CN117715117B - Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation - Google Patents
Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation Download PDFInfo
- Publication number
- CN117715117B CN117715117B CN202410166403.7A CN202410166403A CN117715117B CN 117715117 B CN117715117 B CN 117715117B CN 202410166403 A CN202410166403 A CN 202410166403A CN 117715117 B CN117715117 B CN 117715117B
- Authority
- CN
- China
- Prior art keywords
- uav
- user
- ris
- unloading
- time slot
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000004364 calculation method Methods 0.000 title claims abstract description 49
- 238000001829 resonance ionisation spectroscopy Methods 0.000 claims abstract description 102
- 238000013468 resource allocation Methods 0.000 claims abstract description 36
- 238000005457 optimization Methods 0.000 claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 18
- 230000010363 phase shift Effects 0.000 claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims abstract description 7
- 230000002787 reinforcement Effects 0.000 claims abstract description 7
- 238000004458 analytical method Methods 0.000 claims abstract description 4
- 238000004891 communication Methods 0.000 claims description 22
- 238000005265 energy consumption Methods 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 17
- 230000009471 action Effects 0.000 claims description 13
- 230000005540 biological transmission Effects 0.000 claims description 10
- 239000000654 additive Substances 0.000 claims description 4
- 230000000996 additive effect Effects 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 3
- 101000983970 Conus catus Alpha-conotoxin CIB Proteins 0.000 claims description 2
- 101000932768 Conus catus Alpha-conotoxin CIC Proteins 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 238000006073 displacement reaction Methods 0.000 claims description 2
- 238000005562 fading Methods 0.000 claims description 2
- 230000006870 function Effects 0.000 claims description 2
- 230000002040 relaxant effect Effects 0.000 claims 1
- 230000008901 benefit Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008054 signal transmission Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013021 overheating Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a task unloading method for multi-UAV-RIS auxiliary mobile edge calculation, which comprises the following steps: constructing a large-scale mobile edge computing unloading scene assisted by multiple UAVs-RISs; constructing an optimization model; according to the current optimal RIS phase offset and the UAV track, solving a user unloading decision, an unloading proportion and user and UAV computing resource allocation; solving an RIS phase shift matrix by adopting an analytic method; solving the UAV locus by applying a deep reinforcement learning algorithm; and circularly executing the solving step by using an alternative optimization algorithm to obtain an optimal allocation strategy of the computing resources of the user equipment and the UAV, obtaining an optimal unloading decision and unloading proportion of the computing tasks of the user equipment, and obtaining RIS phase offset and the track of the optimal UAV. The invention can obtain the maximum average throughput of the whole network.
Description
Technical Field
The invention belongs to the technical field of mobile edge calculation assisted by Unmanned aerial vehicles (Unmanned AERIAL VEHICLE, UAV) and intelligent reflecting surfaces (reconfigurated IntelligentSurface, RIS), and particularly relates to a task unloading method for multi-UAV-RIS assisted mobile edge calculation.
Background
In the current digital age, with the rapid development of mobile communication and internet of things technologies, the number and complexity of computationally intensive tasks are also rapidly increasing. These tasks tend to consume more energy. These tasks typically require significant computing power and consume significant amounts of energy. However, the computational resources and energy of the user equipment are limited, which becomes an important bottleneck to improve the quality of service of the user. Mobile edge computing is an innovative technique that allows user devices to offload their computing tasks to an edge server for computation. This technique takes advantage of the high performance computing resources and sufficient energy of the edge servers to minimize the computing burden on the user device. In this way, mobile edge computing effectively improves the performance of the user device in handling computationally intensive tasks, while avoiding the problem of overheating or insufficient energy of the device due to heavy tasks.
However, in a large-scale mobile edge computing offload scenario, communications between a base station and a ground user face challenges such as obstruction, signal interference, and communication capacity limitations. These factors lead to degradation in user quality of service, affecting the reliability and efficiency of communications. To solve these problems, smart reflector technology has attracted extensive attention and research in recent years in industry and academia. The intelligent reflecting surface is a surface capable of intelligently adjusting the reflecting characteristics of the intelligent reflecting surface, and the control and reflection of the incident signal are realized by controlling the shape phase and the phase amplitude of the surface elements of the intelligent reflecting surface. The technology can effectively solve the communication problem between the base station and the ground user, especially when the signal is blocked or interfered by the obstacle. By reflection from the intelligent reflecting surface, the signal can bypass the obstacle and reach the destination, thereby significantly improving the communication quality. In addition, the unmanned aerial vehicle is a powerful tool for solving the communication problem in the urban environment due to the characteristics of high maneuverability, low cost and easiness in deployment. The unmanned aerial vehicle can flexibly adjust the flying position and the flying height, so that better signal coverage and larger communication capacity are provided for users under different geographic positions and communication requirements. Especially in urban environments, tall buildings, dense streets and other obstructions often cause the communication signals to be blocked and faded. However, the unmanned aerial vehicle's mobility enables it to easily avoid these obstacles, point-to-point signal transmission, and thereby establish a more reliable and efficient communication link between urban residents.
Disclosure of Invention
The technical problems to be solved are as follows: aiming at the defects in the prior art, the invention provides a multi-UAV-RIS-assisted mobile edge computing task unloading method, which fully considers UAV computing resources, UAV tracks for obtaining a global optimal solution, user equipment unloading decision and unloading proportion and computing resource allocation when facing a large-scale complex scene, thereby obtaining the maximum average throughput of the whole network.
