CN115134364B - Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things system - Google Patents
Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things system Download PDFInfo
- Publication number
- CN115134364B CN115134364B CN202210741451.5A CN202210741451A CN115134364B CN 115134364 B CN115134364 B CN 115134364B CN 202210741451 A CN202210741451 A CN 202210741451A CN 115134364 B CN115134364 B CN 115134364B
- Authority
- CN
- China
- Prior art keywords
- internet
- things
- delay
- unloading
- ran
- 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 49
- 238000005265 energy consumption Methods 0.000 claims abstract description 82
- 238000012545 processing Methods 0.000 claims abstract description 54
- 238000004364 calculation method Methods 0.000 claims abstract description 48
- 238000013468 resource allocation Methods 0.000 claims abstract description 10
- 230000005540 biological transmission Effects 0.000 claims description 57
- 238000007726 management method Methods 0.000 claims description 40
- 238000005457 optimization Methods 0.000 claims description 32
- 230000008569 process Effects 0.000 claims description 15
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000001934 delay Effects 0.000 claims description 7
- 230000014509 gene expression Effects 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 4
- 239000000654 additive Substances 0.000 claims description 3
- 230000000996 additive effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 239000013256 coordination polymer Substances 0.000 claims description 3
- 235000019580 granularity Nutrition 0.000 abstract description 6
- 239000002699 waste material Substances 0.000 abstract description 5
- 230000003287 optical effect Effects 0.000 abstract description 3
- 206010063385 Intellectualisation Diseases 0.000 abstract description 2
- 230000001105 regulatory effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 16
- 101100194706 Mus musculus Arhgap32 gene Proteins 0.000 description 9
- 101100194707 Xenopus laevis arhgap32 gene Proteins 0.000 description 9
- 238000011160 research Methods 0.000 description 5
- 230000007423 decrease Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000004134 energy conservation Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000006855 networking Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 102100037651 AP-2 complex subunit sigma Human genes 0.000 description 1
- 101000806914 Homo sapiens AP-2 complex subunit sigma Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013461 design 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
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- 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
Abstract
The invention discloses an energy-saving computing unloading system based on an O-RAN (optical network access network) internet of things system, under the O-RAN based internet of things system, the participation and division of a near-real-time energy controller and a non-real-time intelligent controller can obtain the optimal unloading and resource allocation strategy according to global and local information, and the global and local information of the O-RAN are regulated and optimized in real time according to different time granularities and collected, so that congestion is better reduced, and resource waste is avoided; in addition, in order to fully utilize the advantages of the O-RAN architecture and realize the intellectualization and flexibility of the Internet of things system, the invention also considers the energy-saving calculation unloading method based on the O-RAN Internet of things system, and the optimal energy-saving calculation unloading strategy based on the local processing speed, the transmitting power and the local unloading rate can minimize the energy consumption of the Internet of things terminal equipment while meeting the time delay requirement.
Description
Technical Field
The invention relates to the technical field of computing and unloading, in particular to an energy-saving computing and unloading system and method based on an O-RAN (O-radio network access network) internet of things system.
Background
With the development of the next generation mobile communication network, some future scenes may put very strict requirements on the internet of things system, such as mass connection, ultra-low delay and the like; it requires more flexible and intelligent network deployment and management; unfortunately, these extreme requirements are difficult to meet using existing network architectures. The introduction of new network architecture into internet of things systems is urgent.
With the increasing task of internet of things devices being computationally intensive (e.g., intelligent transportation and intelligent medical), most internet of things devices have limited computing power and battery power that can become bottlenecks in the system. Generally, researchers have designed computational offloading based on traditional network architectures to increase computational speed, save energy, and reduce latency.
The traditional partial offloading policy minimizes the energy consumption of the internet of things taking into account the delay limit, transmission power limit, maximum CPU cycles and memory of the MEC network; or unloading and resource allocation combined optimization scheme based on spectrum efficiency, so that the problem of profit maximization under the constraints of load delay, front-end transmission capacity, limited bandwidth and calculation resources is solved; however, all of the above work only considers the application of computational offload in traditional MEC network architectures, resulting in delays and reduced power consumption to bottlenecks.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the traditional network architecture and partial unloading strategy solve the problem of profit maximization under the constraints of load delay, front-end transmission capacity, limited bandwidth and computing resources; however, when the problems are solved, only the application of computing and unloading in the traditional MEC network architecture is considered, and the comprehensive utilization of global and local information is not considered to obtain an optimal unloading strategy, so that the delay and the energy consumption of the whole network are reduced to the bottleneck, and the resource consumption is huge; the invention aims to provide an energy-saving computing unloading system based on an O-RAN (optical network access network) internet of things system, under the O-RAN based internet of things system, the participation and division of a near-real-time energy controller and a non-real-time intelligent controller can obtain an optimal unloading and resource allocation strategy according to global and local information, and the global and local information of the O-RAN are regulated and optimized in real time according to different time granularities and collected, so that congestion is better reduced, and resource waste is avoided; in addition, in order to fully utilize the advantages of the O-RAN architecture and realize the intellectualization and flexibility of the Internet of things system, the invention also considers the energy-saving calculation unloading method based on the O-RAN Internet of things system, and the optimal energy-saving calculation unloading strategy based on the local processing speed, the transmitting power and the local unloading rate can minimize the energy consumption of the terminal equipment of the Internet of things while meeting the time delay requirement.
