in a scene with sparse distribution of the ground base stations, the task unloading capacity is considered to be optimized, and the whole system delay and the energy consumption are ignored, so that the scheme method cannot provide high experience quality for users in some challenging scenes.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide the combined positioning and computing support method for the Internet of things equipment, which reduces the positioning energy consumption of the Internet of things equipment, protects the privacy of the Internet of things equipment and provides stable and reliable computing support for the Internet of things equipment in an environment with sparse distribution of ground base stations and challenges.
The invention provides a joint positioning and computing support method of Internet of things equipment, which comprises the following steps:
step 1: collecting distance information of the anchor nodes and residual electric quantity information of the anchor nodes in a communication range of the IoT device, and obtaining an available anchor node set of each IoT device;
step 2: for the case that the number of available anchor nodes is less than 3, constructing an auxiliary IoT device set according to the communication range of the target IoT device, and calculating distance information between the target IoT device and all auxiliary IoT devices;
step 3: the method comprises the steps that the IoT device sends collected information to the UAV, the UAV builds a positioning model according to the position and distance information, the positioning problem is converted into a semi-positive planning problem through weighted least square and semi-positive relaxation, and finally, a convex optimization tool is adopted to solve the position coordinates of the target IoT device;
step 4: the UAV edge server dynamically plans a flight path and allocates computing resources to process the task request of unloading, aims at minimizing the total energy consumption and delay of the system, and adopts a deep reinforcement learning method to obtain a joint decision of computing task unloading, UAV path planning and UAV computing resource allocation;
step 5: the UAV edge server obtains an optimal downlink power allocation decision by a successive approximation method, and aims at maximizing minimum downlink throughput;
step 6: the UAV performs optimal joint decisions to support offloading requests of IoT devices while performing optimal downlink power allocation decisions to promote downlink throughput of the system.
The combined positioning and computing support method for the Internet of things equipment has at least the following beneficial effects:
1. according to the invention, the mobile edge computing technology is utilized to deploy the edge server on the UAV, so that the mobile edge server is constructed, and the motion trail can be actively planned to better support the offloading task of the IoT device, the flexibility of the computing support system is increased, and the computing resource utilization rate of the UAV is improved. The problem that IoT devices cannot carry high-cost global positioning systems due to cost limitations and stable and reliable communication and computing support cannot be obtained due to sparse ground base station distribution and challenging environments is solved.
2. The combined positioning method in the combined positioning and calculation supporting technical scheme of the Internet of things equipment can obtain more accurate real-time positioning information, and the calculation supporting method accelerates the network training speed, improves the task completion rate and improves the resource utilization rate of the system. According to the invention, performance indexes such as algorithm convergence performance, total time delay and energy consumption of the system, task incomplete proportion, average task unloading data size, unmanned aerial vehicle computing resource utilization rate and the like are evaluated in an experiment, so that good effects are achieved, and the practicability of the system can be greatly improved.
Drawings
FIG. 1 is a system architecture diagram;
FIG. 2 is a joint positioning flow chart of an Internet of things device;
FIG. 3 is a flow chart of joint location and computing support for an Internet of things device;
FIG. 4 is a graph of mean square positioning error as a function of average number of anchor nodes;
FIG. 5 is a graph of energy consumption for positioning as a function of average number of anchor nodes;
FIG. 6 is a graph comparing convergence performance of different algorithms;
FIG. 7 is a graph of total latency and energy consumption of different algorithms as a function of network training process;
FIG. 8 is a graph of the ratio of incomplete tasks of different algorithms as a function of the network training process;
FIG. 9a is a graph showing the average amount of task data offloaded as a function of task data size for a training round number of 50;
FIG. 9b is a graph showing the average amount of task data offloaded as a function of task data size for a training round number of 6000;
FIG. 10 is a graph showing the change of the task unloaded ratio with the network training process after the state coordination;
fig. 11 is a graph of computational resource utilization of a UAV as a function of the number of IoT devices.
Detailed Description
The invention provides a joint positioning and computing support method of Internet of things equipment.
Firstly, an available anchor node set is obtained according to the distance information of the IoT devices and anchor nodes in a communication range and the residual capacity information of the anchor nodes, and an auxiliary IoT device set is obtained according to the distance information of the IoT devices and other IoT devices in the communication range aiming at the situation that the number of the available anchor nodes is rare. In addition, the IoT device solves its own location coordinates geometrically from the location and distance information in the available anchor nodes and auxiliary IoT devices.
Then, performing flight trajectory planning, task offloading, and computing resource allocation using mobile edge computing technology and UAVs deployed with onboard edge servers, providing computing support for IoT devices and minimizing a weighted sum of system total energy consumption and latency. UAV intelligent agent trains in a distributed mode, solves the problem of state conflict through a coordination normalization method, and accelerates the training process.