The technical scheme is as follows:
a method of task offloading multi-UAV-RIS assisted mobile edge computing, the method comprising the steps of:
s1, constructing a large-scale mobile edge computing unloading scene assisted by a plurality of UAVs-RISs according to data collected by a third party;
S2, respectively establishing a user unloading decision, an unloading proportion, an UAV track, RIS phase offset, a communication model, a calculation model, a time delay model and an energy consumption model for user and UAV calculation resource allocation, and establishing an optimization model based on the communication model, the calculation model, the time delay model and the energy consumption model; the communication model is used for calculating the average throughput of the network, the calculation model is used for calculating the representation of tasks, the task quantity and the CPU quantity required by calculation, the time delay model is used for calculating time delay by the UAV server, calculating time delay and unloading time delay by the user, and the energy consumption model is used for calculating the flight energy consumption of the UAV, calculating the energy consumption of two-way transmission and calculating the local energy consumption of the UAV and the user;
s3, setting an initial value of a related optimization variable;
s4, solving a user unloading decision, an unloading proportion and user and UAV calculation resource allocation according to the current optimal RIS phase offset and the UAV locus;
S5, solving the RIS phase offset matrix by adopting an analytic method according to the current optimal user unloading decision, unloading proportion, user and UAV computing resource allocation and UAV locus;
s6, according to the current optimal user unloading decision, unloading proportion, user and UAV calculation resource allocation and RIS phase offset, a deep reinforcement learning algorithm is applied to solve the UAV track;
s7, circularly executing the step S4, the step S5 and the step S6 by using an alternative optimization algorithm until the absolute value of energy or time delay consumed by the whole network between two adjacent iterations is smaller than a preset threshold value or the maximum preset iteration times are reached, ending the iteration, obtaining an optimal allocation strategy of computing resources of user equipment and the UAV, obtaining an optimal unloading decision and unloading proportion of computing tasks of the user equipment, and obtaining RIS phase offset and the track of the optimal UAV;
and S8, based on the current state of the whole mobile edge computing network, applying UAV tracks, an optimal task unloading decision and unloading proportion, RIS phase offset and an optimal resource allocation strategy to load and calculate the computing tasks of all user equipment in a large-scale target area by taking the average throughput of the whole network as a target.
The beneficial effects are that:
the multi-UAV-RIS assisted mobile edge computing task unloading method is suitable for large-scale mobile edge computing complex scenes, can solve user equipment unloading decision and unloading proportion by adopting a mathematical method according to specific environmental conditions, achieves effective unloading, and obtains computing resource allocation strategies of the user equipment and the UAV, so that computing tasks of users are further and efficiently completed, resource maximum utilization is achieved, and service experience of the user equipment is improved. Due to the variability of the environment and the randomness of the user position, the invention introduces a deep reinforcement learning algorithm at the base station end to solve the optimal flight trajectory of the UAV in real time, thereby being capable of adapting to dynamic environment changes, obtaining an optimal UAV trajectory and obviously improving the real-time performance of the system.
Drawings
FIG. 1 is a schematic diagram of a multi-UAV-RIS assisted mobile edge computing task offloading method according to an embodiment of the present invention.
FIG. 2 is a flow chart of a method for multi-UAV-RIS assisted mobile edge computing task offloading in accordance with an embodiment of the present invention.
Fig. 3 is a graph of the number of ues per algorithm versus the average throughput for different algorithms according to an embodiment of the present invention.
Fig. 4 is a graph of the task size versus average throughput of a user device for different algorithms according to an embodiment of the present invention.
FIG. 5 is a graph of RIS element number versus average throughput for various algorithms in accordance with an embodiment of the present invention.
Fig. 6 is a diagram showing the relationship between the number of users and the average throughput of the network for three different trace cases.
Detailed Description
The following examples will provide those skilled in the art with a more complete understanding of the invention, but are not intended to limit the invention in any way.
The invention will now be described in further detail with reference to the accompanying drawings.
It should be noted that the terms like "upper", "lower", "left", "right", "front", "rear", and the like are also used for descriptive purposes only and are not intended to limit the scope of the invention in which the invention may be practiced, but rather the relative relationship of the terms may be altered or modified without materially altering the teachings of the invention.
FIG. 1 is a schematic diagram of a multi-UAV-RIS assisted mobile edge computing task offloading method of an embodiment of the invention. In a physical entity network, which comprises two layers, namely a mobile user layer and an UAV layer, wherein a user in the mobile user layer carries a base station around the mobile equipment, the mobile equipment carried by the mobile user can generate a calculation task at any time, and the position of the user has randomness; the UAV layer is composed of multiple UAVs equipped with MEC servers, which cover the active area of the mobile user and can process the calculation tasks of the user equipment, thus reducing the burden of the user equipment. The quality of service of the user is low because of the possible obstruction between the user and the base station,
Fig. 2 is an algorithm flow of an embodiment of the present invention. The process comprises the following steps:
Step 1, constructing a large-scale mobile edge computing unloading scene assisted by a plurality of UAVs-RISs according to data collected by a third party; the method specifically comprises the following steps:
since the user equipment needs to process the computational tasks in a limited time, in successive cycles Internal/>The computing task is performed in time slots, each time slot numbered/>. Obtaining environmental information using data collected by a third party,/>Individual single antenna users are randomly distributed over a length/>Width is/>In the region of (2), the three-dimensional coordinates of the user are expressed as; UAV quantity is/>The number of the mth UAV is/>; Each UAV can be used as an air base station to provide unloading service for ground users, and the coordinates of each UAV are as follows; RIS quantity is/>Each UAV is provided with a RIS, no./>The number of RIS is/>; The RIS adopts a uniform linear array, so that the signal reflection of the user can be unloaded to a ground base station; 3D coordinates of the base station are fixed as/>。
The user equipment can not only offload the calculation task to the UAV edge server, but also offload the calculation task to the ground base station through RIS reflection to define a new setRepresenting a location where the user device may be offloaded; defining unloading decision variables/>, for the characteristics that users may be unloaded to an UAV edge server or ground base station：
；
When (when)Time representation/>Time slot user/>Deciding to offload part of its computational tasks to the mth UAV; /(I)Representation/>Time slot user/>Deciding to offload part of its calculation task reflection to the base station; when/>Time representation/>Time slot user/>The unloading service is not performed, and all calculation tasks are selected to perform local calculation; only one user equipment per slot can be offloaded and only to one location, there is；
The mth UAV is in time slotSet to/>; Defining the starting and ending positions of the UAV, there are/>，/>Representing UAV initial flight position,/>Representing an UAV end stop position; the maximum travel distance of the UAV in a slot is limited to: /(I)，/>Representing the maximum displacement that the UAV can achieve per unit time; in order to prevent UAVs from colliding during flight, a minimum safe distance must be maintained between UAVs,，/>Is the minimum safe distance, j is the number of the jth UAV.