The invention is realized by the following technical scheme:
the invention provides an energy-saving computing unloading system based on an O-RAN (O-radio network access network) internet of things system, wherein the O-RAN internet of things system comprises an O-RAN and an O-Cloud; the O-RAN is provided with a near real-time intelligent controller and a non-real-time intelligent controller;
the terminal equipment of the Internet of things requests the O-RAN to calculate and unload tasks;
the non-real-time intelligent controller is used for collecting non-real-time information of the O-RAN internet of things system and carrying out data reasoning analysis, wireless resource management and energy-saving calculation unloading strategy optimization based on the non-real-time information;
the near real-time intelligent controller is used for collecting real-time information of the O-RAN internet of things system, carrying out decision adjustment and decision execution by combining the optimized energy-saving calculation unloading strategy, and finally sending an execution result back to the internet of things terminal equipment.
The working principle of the scheme is as follows: performing compute offloading under Mobile Edge Computing (MEC) network architecture is a widely accepted method to reduce internet of things system latency and energy consumption; however, under the traditional network architecture, it is difficult to comprehensively utilize global and local information to obtain an optimal offloading policy on the distributed server, which may lead to delay and reduction of energy consumption to a bottleneck; in order to meet the requirements of ultralow delay and ultralow energy consumption of future internet of things systems, the scheme provides an energy-saving computing unloading system and method based on an O-RAN (optical network access network) internet of things system, an O-RAN architecture is adopted to deploy the internet of things system, under the O-RAN based internet of things system, participation and division of a near-real-time energy controller and a non-real-time intelligent controller can obtain optimal unloading and resource allocation strategies according to global and local information, and according to different time granularities, the strategies are adjusted and optimized in real time by collecting the global and local information of the RAN, so that congestion is reduced better, and resource waste is avoided.
Compared with the traditional network architecture, the O-RAN Internet of things system provides an intelligent, flexible and interoperable standard network architecture; performing computation offload under the O-RAN architecture may solve some of the problems encountered in conventional network architectures, thereby effectively improving the performance of the internet of things system; specifically, under a traditional network architecture, when performing computing offload, it is impossible to collect local and global real-time information simultaneously (e.g., the number of real-time tasks each internet of things needs to handle and the current available resources of all distributed units DU cannot be obtained simultaneously); in this case we cannot obtain the best offloading and resource allocation strategy, which would lead to a suboptimal reduction of delay and energy consumption; however, the participation and division of near real-time intelligent controllers and non-real-time intelligent controllers (RIC) in the O-RAN architecture may solve this problem; the two intelligent controllers can adjust and optimize strategies in real time by collecting global and local information of the O-RAN according to different time granularities, thereby reducing congestion better, avoiding resource waste and improving the energy efficiency of the Internet of things.
The non-real-time information includes information such as the number of terminal devices served, the uplink channel, the quality of service requirements of the devices, server resources of the control unit CUs and the distributed units DUs,
The near-real-time intelligent controller is responsible for collecting and analyzing real-time information, so that the near-real-time intelligent controller RICs can monitor the behavior change of the Internet of things equipment and conduct real-time adjustment, and only the execution result can be transmitted to the Internet of things terminal equipment.
A further optimization scheme is that the O-RAN is further configured to: service management and traffic management framework SMO, control unit CUs, distributed units DUs and radio units RUs; all wireless unit RUs contain transceiver antennas for performing physical layer operations;
the control unit CUs is used for operation of the distributed units DUs and the wireless units RUs;
the near real-time intelligent controller is deployed in the control unit CUs, and the non real-time intelligent controller is deployed in the service management and service management framework SMO;
all the terminal devices of the internet of things send a calculation offloading task request to the distributed units DUs through the connected wireless units RUs.
Consider distributed units DUs (O-DUs) and O-closed in the internet of things device IoTD, O-RAN to cooperatively handle computationally intensive tasks; meanwhile, by arranging a control unit CUs (O-CU) in the O-RAN and separating a control plane from a data plane, hierarchical management of the system is realized through a standardized communication interface; under the O-RAN architecture, the service management and traffic management framework SMO is mainly responsible for managing network functions and coordination control, the control unit CUs is responsible for handling higher level protocols, the connection between the RUs and DUs is a high capacity front-end link,
A further optimization scheme is that the O-RAN is further configured to: an A1 interface, an O2 interface, an F1-C interface and an F1-U interface;
the non-real-time intelligent controller provides an optimized energy-saving calculation unloading strategy for the near-real-time intelligent controller through an A1 interface;
the service management and service management framework SMO manages the control unit CUs, the distributed units DUs and the radio units RUs through the O1 interface;
the service management and service management framework SMO is connected with the O-group through an O2 interface, and intelligent management and operation are provided for the O-group;
the distributed units DUs are connected to the user data plane CU-UP of the control unit CUs via an F1-C interface and to the control plane CU-CP of the control unit CUs via an F1-U interface.