And finally, executing joint track planning, task unloading and computing resource allocation actions obtained by UAV agent training, and simultaneously obtaining and executing an optimal downlink power allocation decision by utilizing a successive approximation technology.
The combined positioning and computing support method of the Internet of things equipment can provide stronger computing resource expansion for the Internet of things equipment, increase the flexibility of a computing support system and improve the resource utilization efficiency of the system.
As shown in fig. 1, the network model of the present invention is divided into two layers, a device layer and a server layer. The device layer includes a large number of IoT devices with computationally intensive and delay sensitive tasks including disaster pre-warning, long pipeline infrastructure detection, underwater infrastructure detection, and military operations, among others. IoT devices move in a gaussian markov random motion pattern and randomly generate task requests with a certain probability. The IoT devices have very limited computing resources and can offload complex tasks as much as possible to edge servers for processing. Further, a number of anchor nodes are fixedly deployed in the vicinity of IoT devices for positioning. The server layer includes a plurality of UAVs deployed with edge servers, each UAV receiving offload requests from IoT devices while planning flight trajectories to better provide computing support for the IoT devices.
The scheme of the invention mainly comprises two parts of combined positioning and calculation support of the Internet of things equipment. As shown in fig. 2, the joint positioning of the internet of things device mainly includes the steps of collecting anchor node energy consumption information, obtaining distance information between the IoT device and the anchor node, forming an available anchor node set and an auxiliary IoT device set, obtaining positioning coordinates by adopting a geometric method, and the like. As shown in fig. 3, the computational support includes the steps of determining joint task offloading, UAV trajectory planning, and computational resource allocation decisions, and determining optimal downlink power allocation decisions. The method specifically comprises the following steps:
step 1: collecting distance information of the anchor nodes and residual electric quantity information of the anchor nodes in a communication range of the IoT device, and obtaining an available anchor node set of each IoT device, wherein the step 1 specifically comprises:
step 1.1: for IoT device u i Constructing a set of candidate anchor nodes for positioningWherein include u i All anchor nodes within communication range;
step 1.2: ioT device u i Transmitting a positioning request signal q to candidate anchor nodes i Including unique identification of IoT devices and location request information;
step 1.3: after the positioning request signal reaches the candidate anchor node, the anchor node n k Calculating a corresponding IoT device u according to the transmission timestamp and the arrival timestamp of the request i And anchor node n k Euclidean distance d between ik The self identification information, the time stamp information of the arrival request, the distance information between the IoT device and the self position information and the residual power information are returned to the IoT device;
step 1.4: for collectionsCandidate anchor node in IoT device u i Screening anchor nodes with the residual electric quantity larger than a threshold value e according to the collected anchor node residual electric quantity information to form an available anchor node set +.>And calculating the number c of anchor nodes in the available anchor node set i 。
Step 2: for the case that the number of available anchor nodes is less than 3, constructing an auxiliary IoT device set according to the communication range of the target IoT device, and calculating distance information between the target IoT device and all auxiliary IoT devices, wherein step 2 specifically comprises:
step 2.1: for IoT device u i Building auxiliary IoT device setsWherein include u i All IoT devices within communication range;
step 2.2: ioT device u i Transmitting a location request signal q to an auxiliary IoT device i Including IoT device u i Is a unique identification and location request information of the mobile terminal;
step 2.3: after the location request signal reaches the auxiliary IoT device, auxiliary IoT device u j According to the sending time stamp and the arrival time stamp of the request, calculating the self and the corresponding IoT device u i Euclidean distance d between ij And combine self-identification information, timestamp information of request arrival, and IoT device u i Information such as distance information of (a) is returned to the target IoT device u i 。
Step 3: the IoT device sends the collected information to the UAV, the UAV builds a positioning model according to the position and distance information, the positioning problem is converted into a semi-positive programming (SDP) problem by weighted least squares and semi-positive relaxation (SDR), and finally, the position coordinates of the target IoT device are solved by adopting a convex optimization tool, and step 3 specifically comprises:
step 3.1: for each IoT device, a positioning problem is defined as estimating the location of the target IoT device from the measured distance containing measured noise and NLOS scene noise given the available anchor node and auxiliary IoT device locations, constructing a positioning model of the IoT device as follows:
wherein d ik Representing IoT device u i To anchor node n k Or auxiliary IoT device u k Is used for the observation distance of the (a),representing IoT device u i To anchor node n k Or auxiliary IoT device u k True distance, n ik Is obeyed to mean 0 and variance sigma 2 Measurement of gaussian distribution of (c)Noise, m ik Is NLOS scene noise, and m ik >>n ik ;/>For the set of available anchor nodes, < > a->For a secondary IoT device set; by means of the collection->Storing distance information, set ∈>Distance information between IoT devices and anchor nodes is deposited.