Step 2, constructing a communication model of user unloading decision, unloading proportion, UAV track, RIS phase offset and user and UAV computing resource allocation according to a task unloading method of multi-UAV-RIS auxiliary mobile edge computing, wherein the communication model comprises the following processes:
calculating time slots according to Euclidean formula The distance between different devices. Specifically, the/>Personal UAV (unmanned aerial vehicle) and (th)The distance between individual user equipments is/>; First/>Individual RIS and/>The distance between individual user equipments is/>; First/>The distance between the individual RIS and the base station is/>：
；
；
。
In the time slot, taking the situation in reality into considerationFrom the/>Personal user equipment to the/>The UAVs are line-of-sight links with channel gains/>The method comprises the following steps:
Wherein, Is the channel power gain at a reference distance of 1 m.
Similarly, in time slotsIn, from the/>Personal user equipment to the/>Channel gain of individual RIS/>And from the/>Channel gain of individual RIS to base station/>The method comprises the following steps:
where d is the spacing between the reflective elements, Is the carrier wavelength; /(I)A cosine value representing the angle of arrival from the kth user to the 2 nd RIS,/>The representation is from the/>The cosine value of the arrival angle of the nth RIS from the individual user; /(I)A cosine value representing the departure angle from RIS 2 to the base station,/>A cosine value representing the departure angle from the mth RIS to the base station. In time slot/>，/>Is/>The diagonal phase shift matrix of each RIS is expressed as:
Wherein, 、/>、…、/>The 1 st reflective element, the 2 nd reflective element, … th reflective element, and the Z-th reflective element of the nth RIS are respectively phase shifted in time slot t, and Z is the number of reflective elements in a single RIS.
If the userIn time slot/>Served by the mth UAV, i.e./>Then corresponding signal to noise ratio/>, of the mth UAVExpressed as:
Wherein, Representing user/>Transmission power of/>Is the power of the additive white gaussian noise at the receiving end of the UAV.Expressed in time slot/>Interference caused by transmissions from all other users within the co-channel.
In this case, the userIn time slot/>Unloading rate/>The method comprises the following steps:
If the user In time slot/>Obtaining base station service, i.e./>The corresponding signal to noise ratio/>, at the base stationThe method comprises the following steps:
；
In this case, the user In time slot/>Unloading rate/>The method comprises the following steps:
Wherein, Representing user/>Transmission power of/>Is the power of the additive white gaussian noise at the receiving end of the UAV.Expressed in time slot/>Interference caused by transmissions from all other users in the co-channel and by reflection from the RIS.
Step 3, constructing a calculation model of user unloading decision, unloading proportion, UAV track, RIS phase offset and user and UAV calculation resource allocation according to a task unloading method of multi-UAV-RIS auxiliary mobile edge calculation, wherein the calculation model comprises the following processes:
with one user facing one computationally intensive task per slot To be performed, the task is defined as:
Wherein, Representing time slot/>User/>Data volume to be processed,/>Representing the total number of CPU cycles required to perform the computational task.
The computing power and battery capacity of the user are limited, and the user needs to be in a time slotAnd (3) completing the computing task, and unloading part of the computing task to an edge server. /(I)Time slot user/>The ratio of local calculations is set to/>The divided local task quantity is as followsThe amount of tasks offloaded to the edge server is/>. By adjusting/>The user may choose the ratio between local calculation and offloading to the edge server to meet the computing resource and battery capacity constraints.
Step 4, constructing a time delay model process of user unloading decision, unloading proportion, UAV track, RIS phase offset and user and UAV computing resource allocation according to a task unloading method of multi-UAV-RIS auxiliary mobile edge computing, wherein the time delay model process comprises the following steps:
Due to the user Local calculation ratio is/>Unloading ratio is/>. Thus, at/>Time slot, userTime needed for local computation/>The method comprises the following steps:
Wherein, Is/>Time slot user/>Representing the local computing power of the slot/>User/>The number of CPUs calculated.
Each user has limited computing power and the user is in a time slotComputing power in/>The method comprises the following steps:
Wherein, Maximum local computing power, here a limitation of user computing resources, is per user.
If the userDeciding on time slot/>Offloading part of its computational tasks to the mth UAV, the time required for offloading/>And calculating the required time/>The method comprises the following steps of:
Wherein, For the mth UAV in slot/>Can be allocated to user/>Is a computing resource of (a).
The computational power of each UAV is also limited, and the UAVs are in time slotsComputing power in/>The method comprises the following steps:
Wherein, Is the maximum computing resource that each UAV can allocate, here the limit of the computing resources of the UAV.
If the userDeciding on time slot/>Offloading part of its tasks to the base station, the time required for offloading/>And calculating the required time/>The method comprises the following steps:
Wherein, Computing resources that can be allocated to users for the base station.