The further optimization scheme is that the energy-saving calculation unloading strategy after optimization is as follows:
optimal unloading ratio lambda of mth Internet of things terminal equipment m The method comprises the following steps:
optimal transmitting power P of mth Internet of things terminal equipment m The method comprises the following steps:
in the formula tm For total delay time, B for total available bandwidth, D m Is the input data size of the mth Internet of things terminal equipment,to delay in distributed units DUs or O-groups after policy enforcement, T MAX For the maximum time delay of the tasks of the Internet of things equipment, mu is Lagrangian multiplier and is- >Г m and βm Is a different equation related to the signal-to-noise ratio during the salifying process.
The invention also provides an energy-saving calculation unloading method based on the O-RAN internet of things system, which is applied to the system and comprises the following steps:
s1, an Internet of things terminal device requests a calculation unloading task to an O-RAN;
s2, a non-real-time intelligent controller collects non-real-time information of an O-RAN (Internet of things) system and performs data reasoning analysis, wireless resource management and energy-saving calculation unloading strategy optimization based on the non-real-time information;
and S3, the near-real-time intelligent controller collects real-time information of the O-RAN internet of things system, performs decision adjustment and decision execution by combining the optimized energy-saving calculation unloading strategy, and finally sends an execution result back to the internet of things terminal equipment.
We assume that each internet of things terminal device has a computationally intensive task (autopilot, collaborative computing, etc.) that needs to be completed. Different tasks have different request numbers, quality of service, data size, and computational power requirements. For example, an automatic driving task has high requirements on service quality and a large request amount; the collaborative computing task needs to consume a large amount of computing resources due to the arrival of a large amount of tasks; although performing computational offloading may reduce energy consumption while meeting the latency requirements of the terminal device tasks; however, how to design an optimal computational offload strategy remains a challenging problem given the heterogeneity of tasks and the limited resources of task units.
Based on the O-RAN internet of things system, in order to minimize the energy consumption of the internet of things terminal equipment, the system is based on meeting the delay requirement of each task.
Firstly, an internet of things terminal device (IoTDs) requests computing tasks to an O-RAN, wherein the tasks need to be offloaded;
secondly, the non-real-time intelligent controller RICs collects information such as the number of IoTDs, an unloading strategy, the distance and computing capacity of a control unit CUs and O-cloud, and a distributed unit DUs and an application server of available computing resources, and utilizes the collected information to analyze data in a non-real-time manner, and a wireless resource management and optimization strategy to be deployed in the non-real-time intelligent controller;
then, the non-real-time intelligent controller downloads analysis, management and optimization results to the near-real-time intelligent controller; the near real-time intelligent controller is responsible for collecting and analyzing the real-time change of the arrived task, the number of requests and the resources required by the service quality of the task;
and then, the near-real-time intelligent controller combines the global information and the strategy optimization provided by the non-real-time intelligent controller, monitors the dynamic change of the terminal equipment (IoTDs) of the Internet of things in real time, optimizes the energy-saving calculation unloading strategy (ECO), namely the unloading ratio, and adjusts and makes a decision. Local processing speed and transmission power of the internet of things terminal devices (IoTDs), and whether tasks are allocated to distributed units DUs or O-groups to minimize energy consumption while meeting latency requirements.
And finally, only the execution result is sent back to the terminal equipment of the Internet of things through the downlink channel so as to reduce congestion.
The further optimization scheme is that the energy-saving calculation unloading strategy optimization process comprises the following steps:
t1, constructing a channel model, a time delay model, an energy consumption model and an unloading model based on an O-RAN (O-radio network access network) internet of things system;
and T2, taking the calculation unloading task as a non-convex optimization problem to consider the energy consumption and delay requirement, and optimizing to obtain an energy conservation calculation unloading strategy by solving the problem.
Assume that there are M internet of things terminal devices (IOTDs), M e {1, 2..m }; in the case of full frequency multiplexing, the frequency spectrums used by the radio units RUs are overlapping, and thus the interference inside the radio units RUs should be taken into account. However, when accessing the same radio units RUs, the spectrum is allocated orthogonally to the terminal device. Therefore, the interference inside the radio units RUs is negligible, assuming that all the downlinks between each radio unit RU and the terminal devices IoTDs of the internet of things have complete channel state information CSI;
the signal-to-noise ratio SINR transmitted by the wireless unit RUr to the terminal device IoTD m of the internet of things is:
wherein ,Pm The transmitting power g of the terminal equipment m of the Internet of things m,r For the channel gain from the wireless unit RUr to the terminal device m of the internet of things, σ2 is the power of the additive white gaussian noise; assuming that the distribution is gaussian, according to Shannon's formula, the channel model is the propagation rate V of the wireless unit RUr to the terminal device m of the internet of things on the wireless link channel m,r :
Wherein B is the total available bandwidth and R represents the total number of wireless units RUs;
the time delay model comprises: a queuing delay model, a transmission delay model and a processing delay model; since O-closed has rich computing resources, the queuing delay of O-closed is not considered.
The queuing delay of each distributed unit DU is modeled as an M/M/1 queue of distributed units DU, where the first M represents the traffic arrival rate subject to a Poisson distribution and the second M represents the service rate exponentially negative. Queue delay calculation based on traffic and node capacity, queuing delay at distributed units DUsThe method comprises the following steps:
wherein ,Yd Representing the traffic arrival rate, X, of a distributed unit DUs d Representing the processing rate of the distributed units DUs.