Step 3.2: the square of two sides of the positioning model is omitted which is far smaller than the rest itemsSimultaneous commandThe positioning model can be converted into: />
Step 3.3: let IoT device coordinates beAnchor node coordinates are +.>Wherein->For IoT device set, ->For the anchor node set, according to the positioning model obtained in the step 3.2, unknown parameters S i And p ik Using a weighted least squares method to estimate, a weighted least squares problem is converted to:
wherein, the weight parameter w ik =1/(d ik ·σ ik ) 2 ,σ ik Representing the variance corresponding to the measured noise whenFrom the collection->When (I)>When->From the collection->When (I)>
Step 3.4: introducing auxiliary variablesI.e. < ->And->The positioning problem can be translated as follows:
step 3.5: converting the non-convex problem in the step 3.4 into an SDP problem by utilizing SDR relaxation constraint and combining with the Shu' S complement theorem, and obtaining an IoT device positioning coordinate S with a relatively stable value through continuous iteration by a convex optimization tool such as CVX solution i 。
To this end, the positioning process of the IoT device ends. Next, at each slot t, the IoT device sends its own position to the UAV while offloading tasks, enabling the UAV to better provide computing support for it.
Step 4: the UAV edge server dynamically plans a flight path and allocates computing resources to process the task request of unloading, aims at minimizing the total energy consumption and delay of the system, adopts a deep reinforcement learning method to obtain a joint decision of computing task unloading, UAV path planning and UAV computing resource allocation, and the step 4 specifically comprises:
step 4.1: defining a weighted sum of total delay and energy consumption of the UAV calculation support system, and then combining and optimizing an unloading decision, a UAV path planning decision and a UAV calculation resource allocation decision by taking the minimum system total energy consumption and delay as targets;
the invention provides a joint optimization scheme based on deep reinforcement learning, wherein the optimization targets are weighted sums of total energy consumption and delay of a system, and the weighted sums comprise transmission and calculation energy consumption and transmission and calculation delay generated by an internet traffic (IoT) device, and flight and calculation energy consumption and calculation delay generated by an UAV edge server:
wherein t is used for indicating time slots, M is used for indicating UAVs, n is the total time slot length, M is the number of unmanned aerial vehicles, alpha is a calculated unloading proportion variable, U is UAV position coordinates, f is a calculated resource allocation variable,in order to keep the time delay t and the energy consumption E in the same magnitude; t (T) t m Representation of UAVv m The sum of the calculated delays of the transmission and calculation corresponding to the task request processed by the calculation delay, i.e. the total task delay,/->Representing the sum of the transmission and calculation energy consumption corresponding to the UAV flight energy consumption and the calculation energy consumption and the task request processed by the UAV flight energy consumption.
UAV assigned computing resource f to IoT device t i,m Can exceed the maximum computing resource of the self
Wherein the UAV has the following limitations with respect to flight speed:
for UAV flight speed, v min And v max The lower limit value and the upper limit value of the UAV flying speed are defined.
Step 4.2: each UAV edge server deploys a deep reinforcement learning module as an Agent, defines a Markov decision process corresponding to the joint optimization problem, and comprises an Agent, an environmental State, action actions and rewards Reward, so that a foundation is provided for the neural network learning and training of the UAV edge server;
agent: each UAV edge server is considered an Agent for which the environment is considered to be fully observable, with observations being equivalent to states. Each Agent contains an Actor and Critic network, acting as an action policy and policy reviewer, respectively. Wherein the parameter of the Critic value network is mu, and the parameter corresponding to the target network is mu - The method comprises the steps of carrying out a first treatment on the surface of the The parameter of the Actor strategy network is theta, and the corresponding parameter of the target network is theta - 。
State: including IoT device information and UAV edge server information, UAV edge server v m The observed state at time slot t is defined asWherein f t m Is the idle computing resource of the UAV at time slot t,is the position coordinates of the UAV in time slot t, R t,u Is an IoT deviceUplink transmission rate vector of C t Is the task information set, req, of an IoT device t Is an offload request set of IoT devices and the elements in the set satisfy + ->Where 0 indicates that no offload request is sent, 1 indicates that a request is sent, and-1 indicates that a request after state coordination is sent.