Step 5, constructing an energy consumption model of user unloading decision, unloading proportion, UAV track, RIS phase offset and user and UAV computing resource allocation according to a task unloading method of multi-UAV-RIS auxiliary mobile edge computing, wherein the energy consumption model comprises the following steps:
If time slot Inner user/>Deciding to perform the computational task locally, then user/>Locally calculating the consumed energy/>The method comprises the following steps:
Wherein, Is the user/>A coefficient greater than zero, called user/>An effective switching capacitance of the CPU;
If time slot Inner user/>Deciding to offload part of its computational tasks to the mth UAV, then offloading the consumed energyAnd calculating the consumed energy/>The method comprises the following steps of:
Wherein, Is unmanned aerial vehicle/>The effective switch capacitance of the CPU is a constant;
If time slot Inner user/>Deciding to offload part of its computational tasks to the base station, then offloading the consumed energyAnd calculating the consumed energy/>The method comprises the following steps of:
Wherein, Is a constant.
Under the condition of considering time delay and energy consumption constraint, the offloading tasks are reasonably distributed to the UAV and the ground server, and all users in the network can obtain the maximum average throughput by adjusting the task offloading proportion and optimizing the calculation resource distribution of the UAV. Furthermore, RIS phase offset and 3D trajectory planning of UAV are considered to optimize signal transmission and coverage. Defining offloading decisionsUnloading ratioComputing resources/>RIS phase offset/>UAV trajectory/>In order to maximize the average throughput of the entire network, the optimization problem is modeled as:
(1.1)
(1.2)
(1.3)
(1.4)
(1.5)
(1.6)
(1.7)
(1.8)
(1.9)
(1.10)
(1.11)
Constraint C1.1 is used to limit the proportion calculated locally for each user; constraint C1.2 and C1.3 ensure that at most one user is offloading computing tasks to the edge server per slot to avoid task conflicts and resource contention; constraint C1.4 and C1.5 limit the maximum computing resources per user and UAV, ensuring that the computing resource allocation in the network is within acceptable limits; constraint C1.6 represents the adjustment range of the RIS phase shift; constraint C1.7 represents the initial and final positions of the UAV, ensuring that the UAV's flight trajectory within the time slot is reasonable; constraint C1.8 limits the maximum travel distance of each slot UAV to ensure the feasibility and safety of the flight path; constraint C1.9 ensures minimum safe distance between UAVs to avoid collisions and collisions; constraint C1.10 indicates that each task is to be completed in one time slot, and timeliness and reliability of task completion are ensured; constraint C1.11 represents an energy consumption constraint within a time slot, limiting the energy available to the entire network, ensuring that the computing and communication tasks can be completed under the energy consumption constraint.
And step 6, setting initial values of related optimization variables, and facilitating the first iteration solution of the steps 7 to 10.
Step 7, given RIS phase offsetAnd UAV trajectory/>Solving for offloading decisions/>Unloading ratio/>UAV and user Equipment computing resource Allocation/>The constructed problem is expressed as:
For optimization problems In order to make this optimization problem easier to handle, the/>, will be thereinThe binary variable of (c) is relaxed to a continuous variable, and this process creates the following convex problem:
For convex optimization problem Offloading decision/>Unloading ratio/>UAV and user Equipment computing resource Allocation/>The solution is performed using a convex optimization tool CVX.
Step 8, solving the RIS phase offset matrix according to the current optimal user unloading decision, unloading proportion, user and UAV calculation resource allocation and UAV locus; the constructed problem is expressed as:
For optimization problems To make this optimization problem easier to handle, we will have the/>The binary variable of (c) is relaxed to a continuous variable, and this process creates the following problems:
For convex optimization problem Offloading decision/>Unloading ratio/>Unmanned aerial vehicle and user equipment computing resource allocation/>The solution is performed using a convex optimization tool CVX.
Step 9, giving an offloading decisionUnloading ratio/>UAV and user Equipment computing resource Allocation/>UAV trajectorySolving RIS phase offset/>。
In particular, passive beamforming is used to find phase shiftersIs a solution to the optimization of (3). In time slot/>The channel gain of the entire link of the user-RIS-base station can be expressed as:
。
In time slot To maximize user/>The unloading rate to the base station can be obtained by carrying out phase alignment on the received signals at the signal receiving end to obtain the/>Slot number/>Individual RIS No./>Phase offset of individual reflective elements/>：
；
RIS phase offset，/>For/>Slot number/>Phase offset of each RIS.
Step 10, given offload decisionUnloading ratio/>UAV and user Equipment computing resource Allocation/>RIS phase offset/>Solving for UAV trajectories/>The following problems are constructed:
/>
two deep neural networks were constructed: actor network and critique network, all parameters of actor network being noted as All parameters of the commentator network are noted as/>; Current network state/>Expressed as:
。
When in a certain time slot State/>When the method is used, the deep reinforcement learning algorithm continuously updates actions according to the current state, and the actions are performed in time slots/>Under the action space/>Expressed as:
。
Reward function Expressed as:
。
To solve the optimization problem Obtain optimal UAV locus/>. In time slot/>The input to the actor network is the state/>, of the current networkThe output is the next state/>Action/>Reward/>State/>Average throughput of whole network after each user equipment finishes unloading action selection, wherein action/>Use/>A greedy algorithm makes the selection. State/>Respectively input to the critics network and calculate the time differenceWherein/>，，/>Is a fading factor,/>. In performing selected actions/>After entering the next state/>Obtain rewards/>Actor network parameters/>And critics network parameters/>Is updated according to the update of the update program.