The transmission delay model includes a transmission delay to a distributed unit DUsAnd a transmission delay to O-closed->
Setting the input data size of the mth IOT terminal equipment IOTD as D m An unloading ratio of 1-lambda m (0<λ m < 1), i.e. the amount of data executed locally is lambda m D m The amount of data offloaded is (1-lambda m )D m The method comprises the steps of carrying out a first treatment on the surface of the The front-end link transmission is transmitted by the wireless unit RUr to the distributed unit DUd, with the front-end link transmission capability being C r,d The method comprises the steps of carrying out a first treatment on the surface of the Then
wherein ,dm Representing the delay caused by the wired transmission to the O-closed;
the processing delay model comprises processing delay local to the terminal equipment of the Internet of thingsProcessing delay in distributed units DUs>And processing delay on O-closed +.>
Definition of the definition and />Respectively representing the computing power of IoTDs, the computing power of DUs and the computing power of O-group; is provided with->Then:
the further optimization scheme is that the energy consumption model comprises: transmission energy consumption to distributed units DUsTransmission energy consumption for transmission to O-closed>And local processing energy consumption->
wherein ,rm And the epsilon is a chip factor for the energy consumption generated by occupying a channel after the data transmission is completed.
A further optimization is that, in order to make our system perform better, two cost metrics are defined to decide whether the computationally intensive task should be transferred to the distributed unit DU or O-group; the option of offloading each device is defined as a m, wherein am =1 means that the task is offloaded to O-closed, a m =0 means that the task is offloaded to the distributed unit DUs;
the delay costs in the distributed units DU and O-closed are calculated as follows:
thus, the total cost and />This means that the task of setting the IOTD m for the internet of things terminal is offloaded to the distributed units DU and O-closed, respectively.
wherein , the terminal can be adjusted according to the state of the terminal of the Internet of things; if the task has a high latency requirement, it is possible to add +.>A bit is enlarged; if the device has less battery remaining +.>And (5) enlarging a bit.
If the task of the mth Internet of things terminal equipment is selected to be offloaded to the O-group, an offloading model of the mth Internet of things terminal equipmentThe method comprises the following steps:
if the task of the mth Internet of things terminal device is selected to be offloaded to the distributed unit DU, an offloading model thereofThe method comprises the following steps:
firstly, the data size of fixed unloading of the mth Internet of things terminal equipment is as followsPower of completed transmission->For simplicity we first ignore the delay of the front-end link so we can calculate a sum d m Related threshold d thr :
If d m <d thr Then a m =1; otherwise, a m =0。
The further optimization scheme is that the energy-saving calculation unloading strategy after T2 optimization is as follows:
optimal unloading ratio lambda of mth Internet of things terminal equipment m The method comprises the following steps:
optimal transmitting power P of mth Internet of things terminal equipment m The method comprises the following steps:
in the formula tm For total delay time, B for total available bandwidth, D m Is the input data size of the mth Internet of things terminal equipment,to delay in distributed units DUs or O-groups after policy enforcement, T MAX For the maximum time delay of the tasks of the Internet of things equipment, mu is Lagrangian multiplier and is->Γ m and βm Is a different equation related to the signal-to-noise ratio during the salifying process.
In the scheme, the calculation unloading problem is taken as a non-convex optimization problem to consider the energy consumption and delay requirement, and the calculation unloading strategy of energy conservation is obtained by solving the problem.
A. Problem formula
According to the channel model, the time delay model, the energy consumption model and the unloading model, the total energy consumption of the mth Internet of things terminal equipment task can be calculated as e m A total delay time of t m :
Comprehensively considering local calculation speedOffloading policy->Transmitting powerUnloading ratio->Obtain the weighted sum of the IoTDs energy consumption of all the terminal devices of the Internet of thingsThe weighted sum may be considered an energy consumption tradeoff between all internet of things devices; our goal is to consider F L In the case of A, P and λ, a combined strategy is obtained that minimizes energy consumption; the problems established are expressed as follows:
s.t:
wherein C1 For maximum local processing speed constraint, C 2 For non-negative transmit power constraints, C 3 To unload the ratio constraint, C 4 For maximum delay constraint, C 5 and C6 For offloading policy constraints; due to the problem P 1 Is non-convex, we divide it into two sub-problems, namely confirm the unloading point among terminal equipment IoTDM of the thing networking, distributed unit O-DU, O-closed, and distribute unloading rate, transmission power and local processing speed according to the unloading point confirmed; the continuous convex optimization framework is adopted to perform convex approximation on the two sub-problems, and the non-convex optimization problem is solved by iteratively solving the two sub-problems.