Action: including offloading the proportional decisions, computing the resource allocation decisions, and the next slot position coordinates of the UAV, the UAV edge server v m The action performed in time slot t is defined asWherein->By the speed angle change during UAV flight +.>And speed size->To express, will a t Redefined as
Reward: when an Agent performs an action, the system will transition from one environmental state to another and get rewards that will guide each Agent to its optimal policy. The definition of the reward function is generally related to the optimization objective of the system, and therefore, the reward function is defined as the inverse of the weighted sum of energy consumption and delay:
wherein the opposite number is taken to convert the cost into a prize, the log (-) function is used to smooth the prize,is a penalty when multiple agents generate different offloading actions for the same IoT device, such that a state conflict occurs.
Step 4.3: the Agent of each UAV edge server selects an action a through an Actor policy network t ;
Step 4.4: agent executes action a obtained in step 4.3 t And observe rewards r t And next state s t+1 ;
Step 4.5: agents will experience tuples R (s t ,a t ,r t ,s t+1 ) Storing in an experience buffer zone, sampling small batches of experiences in the experience buffer zone to update a neural network so as to accelerate the training process, updating a main Actor network by minimizing a strategy objective function, and updating a main Critic network by minimizing a loss function;
step 4.6: the Agent updates the target Actor network and the target Critic network through a soft update strategy in each time slot;
step 4.7: aiming at the problems of action conflict caused by the state conflict of related agents and the amplitude difference between input state elements when an IoT device is in a plurality of UAV coverage areas, each Agent communicates after generating actions, and corresponding punishments are added in rewards of the conflicting actions so as to realize coordination normalization of states;
step 4.8: the Agent continuously repeats the steps 4.3-4.6 for trial and error and learning, and finally obtains the optimal joint decision { alpha, U, f } of task unloading, UAV track planning and computing resource allocation.
Step 5: the UAV edge server obtains an optimal downlink power allocation decision by a successive approximation method, and aims at maximizing minimum downlink throughput, wherein the step 5 specifically comprises the following steps:
step 5.1: defining a minimum average downlink transmission rate of a UAV calculation support system to optimize downlink power allocation so as to improve downlink throughput;
wherein P is d A variable is allocated for the downlink transmission power,representing the proportion of IoT device tasks offloaded to UAV edge server processing, +.>For IoT device u k With UAV edge server v m The downlink transmission rate between the two is defined as follows:
wherein W represents the bandwidth,represents noise power +.>Indicating channel gain, +.>Representation of UAVv m Is->Representing co-channel interference caused by other UAVs at time slot t.
By introducing auxiliary variables R dmin Redefining the problem as maximizing the minimum average downlink transmission rate problem:
the downlink transmission power of the UAV edge server cannot exceed the specified maximum downlink transmission power:
step 5.2: the problem of maximizing the minimum average downlink transmission rate is converted into a convex problem by utilizing the property of a logarithmic function in the downlink transmission power and a first-order Taylor expansion, an optimal power value is obtained in each iteration by utilizing a successive convex approximation algorithm SCA until the difference between the optimal values of two successive iterations is smaller than a certain threshold value, so that the optimization of the maximum downlink throughput is realized, and the downlink transmission power distribution is optimized.
Step 6: the UAV performs an optimal joint decision to support the offloading request of the IoT device, and performs an optimal downlink power allocation decision to improve the downlink throughput of the system, step 6 specifically is:
step 6.1: and (3) the UAV adjusts the flight trajectory of the UAV according to the flight trajectory planning decision obtained in the step (4), and processes the task unloading request of the IoT device according to the task unloading decision and the computing resource allocation decision in the step (4).
Step 6.2: and (3) the UAV transmits the task result back to the IoT device according to the optimal downlink power allocation decision obtained in the step (5), and the downlink throughput of the system is improved by maximizing the minimum downlink transmission power.
The technical scheme of the invention is described in detail below in combination with a specific experimental platform and experimental results.
According to the simulation implementation method, based on Matlab and PyCharm platforms, an IoT device working area of 1000m multiplied by 1000m is considered, wherein 3 unmanned aerial vehicle edge servers are deployed, initial positions of 100 IoT devices are subjected to uniform distribution, 30 anchor nodes are fixed in the considered area in a uniform random distribution mode, mobility of the IoT devices is simulated by adopting a Gaussian Markov random motion model of the nodes, and the unmanned aerial vehicle flies at a fixed height and provides calculation support for the IoT devices. The IoT devices have a computational frequency size of [0.1,0.5]Evenly distributed in GHz, tasksData size I of (2) k In [100,1000]The computing resources gamma required by the unit task data volume are uniformly distributed in KB k In [500,1000]The cycles/bit are uniformly distributed, the maximum computing resource capacity of the unmanned aerial vehicle is set to be 20GHz, and the flying speed is uniformly distributed in [10,15 ]]m/s. The experiment simulates 6000 slots, during each time period, the IoT device generates a computing task with a probability of 0.99, the UAV hovers to a specified location according to the action generated during the previous time period, receives the offload task, continues flying while processing the computing task, and therefore, the positions of the IoT device and the UAV are dynamically changed during different time periods. The parameter list is shown in table 1:
TABLE 1 parameter settings
Where K is the number of IoT devices, M is the number of UAVs, N is the number of anchor nodes, φ 1 And phi 3 Are constant parameters related to IoT device hardware, η is the discount rate of the reward, θ is the Actor network parameter, μ is the Critic network parameter,is unmanned plane v m Maximum computational resource of->Representation of UAVv m Maximum downlink transmission power of +.>Representing the noise power.