And 11, circularly iterating the steps 7 to 10 by using an alternative optimization algorithm until the absolute value of the energy or time delay of the whole network under two adjacent iterations is smaller than a preset threshold value or the maximum preset iteration times are reached, ending the iteration, obtaining an optimal allocation strategy of the computing resources of the user equipment and the UAV, obtaining an optimal unloading decision and unloading proportion of the computing tasks of the user equipment, and obtaining the RIS phase offset and the track of the optimal UAV. Based on the current state of the whole mobile edge computing network, UAV locus, optimal unloading decision and unloading proportion of computing tasks, RIS phase offset and optimal allocation strategy of computing resources are applied, so that the unloading and computing of computing tasks of all user equipment in a large-scale target area are realized, and the maximization of the average throughput of the whole network is realized.
Figures 3 and 4 show how the variation of the number of users and the amount of tasks, respectively, affects the average throughput for different schemes in embodiments of the present invention. By observing both graphs, it is evident that the average throughput of the various schemes has a tendency to rise gradually as the number of users and the amount of tasks increases. It is particularly notable that the proposed solution of the present invention exhibits significant advantages in this growing trend. The invention has the capability of real-time decision and adjustment because the number of users and the resource allocation situation can change at any time in the complex urban environment. The system can not only carry out intelligent adjustment according to real-time feedback, but also can be rapidly adapted to different user quantity and resource allocation conditions. This high degree of adaptability and dynamics enables the inventive solution to maintain excellent performance in various situations.
The relationship between the average throughput of the network and the number of RIS reflective elements is shown in fig. 5. Simulation results show that as the number of RIS reflective elements increases, the average throughput of the network tends to increase, and that this trend is more pronounced as the number of users increases. Under the constraint conditions of considering the local computing capacity of a user, the UAV computing resource allocation, the total network time delay, the power consumption and the like, more efficient signal transmission and task unloading can be realized by carrying out phase shift optimization on the RIS reflecting element. Increasing the number of RIS reflective elements helps to further enhance the communication link gain, thereby increasing the offloading rate and enhancing the base station task offloading capability of the MEC network.
Figure 6 shows the number of users versus the average throughput of the network for three different trace cases. In a straight line UAV route, which refers to flying straight from one coordinate to another, the increase in the number of users under this scheme has a relatively small impact on average throughput, which is very weak and slow, although the average throughput increases as the number of users increases. This is because the position of the UAV is not adjusted to the change in the number of users and thus the improvement in average throughput is limited. The UAV circular track is a two-dimensional circular track with a region central point as a circle center and a radius length of r, and the average throughput is better represented as the number of users increases than the case of a fixed UAV route. This is because the UAV is able to better cover the user area under a circular trajectory, providing better quality of service. The UAV locus designed by the scheme adopts 3D locus optimization, and compared with other two cases, the performance advantage of the UAV locus is continuously expanded. This is because the proposal of the invention can optimize the trajectory of the UAV according to the user position and real-time requirement so as to ensure the service coverage and QoS to the maximum extent.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above examples, and all technical solutions belonging to the concept of the present invention belong to the protection scope of the present invention. It should be noted that modifications and adaptations to the invention without departing from the principles thereof are intended to be within the scope of the invention as set forth in the following claims.
Claims (1)
1. A method of task offloading multi-UAV-RIS assisted mobile edge computing, the method comprising the steps of:
s1, constructing a large-scale mobile edge computing unloading scene assisted by a plurality of UAVs-RISs according to data collected by a third party;
S2, respectively establishing a user unloading decision, an unloading proportion, an UAV track, RIS phase offset, a communication model, a calculation model, a time delay model and an energy consumption model for user and UAV calculation resource allocation, and establishing an optimization model based on the communication model, the calculation model, the time delay model and the energy consumption model; the communication model is used for calculating the average throughput of the network, the calculation model is used for calculating the representation of tasks, the task quantity and the CPU quantity required by calculation, the time delay model is used for calculating time delay by the UAV server, calculating time delay and unloading time delay by the user, and the energy consumption model is used for calculating the flight energy consumption of the UAV, calculating the energy consumption of two-way transmission and calculating the local energy consumption of the UAV and the user;
s3, setting an initial value of a related optimization variable;
s4, solving a user unloading decision, an unloading proportion and user and UAV calculation resource allocation according to the current optimal RIS phase offset and the UAV locus;
S5, solving the RIS phase offset matrix by adopting an analytic method according to the current optimal user unloading decision, unloading proportion, user and UAV computing resource allocation and UAV locus;
s6, according to the current optimal user unloading decision, unloading proportion, user and UAV calculation resource allocation and RIS phase offset, a deep reinforcement learning algorithm is applied to solve the UAV track;