B. Problem analysis
For problem P 1 Firstly, the adjustment of local processing speed is focused on, and the optimal calculation speed of each terminal device of the Internet of things is obtained in a closed form. Due to e m Along with itMonotonically increasing, so that it is possible to reduce/>And reduce the dimension of the original problem to solve for P 1 Can be expressed as:
so that
In the calculation unloading model part, we describe in detail the two costs of energy consumption and delay to determine whether the task of the IOTDM is to be unloaded to the distributed units DUs or O-Cloud; so we can according to Calculate d thr And comparing the values of d thr and dm Searching for the value of A, defined as P 2 :
P 2 :
s.t:C 4 ,C 5 ,C 6
wherein C7 From C 1 And (21) obtained, C 8 C for local processing speed constraint 9 Processing speed constraints for maximum distributed units DUs, C 10 and C11 Respectively restricting the maximum task number and the maximum transmitting power of the IOTDM terminal equipment. Then, the value of A is searched according to P2, which can be expressed asP 1 and P2 Can be converted into P 3 :
P 3 :
s.t:C 2
C 12 and C13 Converted from (22) and (20); according toWe mean +.>Representing the delay in the distributed units DUs or O-groups after executing the policy, including queuing delay and processing delay. Thus C 13 Can be expressed as:
since (23), the problem is still non-convex, it can be divided into (24) and (25) according to the non-convexity, expressed as:
non-convex:
convex:
wherein ,
h m ′(P m )+h m ′(λ m )≤0, (26)
we use a continuous inward-projecting framework to optimize a series of approximate convex problems, which allows the development of a computationally efficient algorithm. The algorithm utilizes an iterative method to solve a series of convex optimization problems, and is similar to solving the original problems, and a standing point of the original problems is found. In each iteration, the non-convex objective function and constraints are replaced by appropriate convex approximations by solving a convex approximation solution of the original problem. Segment(s)We have:
wherein At the kth th Each convexityThe sequence is denoted->For h m ′(q m (k) ) We require that the following three properties be met:
Arbitrary convergence sequence q m (k) The limit of (2) is a Karush-Kuhn-Tucker (KKT) point. These properties mean that the sub-problem is constructed approximately from convex substitution functions of the original objective function and convex subsets within the feasible solution.
Kth th The iteration point is q m (k) Constructing h on the basis m ′(q m ) Is a proxy function of (a). On the kth salified sequence, according to (30), q is locally preserved m (k) Middle h m ′(q m ) And is strongly convex, can be expressed as:
equation (29) represents h m ′(q m (k) ) At any feasible point it need not be its own upper bound, then there is:
thus:
through k th The sequence is raised so that a proper h can be found m ′(q m ) Is a similar value to (a) in the above. We can put P 3 Conversion to P 4 Can be expressed as:
P 4 :
s.t:
C 14 transformed by (36), C 15 Similar to C 2 . Then, solve the convex problem P using Lagrangian multiplier 4 The closed form expression is found using the KKT condition.
The KKT condition is given by:
the process of updating the resource allocation and the lagrangian multiplier is repeated until a predetermined maximum number of iterations is converged or reached, i.e. a minimum value is reached.
The one-to-one correspondence between transmission power and unloading ratio is established;
the expressions (42) and (43) of the optimal unloading ratio are finally obtained.
In order to make the optimization problem easy to process, we decouple the original problem into two sub-problems and solve the non-convex problem in the joint optimization by iteratively solving the two sub-problems; experimental results show that under the O-RAN-based Internet of things system, the average energy consumption of all Internet of things equipment can be reduced by more than 20% by using the obtained strategy.
Compared with the prior art, the invention has the following advantages and beneficial effects:
according to the energy-saving computing unloading system and method based on the O-RAN (object-oriented radio network) system, the O-RAN architecture is adopted to deploy the Internet of things system, under the O-RAN-based Internet of things system, the participation and division of the near-real-time energy controller and the non-real-time intelligent controller can obtain the optimal unloading and resource allocation strategy according to global and local information, and the strategy is adjusted and optimized in real time by collecting the global and local information of the RAN according to different time granularities, so that congestion is better reduced, and resource waste is avoided.
Drawings
In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings that are needed in the examples will be briefly described below, it being understood that the following drawings only illustrate some examples of the present invention and therefore should not be considered as limiting the scope, and that other related drawings may be obtained from these drawings without inventive effort for a person skilled in the art. In the drawings:
fig. 1 is a schematic diagram of an energy-saving computing and unloading system based on an O-RAN internet of things system;
FIG. 2 is a schematic diagram of a data flow of an energy-saving computing offload system based on an O-RAN Internet of things system;
FIG. 3 is a graph showing average energy consumption versus different strategies;
FIG. 4 is a schematic diagram showing the comparison of time delays of different strategies under the same energy consumption;
FIG. 5 is a graph showing average energy consumption under different delay constraints;
FIG. 6 is a graph showing average energy consumption comparisons at different local processing speeds;
FIG. 7 is a graph of local processing speed, transmission power, and unloading ratio variation;
FIG. 8 is a pseudo code schematic;
FIG. 9 shows other simulation parameters for intent.
Detailed Description
For the purpose of making apparent the objects, technical solutions and advantages of the present invention, the present invention will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present invention and the descriptions thereof are for illustrating the present invention only and are not to be construed as limiting the present invention.