Meanwhile, the invention sets a comparison algorithm when the simulation is realized. Aiming at a positioning algorithm, in order to evaluate the mean square positioning error (Root Mean Squared Error, RMSE) of a positioning position and an actual position and the positioning energy consumption of an IoT device, a comparison experiment is carried out on a Matlab platform on the least square method, the single-device positioning method based on SDR, the multi-device positioning method based on SDR and the combined positioning method provided by the patent.
Fig. 4 shows the measurement noise sigma 2 When=1, the mean square positioning error of the positioning location and the actual location varies with the number of average anchor nodes around the IoT device. The RMSE obtained by the combined positioning method is lower than other comparison algorithms, because the combined positioning method has the advantages of a single-device positioning method based on SDR and a multi-device positioning method based on SDR, and the algorithm can be adaptively adjusted according to the distribution condition of the anchor node.
Fig. 5 shows the variation of average location energy consumption of an IoT device with the number of average anchor nodes around the IoT device. The positioning energy consumption of the method is similar to that of the least square method and far lower than that of other comparison algorithms, because the combined positioning method uses the least square principle, and when the number of anchor nodes is increased, the combined positioning method gradually deviates to the positioning method using single SDR-based equipment, so that energy conservation is realized.
Aiming at a joint optimization algorithm of task unloading, UAV computing resource allocation and track planning, in order to evaluate algorithm convergence, weighted sum of total energy consumption and time delay, task incomplete proportion, average unloading task data size, proportion of tasks which are not unloaded after state coordination to coordinated tasks and UAV computing resource utilization rate, a comparison experiment is carried out on a PyCharm platform by a depth deterministic strategy gradient algorithm (PF-DDPG) adopting a penalty mechanism, a depth Q network algorithm (PF-DQN) adopting a penalty mechanism, a Greedy algorithm (Greedy) and a multi-agent depth reinforcement learning algorithm (MASC-DDPG) based on state coordination normalization.
FIG. 6 shows the convergence of the algorithms, characterized by the change in total rewards of UAV agents. As can be seen from the figure, the rewards of the greeny method do not change greatly with the increase of the time slot, the rewards of other deep reinforcement learning methods become larger gradually and finally tend to be stable, and the rewards after training and stabilization are larger than those obtained by the greeny method. In addition, compared to the other two deep reinforcement learning algorithms, our algorithm converges faster and gets a higher total prize because the MASC-DDPG method considers the state coordination normalization of multiple UAV agents and the speed of the DQN method decreases significantly as the number of generated motion quantization increases.
Figure 7 shows the weighted sum of total energy consumption and time delay as a function of training time. It can be seen that the weighted sum of the total energy consumption and the time delay of the greeny method does not change greatly with the increase of the time slot, and the weighted sum of the total energy consumption and the time delay of other deep reinforcement learning methods gradually decreases to finally tend to be stable, and is smaller than the weighted sum of the total energy consumption and the time delay of the greeny method after training is stable. Furthermore, the weighted sum of our algorithm total energy consumption and time delay is smaller than the other two deep reinforcement learning algorithms.
Fig. 8 shows a variation of the unfinished scale of the task. As can be seen from the figure, as the time slot increases, the task incomplete ratio gradually decreases, and reaches a minimum value after convergence. Compared with other algorithms, the MASC-DDPG algorithm can achieve fewer unfinished proportion, has smaller amplitude fluctuation and stable performance.
Figures 9a and 9b show average offloaded task data size from IoT devices to UAV at different training wheel numbers, different task data sizes. As can be seen from fig. 9a, at a training round number of 50, the average task data amount of the greeny algorithm is highest, and the unloading proportion of other algorithms based on deep reinforcement learning is smaller. As can be seen from fig. 9b, when the training round number reaches 6000, the task load off of the deep reinforcement learning-based algorithm is higher than the greeny algorithm. This is because at an initial stage, the training effect has not yet emerged, and the deep reinforcement learning-based algorithm will choose to offload tasks as much as possible based on rewards when training reaches steady state.