s7, circularly executing the step S4, the step S5 and the step S6 by using an alternative optimization algorithm until the absolute value of energy or time delay consumed by the whole network between two adjacent iterations is smaller than a preset threshold value or the maximum preset iteration times are reached, ending the iteration, obtaining an optimal allocation strategy of computing resources of user equipment and the UAV, obtaining an optimal unloading decision and unloading proportion of computing tasks of the user equipment, and obtaining RIS phase offset and the track of the optimal UAV;
S8, based on the current state of the whole mobile edge computing network, applying UAV tracks, computing task optimal unloading decisions and unloading proportions, RIS phase offset and computing resource optimal allocation strategies, and unloading and computing all user equipment computing tasks in a large-scale target area with the aim of maximizing the average throughput of the whole network;
In step S1, the process of constructing a multi-UAV-RIS assisted large scale mobile edge computing offload scene from data collected by a third party includes the steps of:
Setting T time slots in a continuous period G period to execute calculation tasks, wherein each time slot is numbered as T; obtaining environmental information by utilizing data collected by a third party, wherein K single-antenna users are randomly distributed in an area with length of l and width of w, and three-dimensional coordinates of the users are expressed as The number of UAVs is M, and the number of the mth UAV isEach UAV is allowed to serve as an air base station to provide unloading service for ground users, and the coordinates of each UAV are as followsThe RIS number is N, each UAV is provided with at least one RIS, and the nth RIS number is N,/>The RIS adopts a uniform linear array to reflect and unload the signals of the users to a ground base station; the 3D coordinates of the base station are fixed to (x B,yB,hB);
setting computational task offloading constraints, comprising:
Defining a new set Representing a location where the user device is allowed to offload; for the feature that the user may offload to a UAV edge server or ground base station, an offload decision variable a k,m (t) is defined:
When a k,m (t) =1, m+.m+1, it is indicated that t-slot user k decides to offload part of its computation task to the mth UAV; a k,m (t) =1, m=m+1 means that t slot user k decides to offload part of its calculation task reflection to the base station; when a k,m (t) =0, it means that t time slot user k does not perform unloading service, and all calculation tasks are selected to perform local calculation; only one user equipment per slot can be offloaded and only to one location, there is
Setting flight constraints, comprising:
Let the 3D coordinates of the mth UAV at slot t be q m(t)＝(xm(t),ym(t),hm (t)); defining the starting and ending positions of the UAV, there is q m(0)＝qinit,qm(T)＝qfinal, Q init represents the UAV start flight position, q final represents the UAV end stop position; the maximum travel distance of the UAV in a slot is limited to: II q m(t+1)-qm(t)‖≤Vmaxt,t＝1,...T-1,Vmax represents the maximum displacement that the UAV can achieve per unit time; the minimum safe distance between UAVs is maintained to prevent UAVs from colliding during flight, ||q m(t)-qj(t)||≥dmin,/>J+.m, d min is the minimum safe distance;
in step S2, the process of constructing a communication model for user unloading decision, unloading ratio, UAV locus, RIS phase shift, and user and UAV computing resource allocation is as follows:
according to Euclidean formula, calculating the distance between different devices in the time slot t:
Wherein d k,m (t) is the distance between the mth UAV and the kth user equipment; d k,n (t) is the distance between the nth RIS and the kth user equipment; d n,B (t) is the distance between the nth RIS and the base station;
The line-of-sight link from the kth user equipment to the mth UAV at time slot t has a channel gain g k,m (t) of:
where α 0 is the channel power gain at a reference distance of 1 m;
The channel gain g k,n (t) from the kth ue to the nth RIS and the channel gain g n,B (t) from the nth RIS to the base station are:
Where d is the spacing between the reflective elements and λ is the carrier wavelength; A cosine value representing the angle of arrival from the kth user to the 2 nd RIS,/> A cosine value representing an angle of arrival from the kth user to the nth RIS; /(I)A cosine value representing a departure angle from the 2 nd RIS to the base station; /(I)A cosine value representing a departure angle from the Mth RIS to the base station;
At time slot t, θ n (t) is the diagonal phase shift matrix of the nth RIS, denoted as:
Wherein θ n,1(t)、θn,2(t)、…、θn,Z (t) is the phase shift of the 1 st reflective element, the 2 nd reflective element, … th reflective element, and the Z-th reflective element of the n-th RIS at time slot t, respectively, and Z is the number of reflective elements in a single RIS;
If user k is served by the mth UAV in time slot t, i.e. a k,m (t) =1, m+.m+1, then the corresponding signal-to-noise ratio r k,m (t) of the mth UAV is expressed as:
Wherein P k represents the transmission power of user k, and sigma 2 is the power of the additive white Gaussian noise of the UAV receiving end; indicating interference caused by transmissions of all other users within the co-channel in time slot t;
The offloading rate R k,m (t) for user k at time slot t is:
If user k gets the base station service at time slot t, i.e. a k,m (t) =1, m=m+1, then the corresponding signal to noise ratio r k,B (t) at the base station is:
The offloading rate R k,B (t) for user k at time slot t is:
Wherein P k represents the transmission power of user k, and sigma 2 is the power of the additive white Gaussian noise of the UAV receiving end; Representing interference caused by transmissions of all other users within the co-channel and reflection of the RIS in time slot t;
In step S2, the process of constructing a calculation model of user unloading decision, unloading ratio, UAV locus, RIS phase shift, and user and UAV calculation resource allocation is as follows:
Each slot has a user facing a computationally intensive task S k (t) to perform, defined as:
Wherein D k (t) represents the amount of data to be processed by user k for time slot t, and F k (t) represents the total number of CPU cycles required to perform the computational task;
Setting the ratio of the local calculation of the user k in the t time slot as rho k (t), setting the divided local task quantity as rho k(t)Dk (t), and the task quantity unloaded to the edge server as (1-rho k(t))Dk (t), and enabling the user to select the ratio between the local calculation and the unloading to the edge server by adjusting the value of rho k (t) so as to meet the limitation of calculation resources and battery capacity;