With rapid developments in Software Defined Networking (SDN), network Function Virtualization (NFV), dynamic function decomposition, high-capacity data centers, and cloud computing, conventional network architectures have failed to support various functions and business requirements due to a single function, lacking sufficient flexibility. Thus, the O-RAN is generated on the basis of the original C-AN and V-RAN. The C-RAN has previously been considered as one of the potential technologies for satisfying the 5G underlying radio access, mainly using a baseband unit (BBU) pool to share networks and resources, achieving flexible scheduling and improving efficiency. As C-RANs support software definition and function virtualization, the evolution is proceeding to V-RANs. The V-RAN increases the scalability and flexibility of the wireless system based on the C-RAN, overcoming some drawbacks in terms of radio interference and functionality. Many scholars have studied on the basis of C-RAN and V-RAN with great success. However, in face of the increasing demands and complexity of wireless networks, research efforts based on C-RAN and V-RAN network architectures increasingly exhibit some limitations.
Open, intelligent has become the subject of next generation wireless access networks, and is also a necessary condition for deployment and operation of next generation wireless networks. Thus, the advent of O-RANs is based on openness and intelligence. First, the O-RAN virtualizes Base Station (BSs) functions as network functions, dividing the network functions into a plurality of network nodes. The plurality of network nodes are the control units CUs, distributed units DUs and wireless units RUs mentioned herein before, which help to improve efficiency by performing different network processes. This means that the higher layer control unit CUs handles operations with a larger time granularity to implement functions, the lower layer distributed units DUs handle time critical operations, and the wireless units RUs manage Radio Frequency (RF) components and lower layer physical layer (PHY) components. Second, the O-RAN technology separates software, hardware and vendors, which define the open interfaces of the control unit CUs, distributed units DUs and wireless units RUs through the functions of the hardware and software, and improves reliability and usability through modularization and software-based capacity management. Then, the O-RAN can be designed simply and quickly by expanding software, thereby reducing the construction cost of the RAN and improving the flexibility of the RAN. Next, the O-RAN also embeds intelligence and expands SDN to optimize performance and reduce operational complexity. Finally, the O-RAN divides the control unit CUs into a control plane and a user plane, enabling more efficient control and management. And intelligent controller RICs, including non-real-time RICs and near-real-time RICs, are introduced, allowing operators to customize the implementation and deployment of control plane functions according to the operator's requirements to better utilize resources.
B. Computing offloaded related work
In recent years, computing offloading under different network architectures has received increasing attention, and a great deal of research work has been carried out on computing offloading under different network architectures, in which some research is mainly focused on designing computing offloading policies in architectures such as C-RAN, V-RAN, MEC, etc., and solving the effective computing offloading problem in combination with offloading rate, computing resource allocation or radio resource allocation. At present, no research is yet performed on meeting the time delay requirement of an O-RAN-based internet of things system through calculation and unloading, and reducing the energy consumption of IoTDs.
Example 1
The embodiment provides an energy-saving computing unloading system based on an O-RAN (O-radio access network) internet of things system, wherein the O-RAN internet of things system comprises an O-RAN and an O-Cloud; the O-RAN is provided with a near real-time intelligent controller and a non-real-time intelligent controller;
the terminal equipment of the Internet of things requests the O-RAN to calculate and unload tasks;
the non-real-time intelligent controller is used for collecting non-real-time information of the O-RAN internet of things system and carrying out data reasoning analysis, wireless resource management and energy-saving calculation unloading strategy optimization based on the non-real-time information;
the near real-time intelligent controller is used for collecting real-time information of the O-RAN internet of things system, carrying out decision adjustment and decision execution by combining the optimized energy-saving calculation unloading strategy, and finally sending an execution result back to the internet of things terminal equipment.
The O-RAN is further configured to: service management and traffic management framework SMO, control unit CUs, distributed units DUs and radio units RUs; all wireless unit RUs contain transceiver antennas for performing physical layer operations;
the control unit CUs is used for operation of the distributed units DUs and the wireless units RUs;
the near real-time intelligent controller is deployed in the control unit CUs, and the non real-time intelligent controller is deployed in the service management and service management framework SMO;
all the terminal devices of the internet of things send a calculation offloading task request to the distributed units DUs through the connected wireless units RUs.
The O-RAN is further configured to: an A1 interface, an O2 interface, an F1-C interface and an F1-U interface;
the non-real-time intelligent controller provides an optimized energy-saving calculation unloading strategy for the near-real-time intelligent controller through an A1 interface;
the service management and service management framework SMO manages the control unit CUs, the distributed units DUs and the radio units RUs through the O1 interface;
the service management and service management frames SMO and O-closed are connected through an O2 interface;
the distributed units DUs are connected to the user data plane CU-UP of the control unit CUs via an F1-C interface and to the control plane CU-CP of the control unit CUs via an F1-U interface.
The optimized energy-saving calculation unloading strategy is as follows:
optimal unloading ratio lambda of mth Internet of things terminal equipment m The method comprises the following steps:
optimal transmitting power P of mth Internet of things terminal equipment m The method comprises the following steps:
in the formula tm For total delay time, B for total available bandwidth, D m Is the input data size of the mth Internet of things terminal equipment,to delay in distributed units DUs or O-groups after policy enforcement, T MAX For the maximum time delay of the tasks of the Internet of things equipment, mu is Lagrangian multiplier and is->Г m and βm Is a different equation related to the signal-to-noise ratio during the salifying process.