FIG. 10 shows the proportion of unloaded tasks in the actions generated by the MASC-DDPG algorithm to coordinated tasks after the coordinated state. It can be obtained that the initial unloaded ratio is about 0.47, gradually decreases with the progress of training, and finally converges to about 0.05. Analysis shows that after the penalty factors for the actions which are not unloaded are added in the reward functions, the network continuously tends to unload tasks during training, so that the conditions of the actions which are not unloaded after the state coordination are obviously reduced.
Fig. 11 shows UAV computing resource utilization as a function of IoT device number. As can be seen from the figure, as the number of IoT devices increases, the computing resource utilization gradually increases and tends to stabilize. Because the computational resources of the UAV are limited, as the number of IoT devices increases, the total number of tasks increases, and the computational resources allocated to the tasks by the UAV eventually approach their maximum computational resource capacity, at which point the resource utilization cannot continue to increase. It can also be seen from the figure that our computing resources are most utilized.
Considering the hardware limitations and mobility of IoT devices, we first propose a joint positioning algorithm based on semi-positive relaxation and arrival time. Then, for complex internet of things scenes of multiple IoT devices and multiple UAVs, a MASC-DDPG algorithm is provided, state coordination normalization of multiple UAV agents is considered to reduce state conflicts and action conflicts, and convergence speed of network training is improved. Experimental results show that the scheme provided by the patent can accurately locate the IoT device in real time, and compared with a comparison algorithm, the method achieves the lowest location error and the lowest location energy consumption. The scheme shortens the convergence time of network training, effectively reduces the weighted sum of the total energy consumption and time delay of the system, improves the average task unloading data quantity, improves the utilization rate of the computing resources of the UAV, and provides dynamic and reliable computing support for the IoT equipment.
The foregoing description of the preferred embodiments of the invention is not intended to limit the scope of the invention, but rather to enable any modification, equivalent replacement, improvement or the like to be made without departing from the spirit and principles of the invention.
Claims (8)
1. The combined positioning and computing support method for the Internet of things equipment is characterized by comprising the following steps of:
step 1: collecting distance information of the anchor nodes and residual electric quantity information of the anchor nodes in a communication range of the IoT device, and obtaining an available anchor node set of each IoT device;
step 2: for the case that the number of available anchor nodes is less than 3, constructing an auxiliary IoT device set according to the communication range of the target IoT device, and calculating distance information between the target IoT device and all auxiliary IoT devices;
step 3: the method comprises the steps that the IoT device sends collected information to the UAV, the UAV builds a positioning model according to the position and distance information, the positioning problem is converted into a semi-positive planning problem through weighted least square and semi-positive relaxation, and finally, a convex optimization tool is adopted to solve the position coordinates of the target IoT device;
step 4: the UAV edge server dynamically plans a flight path and allocates computing resources to process the task request of unloading, aims at minimizing the total energy consumption and delay of the system, and adopts a deep reinforcement learning method to obtain a joint decision of computing task unloading, UAV path planning and UAV computing resource allocation;
step 5: the UAV edge server obtains an optimal downlink power allocation decision by a successive approximation method, and aims at maximizing minimum downlink throughput;
step 6: the UAV performs optimal joint decisions to support offloading requests of IoT devices while performing optimal downlink power allocation decisions to promote downlink throughput of the system.
2. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein step 1 specifically comprises:
step 1.1: constructing a set of candidate anchor nodes for positioning for the IoT device, including all anchor nodes within communication range;
step 1.2: the IoT device sends a location request signal to the candidate anchor node, wherein the location request signal comprises a unique identification of the IoT device and location request information;
step 1.3: after the positioning request signal reaches the candidate anchor node, the anchor node calculates Euclidean distance between the corresponding IoT device and the anchor node according to the sending timestamp and the arrival timestamp of the request, and returns self identification information, timestamp information of the arrival request, distance information with the IoT device, self position information and residual electric quantity information to the IoT device;
step 1.4: for candidate anchor nodes in the set, the IoT device screens anchor nodes with residual power greater than a threshold according to the collected anchor node residual power information to form an available anchor node set, and calculates the number of anchor nodes in the available anchor node set.
3. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein step 2 is specifically as follows:
step 2.1: building a secondary IoT device set for IoT devices, including all IoT devices within communication range;
step 2.2: the IoT device sends a location request signal to the auxiliary IoT device, wherein the location request signal comprises a unique identification of the IoT device and location request information;
step 2.3: after the positioning request signal arrives at the auxiliary IoT device, the auxiliary IoT device calculates the euclidean distance between itself and the corresponding IoT device according to the sending timestamp and the arrival timestamp of the request, and sends the self identification information, the timestamp information of the arrival request and the IoT device u i Information such as distance information of (a) is returned to the target IoT device.
4. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein the step 3 is specifically as follows:
step 3.1: for each IoT device, a positioning problem is defined as estimating the location of the target IoT device from the measured distance containing measured noise and NLOS scene noise given the available anchor node and auxiliary IoT device locations, constructing a positioning model of the IoT device as follows:
wherein d ik Representing IoT device u i To anchor node n k Or auxiliary IoT device u k Is used for the observation distance of the (a),representing IoT device u i To anchor node n k Or auxiliary IoT device u k True distance, n ik Is obeyingMean 0, variance sigma 2 Measurement noise of gaussian distribution, m ik Is NLOS scene noise, and m ik >>n ik ;/>For the set of available anchor nodes, < > a->For a secondary IoT device set; by collectionStoring distance information, set ∈>Storing distance information between the IoT device and the anchor node;
step 3.2: the square of two sides of the positioning model is omitted which is far smaller than the rest itemsSimultaneous make->The positioning model can be converted into:
step 3.3: let IoT device coordinates beAnchor node coordinates are +.>Wherein the method comprises the steps ofFor IoT device set, ->For the anchor node set, according to the positioning model obtained in the step 3.2, unknown parameters S i And p ik Using a weighted least squares method to estimate, a weighted least squares problem is converted to:
wherein, the weight parameter w ik =1/(d ik ·σ ik ) 2 ,σ ik Representing the variance corresponding to the measured noise whenFrom the collection->In the time-course of which the first and second contact surfaces,when->From the collection->When (I)>Step 3.4: introducing the auxiliary variable +.>I.e.And->The positioning problem can be translated as follows:
step 3.5: converting the non-convex problem in the step 3.4 into an SDP problem by utilizing SDR relaxation constraint and combining with the Shu' S complement theorem, solving the SDP problem through a convex optimization tool, and obtaining an IoT (Internet traffic control) device positioning coordinate S with a relatively stable value through continuous iteration i 。
5. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein the step 4 is specifically:
step 4.1: defining a weighted sum of total delay and energy consumption of the UAV calculation support system, and then combining and optimizing an unloading decision, a UAV path planning decision and a UAV calculation resource allocation decision by taking the minimum system total energy consumption and delay as targets;
step 4.2: each UAV edge server deploys a deep reinforcement learning module as an Agent, defines a Markov decision process corresponding to the joint optimization problem, and comprises an Agent, an environmental State, action actions and rewards Reward, so that a foundation is provided for the neural network learning and training of the UAV edge server;
step 4.3: the Agent of each UAV edge server selects an action a through an Actor policy network t ;
Step 4.4: agent executes action a obtained in step 4.3 t And observe rewards r t And next state s t+1 ;
Step 4.5: agents will experience tuples R (s t ,a t ,r t ,s t+1 ) Storing in an experience buffer zone, sampling small batches of experiences in the experience buffer zone to update a neural network so as to accelerate the training process, updating a main Actor network by minimizing a strategy objective function, and updating a main Critic network by minimizing a loss function;
step 4.6: the Agent updates the target Actor network and the target Critic network through a soft update strategy in each time slot;
step 4.7: aiming at the problems of action conflict caused by the state conflict of related agents and the amplitude difference between input state elements when an IoT device is in a plurality of UAV coverage areas, each Agent communicates after generating actions, and corresponding punishments are added in rewards of the conflicting actions so as to realize coordination normalization of states;
step 4.8: and (3) continuously repeating the steps 4.3-4.6 by the Agent to perform trial and error and study, and finally obtaining the optimal joint decision of task unloading, UAV track planning and computing resource allocation.
6. The method for supporting joint positioning and computing of internet of things equipment according to claim 5, wherein the step 4.1 aims at minimizing total energy consumption and delay of the system, and the specific expression is as follows:
wherein t is used for indicating time slots, M is used for indicating UAVs, n is the total time slot length, M is the number of unmanned aerial vehicles, alpha is a calculated unloading proportion variable, U is UAV position coordinates, f is a calculated resource allocation variable,in order to keep the time delay t and the energy consumption E in the same magnitude; />Representation of UAVv m Is added to the sum of the calculated delays and the transmission corresponding to the task request processed by the same>Representing the sum of the transmission and calculation energy consumption corresponding to the UAV flight energy consumption and the calculation energy consumption and the task request processed by the UAV flight energy consumption.
7. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein the step 5 is specifically:
step 5.1: defining a minimum average downlink transmission rate of a UAV calculation support system to optimize downlink power allocation so as to improve downlink throughput;
wherein P is d A variable is allocated for the downlink transmission power,representing the proportion of IoT device tasks offloaded to UAV edge server processing, +.>For IoT device u k With UAV edge server v m Downlink transmission rate between the two;
step 5.2: the problem of maximizing the minimum average downlink transmission rate is converted into a convex problem by utilizing the property of a logarithmic function in the downlink transmission power and a first-order Taylor expansion, an optimal power value is obtained in each iteration by utilizing a successive convex approximation algorithm SCA until the difference between the optimal values of two successive iterations is smaller than a certain threshold value, so that the optimization of the maximum downlink throughput is realized, and the downlink transmission power distribution is optimized.
8. The method for supporting joint positioning and computing of internet of things equipment according to claim 1, wherein the step 6 is specifically:
step 6.1: and (3) the UAV adjusts the flight trajectory of the UAV according to the flight trajectory planning decision obtained in the step (4), and processes the task unloading request of the IoT device according to the task unloading decision and the computing resource allocation decision in the step (4).
Step 6.2: and (3) the UAV transmits the task result back to the IoT device according to the optimal downlink power allocation decision obtained in the step (5), and the downlink throughput of the system is improved by maximizing the minimum downlink transmission power.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310237384.8A CN116249202A (en) | 2023-03-13 | 2023-03-13 | Combined positioning and computing support method for Internet of things equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310237384.8A CN116249202A (en) | 2023-03-13 | 2023-03-13 | Combined positioning and computing support method for Internet of things equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116249202A true CN116249202A (en) | 2023-06-09 |
Family
ID=86633002
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310237384.8A Pending CN116249202A (en) | 2023-03-13 | 2023-03-13 | Combined positioning and computing support method for Internet of things equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116249202A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117915481A (en) * | 2024-01-18 | 2024-04-19 | 重庆理工大学 | Resource allocation method and system of ultra-dense industrial Internet of things |
-
2023
- 2023-03-13 CN CN202310237384.8A patent/CN116249202A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117915481A (en) * | 2024-01-18 | 2024-04-19 | 重庆理工大学 | Resource allocation method and system of ultra-dense industrial Internet of things |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112351503B (en) | Task prediction-based multi-unmanned aerial vehicle auxiliary edge computing resource allocation method | |
CN111800828B (en) | Mobile edge computing resource allocation method for ultra-dense network | |
CN113543176B (en) | Unloading decision method of mobile edge computing system based on intelligent reflecting surface assistance | |
Li et al. | Downlink transmit power control in ultra-dense UAV network based on mean field game and deep reinforcement learning | |
Chen et al. | Deep reinforcement learning based resource allocation in multi-UAV-aided MEC networks | |
CN113359480B (en) | Multi-unmanned aerial vehicle and user cooperative communication optimization method based on MAPPO algorithm | |
CN113254188B (en) | Scheduling optimization method and device, electronic equipment and storage medium | |
WO2022242468A1 (en) | Task offloading method and apparatus, scheduling optimization method and apparatus, electronic device, and storage medium | |
Chen et al. | Learning-based computation offloading for IoRT through Ka/Q-band satellite–terrestrial integrated networks | |
CN114567888A (en) | Multi-unmanned aerial vehicle dynamic deployment method | |
CN116546559B (en) | Distributed multi-target space-ground combined track planning and unloading scheduling method and system | |
CN113507492A (en) | Internet of things three-dimensional space intelligent positioning and distributed edge calculation support method | |
CN113919483A (en) | Method and system for constructing and positioning radio map in wireless communication network | |
CN116249202A (en) | Combined positioning and computing support method for Internet of things equipment | |
Nasr-Azadani et al. | Single-and multiagent actor–critic for initial UAV’s deployment and 3-D trajectory design | |
CN117528649A (en) | Method for establishing end-edge cloud system architecture, task unloading and resource allocation optimization method and end-edge cloud system architecture | |
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 | |
CN117499867A (en) | Method for realizing high-energy-efficiency calculation and unloading through strategy gradient algorithm in multi-unmanned plane auxiliary movement edge calculation | |
Xue et al. | Deep learning based channel prediction for massive MIMO systems in high-speed railway scenarios | |
CN116882270A (en) | Multi-unmanned aerial vehicle wireless charging and edge computing combined optimization method and system based on deep reinforcement learning | |
CN115766478A (en) | Unloading method of air-ground cooperative edge computing server | |
CN114916013A (en) | Method, system and medium for optimizing unloading time delay of edge task based on vehicle track prediction | |
CN114520991A (en) | Edge network self-adaptive deployment method based on unmanned aerial vehicle cluster | |
CN114513814A (en) | Edge network computing resource dynamic optimization method based on unmanned aerial vehicle auxiliary node | |
Gao et al. | MO-AVC: Deep Reinforcement Learning Based Trajectory Control and Task Offloading in Multi-UAV Enabled MEC Systems |
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 |