in step S2, the process of constructing a delay model for user unloading decision, unloading ratio, UAV locus, RIS phase shift, and user and UAV computing resource allocation is as follows:
Let the local calculation proportion of user k be ρ k (t), the offload proportion be 1- ρ k (t), at time slot t, the time required for local calculation of user k The method comprises the following steps:
Wherein, The local computing capacity of the time slot t user k is represented by the CPU number calculated by the time slot t user k;
the computing power of each user in time slot t is:
Wherein, Maximum local computing power for each user;
If user k decides to offload part of his calculation tasks to the mth UAV in time slot t, then offload the required time And calculating the required time/>The method comprises the following steps of:
Wherein, Allowing computing resources allocated to user k at time slot t for the mth UAV;
computing power of each UAV in time slot t The method comprises the following steps:
Wherein, Is the maximum computing resource that each UAV can allocate;
if user k decides to offload part of his task to the base station in time slot t, then the time required for offloading And calculating the required time/>The method comprises the following steps:
Wherein f BS is the computing resource that the base station can allocate to the user;
In step S2, the energy consumption model process of constructing the user unloading decision, unloading proportion, UAV locus, RIS phase shift, and user and UAV computing resource allocation is as follows:
if user k decides to perform the calculation task locally in time slot t, user k calculates the consumed energy locally The method comprises the following steps:
Wherein, Is a coefficient greater than zero for user k;
If user k decides to offload part of his computational tasks to the mth UAV during time slot t, then the consumed energy is offloaded And calculating the consumed energy/>The method comprises the following steps of:
if user k decides to offload part of his calculation tasks to the base station in time slot t, then the consumed energy is offloaded And calculating the consumed energy/>The method comprises the following steps of:
In step S2, the optimization model is:
Defining offloading decisions Unloading ratio/>Computing resource/>RIS phase offset/>UAV trajectory/>
With the average throughput of the whole network maximized as an optimization target, the optimization problem is modeled as:
C1.7:qm(0)＝qinit,qm(T)＝qfinal(1.7)
C1.8:‖qm(t+1)-qm(t)‖≤Vmaxt,t＝1,...T-1(1.8)
Constraint C1.1 is used to limit the proportion calculated locally for each user; constraint C1.2 and C1.3 are used to ensure that at most one user is offloading computing tasks to the edge server per slot to avoid task conflicts and resource contention; constraint C1.4 and C1.5 are used to limit the maximum computing resources per user and UAV to ensure that the computing resource allocation in the network is within acceptable limits; constraint C1.6 is used to represent the adjustment range of RIS phase shift; constraint C1.7 is used for representing the initial position and the final position of the UAV so as to ensure that the flight path of the UAV in a time slot is reasonable; constraint C1.8 is used to limit the maximum travel distance of each slot UAV to ensure the feasibility and safety of the flight path; constraint C1.9 is used to ensure minimum safe distance between UAVs to avoid collisions and collisions; constraint C1.10 is used to indicate that each task is to be completed in one time slot to ensure timeliness and reliability of task completion; constraint C1.11 is used to represent an energy consumption constraint within a time slot, by limiting the energy available to the entire network to ensure that the computation and communication tasks can be completed under the energy consumption constraint;
In step S4, according to the current optimal RIS phase offset and the UAV trajectory, the process of solving the user unloading decision, unloading proportion, and the user and UAV computing resource allocation includes the following steps:
given the RIS phase offset psi and the UAV locus Q, solving an unloading decision A, an unloading proportion rho, the UAV and user equipment computing resource allocation f, and constructing a problem expressed as:
For the optimization problem P2, relaxing the binary variable of C2.2 therein to a continuous variable, creates the following convex problem:
For the convex optimization problem P3, solving an unloading decision A, an unloading proportion rho, a UAV and user equipment computing resource allocation f by using a convex optimization tool CVX;
In step S5, according to the current optimal user unloading decision, unloading proportion, user and UAV computing resource allocation and UAV trajectory, the process of solving the RIS phase shift matrix by using the analytic method includes the following steps:
In time slot t, in order to maximize the unloading rate from user k to the base station, the received signals are phase aligned at the signal receiving end, so as to obtain the phase offset θ n,z (t) of the nth reflection element of the nth time slot n RIS z:
RIS phase offset Θ n (t) is the phase offset of the nth RIS of the nth slot;
in step S6, according to the current optimal user unloading decision, unloading proportion, user and UAV computing resource allocation and RIS phase offset, the process of solving the UAV trajectory by applying the deep reinforcement learning algorithm includes the following steps:
The following problems are constructed:
s.t.C4.7:qm(0)＝qinit,qm(T)＝qfinal(4.1)
C4.8:‖qm(t+1)-qm(t)‖≤Vmaxt,t＝1,...T-1(4.2)
Two deep neural networks were constructed: the system comprises an actor network and a critter network, wherein all parameters of the actor network are marked as theta, and all parameters of the critter network are marked as omega; the current network state S (t) is expressed as:
when the state S (t) is in a certain time slot t, the deep reinforcement learning algorithm continuously updates actions according to the current state, and under the time slot t, an action space A (t) is expressed as:
The bonus function R (t) is expressed as:
R(t)＝(1-ρk(t))Rk,m(t)t+(1-ρk(t))Rk,B(t)t+Dk(t)t；