Example 2
The present embodiment provides an energy-saving computing unloading method based on an O-RAN internet of things system, which is applied to the system described in the previous embodiment, and includes the steps of:
s1, an Internet of things terminal device requests a calculation unloading task to an O-RAN;
s2, a non-real-time intelligent controller collects non-real-time information of an O-RAN (Internet of things) system and performs data reasoning analysis, wireless resource management and energy-saving calculation unloading strategy optimization based on the non-real-time information;
and S3, the near-real-time intelligent controller collects real-time information of the O-RAN internet of things system, performs decision adjustment and decision execution by combining the optimized energy-saving calculation unloading strategy, and finally sends an execution result back to the internet of things terminal equipment.
The energy-saving calculation unloading strategy optimization process comprises the following steps:
t1, constructing a channel model, a time delay model, an energy consumption model and an unloading model based on an O-RAN (O-radio network access network) internet of things system;
and T2, taking the calculation unloading task as a non-convex optimization problem to consider the energy consumption and delay requirement, and optimizing to obtain an energy conservation calculation unloading strategy by solving the problem.
The channel model is the propagation rate V of the wireless unit RUr to the terminal equipment m of the Internet of things on the wireless link channel m,r :
Wherein B is the total available bandwidth, R is the terminal equipment number of the Internet of things, and p m The transmitting power g of the terminal equipment m of the Internet of things m,r For the channel gain from the wireless unit RUr to the terminal equipment m of the Internet of things, sigma 2 is the power of additive Gaussian white noise;
the time delay model comprises: a queuing delay model, a transmission delay model and a processing delay model;
the transmission delay model includes a transmission delay to a distributed unit DUsAnd a transmission delay to O-closed->
Let the input data size of the mth Internet of things terminal device be D m An unloading ratio of 1-lambda m (0<λ m < 1), the front-end link transmission is that the wireless unit RU r is transmitted to the distributed unit DU d, and the front-end link transmission capacity is C r,d The method comprises the steps of carrying out a first treatment on the surface of the Then
wherein ,dm Representing the delay caused by the wired transmission to the O-closed;
the processing delay model comprises processing delay local to the terminal equipment of the Internet of thingsProcessing delay in distributed units DUs>And processing delay on O-closed +.>
Definition of the definition and />Respectively representing the computing power of the terminal equipment of the Internet of things, the computing power of the distributed units DUs and the computing power of the O-closed; is provided with->Then:
wherein ,Yd Representing the traffic arrival rate, X, of a distributed unit DUs d Representing the processing rate of the distributed units DUs.
The energy consumption model comprises: transmission energy consumption to distributed units DUsTransmission energy consumption for transmission to O-closed>And local processing energy consumption->
wherein ,rm And the epsilon is a chip factor for the energy consumption generated by occupying a channel after the data transmission is completed.
The option of offloading each device is defined as a m, wherein am =1 means that the task is offloaded to O-closed, a m =0 means that the task is offloaded to the distributed unit DUs;
selecting to offload tasks of mth Internet of things terminal equipment to O-group, and offloading model thereofThe method comprises the following steps:
selecting to offload tasks of mth Internet of things terminal equipment to O-DU, and offloading model thereofThe method comprises the following steps:
according to the state adjustment of the terminal equipment of the Internet of things; the data size of the fixed unloading of the mth Internet of things terminal equipment is +. >Power of completed transmission->
The energy-saving calculation unloading strategy after T2 optimization is as follows:
optimal unloading ratio lambda of mth Internet of things terminal equipment m The method comprises the following steps:
optimal transmitting power P of mth Internet of things terminal equipment m The method comprises the following steps:
in the formula tm For total delay time, B for total available bandwidth, D m Is the input data size of the mth Internet of things terminal equipment,to delay in distributed units DUs or O-groups after policy enforcement, T MAX For the maximum time delay of the tasks of the Internet of things equipment, mu is Lagrangian multiplier and is->Γ m and βm Is a different equation related to the signal-to-noise ratio during the salifying process; the pseudo code is shown in fig. 8.
Example 3
The implementation and simulation calculation unloading process in the IOT system based on the O-RAN are realized to display the effectiveness of the energy-saving strategy proposed by us, and the coverage radius of the IOT system based on the O-RAN is considered to be 1000m; this region deploys 3 RUs,2 DUs,1 CU, randomly distributing 30 IoTDs. For wireless channels, we set the channel loss model as g m,r =37.6×log (dist) +148.1 where dist is propagation distance. Other simulation parameters are shown in the table of fig. 9.
We compare the average energy consumption of the whole system based on different strategies. In terms of energy consumption, the energy consumption of local task execution and the energy consumption of offloading are considered. The four strategies available for comparison are an energy saving offload (ECO) strategy, a no offload (WOO) strategy, a no leave offload (WLO) strategy, and a Legacy Offload (LO) strategy, respectively.
A no offload (WOO) policy means that all tasks are computed on the internet of things device. The strategy of using Legacy Offload (LO) means that the task is partially offloaded to the O-DU taking into account the amount of data. A no-leave offload (WLO) strategy refers to completely offloading tasks onto an O-DU or O-close with minimal energy and delay considerations.