In the time slot t, the input of the actor network is the state S (t) of the current network, and the output is the next state S (t+1), the action A (t), the reward R (t) and the average throughput of the whole network after each user equipment in the state S (t) finishes the unloading action selection, wherein the action A (t) is selected by using an epsilon-greedy algorithm; inputting states S (t) and S (t+1) to the critics network, respectively, and calculating a time difference g=v t-(R(t)+γVt+1, wherein V t＝V(A(t),S(t);ω(t)),Vt+1 =v (a (t+1), S (t+1); omega (t+1)), gamma is a fading factor, and gamma is more than or equal to 0 and less than or equal to 1; after executing the selected action A (t), entering the next state S (t+1), obtaining the rewards R (t+1), and updating the actor network parameters theta and the commentator network parameters omega.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410166403.7A CN117715117B (en) | 2024-02-06 | 2024-02-06 | Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410166403.7A CN117715117B (en) | 2024-02-06 | 2024-02-06 | Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117715117A CN117715117A (en) | 2024-03-15 |
| CN117715117B true CN117715117B (en) | 2024-05-28 |
Family
ID=90155675
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410166403.7A Active CN117715117B (en) | 2024-02-06 | 2024-02-06 | Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117715117B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114125708A (en) * | 2022-01-20 | 2022-03-01 | 南京信息工程大学 | Unmanned aerial vehicle cluster trajectory optimization and task unloading method based on digital twinning |
| CN115833907A (en) * | 2022-11-17 | 2023-03-21 | 东南大学 | Joint trajectory, unloading and resource allocation optimization method in multi-unmanned aerial vehicle assisted MEC system |
| CN116634544A (en) * | 2023-06-16 | 2023-08-22 | 西安邮电大学 | MEC energy efficiency maximization method for enabling back-scattering RIS auxiliary unmanned aerial vehicle |
| CN116684925A (en) * | 2023-07-24 | 2023-09-01 | 南京图策信息科技有限公司 | Unmanned aerial vehicle-mounted intelligent reflecting surface safe movement edge calculation method |
| CN116866933A (en) * | 2023-07-24 | 2023-10-10 | 重庆邮电大学 | Unmanned aerial vehicle assisted edge computing network resource allocation method based on intelligent reflection surface assistance |
| CN116916337A (en) * | 2023-07-24 | 2023-10-20 | 重庆邮电大学 | Task unloading resource allocation method for unmanned aerial vehicle mobile edge computing network system based on intelligent reflection surface assistance |
-
2024
- 2024-02-06 CN CN202410166403.7A patent/CN117715117B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114125708A (en) * | 2022-01-20 | 2022-03-01 | 南京信息工程大学 | Unmanned aerial vehicle cluster trajectory optimization and task unloading method based on digital twinning |
| CN115833907A (en) * | 2022-11-17 | 2023-03-21 | 东南大学 | Joint trajectory, unloading and resource allocation optimization method in multi-unmanned aerial vehicle assisted MEC system |
| CN116634544A (en) * | 2023-06-16 | 2023-08-22 | 西安邮电大学 | MEC energy efficiency maximization method for enabling back-scattering RIS auxiliary unmanned aerial vehicle |
| CN116684925A (en) * | 2023-07-24 | 2023-09-01 | 南京图策信息科技有限公司 | Unmanned aerial vehicle-mounted intelligent reflecting surface safe movement edge calculation method |
| CN116866933A (en) * | 2023-07-24 | 2023-10-10 | 重庆邮电大学 | Unmanned aerial vehicle assisted edge computing network resource allocation method based on intelligent reflection surface assistance |
| CN116916337A (en) * | 2023-07-24 | 2023-10-20 | 重庆邮电大学 | Task unloading resource allocation method for unmanned aerial vehicle mobile edge computing network system based on intelligent reflection surface assistance |
Non-Patent Citations (2)
| Title |
|---|
| Secure Multi-Layer MEC Systems With UAV-Enabled Reconfigurable Intelligent Surface Against Full-Duplex Eavesdropper;Yi Zhou等;IEEE Transactions on Communications;20231231;第1-12页 * |
| 智能反射面赋能无人机边缘网络计算卸载方案;李斌等;通信学报;20221031;第43卷(第10期);第223-233页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117715117A (en) | 2024-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109068391B (en) | Internet of vehicles communication optimization algorithm based on edge calculation and Actor-Critic algorithm | |
| CN111787509B (en) | Unmanned aerial vehicle task unloading method and system based on reinforcement learning in edge calculation | |
| CN112118287B (en) | Network resource optimization scheduling decision method based on alternative direction multiplier algorithm and mobile edge calculation | |
| CN114422363B (en) | Capacity optimization method and device for unmanned aerial vehicle-mounted RIS auxiliary communication system | |
| Zhai et al. | Energy-efficient UAV-mounted RIS assisted mobile edge computing | |
| CN113316169B (en) | UAV auxiliary communication energy efficiency optimization method and device for smart port | |
| CN114650567B (en) | Unmanned aerial vehicle auxiliary V2I network task unloading method | |
| CN113254188B (en) | Scheduling optimization method and device, electronic equipment and storage medium | |
| CN116684925B (en) | Unmanned aerial vehicle-mounted intelligent reflecting surface safe movement edge calculation method | |
| CN109991998A (en) | A kind of unmanned plane relaying route optimization method based on received signal strength | |
| CN116436512A (en) | Multi-objective optimization method, system and equipment for RIS auxiliary communication | |
| CN116208619A (en) | Intelligent reflection surface-assisted Internet of vehicles safety calculation unloading method, system, equipment and medium | |
| Qu et al. | Empowering the edge intelligence by air-ground integrated federated learning in 6G networks | |
| Hwang et al. | Deep reinforcement learning approach for uav-assisted mobile edge computing networks | |
| Tang et al. | Machine learning for space–air–ground integrated network assisted vehicular network: A novel network architecture for vehicles | |
| Zhou et al. | Flying IRS: QoE-driven trajectory optimization and resource allocation based on adaptive deployment for WPCNs in 6G IoT | |
| CN117715117B (en) | Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation | |
| CN116208968B (en) | Track planning method and device based on federal learning | |
| Lee et al. | Multi-Agent Reinforcement Learning in Controlling Offloading Ratio and Trajectory for Multi-UAV Mobile Edge Computing | |
| CN116723548A (en) | Unmanned aerial vehicle auxiliary calculation unloading method based on deep reinforcement learning | |
| CN116669073A (en) | Resource allocation and track optimization method based on intelligent reflecting surface auxiliary unmanned aerial vehicle cognitive network | |
| Sheltami et al. | An optimization approach of iod deployment for optimal coverage based on radio frequency model | |
| CN116528250A (en) | Unmanned aerial vehicle auxiliary MEC resource optimization method based on NOMA | |
| CN116321181A (en) | Online track and resource optimization method for multi-unmanned aerial vehicle auxiliary edge calculation | |
| CN115664486A (en) | Energy efficiency optimization method for wireless energy supply in RIS (RIS) assisted UAV (unmanned aerial vehicle) edge computing system |
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 |