As shown in fig. 3, the offloading-based LO, WLO and ECO strategies are significantly superior to the locally-based ones in terms of average energy consumption. The average energy consumption of ECO strategies is reduced by nearly 60% with increasing number of tasks, compared to locally executed strategies (WOO strategies). Since all tasks of the WOO strategy are performed on IOTDs, the local processing power consumption is very high, while the emission power consumption is very low. Whereas for offloading-based policies, only a portion of the tasks are performed on the IOTDs, the energy consumption of local processing and transmission may be better balanced, thereby reducing overall energy consumption. Clearly, WLO strategies consume more energy and grow faster than LO and ECO strategies. This is because the WLO policy considers offloading all tasks to DUs or O-closed. All offloading minimizes the energy consumption of local processing while increasing latency. Therefore, more transmission power is required to reduce the delay, resulting in an increase in transmission energy consumption. Furthermore, the overall energy consumption of the LO strategy is also higher than the proposed ECO strategy. This is because the LO strategy is performed under the same local processing speed conditions when offloading is performed. If the number of tasks increases or the delay requirement increases, the local processing may require more energy consumption, which will result in an increase of the total energy consumption. Since the proposed ECO strategy enables flexible changes of local processing speed and offloading rate and limits offloading transmission power taking into account delay constraints, the total energy consumption of local processing and transmission can be better optimized. Therefore, the energy consumption of the ECO strategy we propose is the lowest of the four strategies, reduced by 26% and 12% compared to the WLO strategy and the LO strategy, respectively.
Fig. 4 is a comparison of transmission delays for three strategies under the same energy consumption conditions. Since the WOO policy is not offloaded, we do not need to consider the transmission delay of the WOO policy. In the case of the same number of arriving tasks, as shown in fig. 4, the delay time of the three offloading strategies gradually decreases with increasing energy consumption. This is because an increase in energy consumption may increase the local processing speed, allow more tasks to be processed locally, or increase the transmission power, allow more tasks to be offloaded, thereby reducing processing delay and transmission delay, respectively. Under the same energy consumption condition, the delay time required by the ECO strategy is the lowest in the three strategies, which are respectively lower than the LO strategy and the WLO strategy.
Here we minimize the average energy consumption by meeting the requirement of maximum delay. Thus, fig. 5 shows the average energy consumption under the maximum delay constraint at different arrival task numbers. It is apparent that the average energy consumption of the low latency constraint is higher than the high latency constraint, and the larger the number of tasks reached the larger the average energy consumption. This is because of T MAX The smaller the delay constraint, the more stringent the less IoTDs that is suitable for transferring tasks to DUs or O-closed. In order to meet the delay requirement when the task arrives, more IoTDs need to increase the local speed, and the energy consumption of increasing the transmission power and the IoTDs will increase. Furthermore, there will be more delay-sensitive IoTDs when the maximum delay constraint is smaller. For delay-sensitive IoTDs, a large amount of radio resources are allocated in the unloading process, and co-channel interference based on multiplexing frequency is more serious and energy consumption is increased. At the same time, FIG. 5 reflects the effectiveness of our ECO strategy to reduce energy consumption by adjusting the unloading ratio, local speed and transmission power.
Here we minimize the average energy consumption by meeting the requirement of maximum delay. Thus, fig. 5 shows the average energy consumption under the maximum delay constraint at different arrival task numbers. It is apparent that the average energy consumption of the low latency constraint is higher than the high latency constraint, and the larger the number of tasks reached the larger the average energy consumption. This is because of T MAX The smaller the delay constraint, the more stringent the less IoTDs that is suitable for transferring tasks to DUs or O-closed. In order to meet the delay requirement when the task arrives, more IoTDs need to increase the local speed, and the energy consumption of increasing the transmission power and the IoTDs will increase. Furthermore, there will be more delay-sensitive IoTDs when the maximum delay constraint is smaller. For delay-sensitive IoTDs, a large amount of radio resources are allocated in the unloading process, and co-channel interference based on multiplexing frequency is more serious and energy consumption is increased. At the same time, FIG. 5 reflects the effectiveness of our ECO strategy to reduce energy consumption by adjusting the unloading ratio, local speed and transmission power.
We have mentioned that the energy consumption is directly related to the local processing speed, the transmission power and the unloading ratio. Thus, fig. 7 shows the local processing speed, transmission power and unloading ratio changes during the unloading process. As the number of tasks arrives increases, the local processing speed and transmission power will gradually increase and the offloading rate will gradually decrease. This suggests that IoTDs is willing to offload more tasks to reduce energy consumption. The local speed is related to the maximum delay and unloading ratio of the task. Although the unloading rate gradually decreases, the number of task arrivals increases faster. The description tasks need to be processed on the local device, and the processing speed of the local device is still in an increasing trend, because the local processing speed is gradually increased. To increase the transmission rate to meet latency requirements, the transmission power increases as the number of offloaded tasks increases. However, due to the limitation of the maximum power, it will gradually approach 0.1w, and this value will not be exceeded. Due to task delays and local processing speed limitations, there is a need to reduce the offloading rate, which means that the proportion of processing locally gradually decreases, and more tasks tend to be offloaded to DUs or O-groups.
The invention fully utilizes the technologies of flexible O-RAN architecture, intelligent management and the like, reduces the time delay and the energy consumption of the Internet of things system, and becomes a research hotspot in the field of the Internet of things; in combination with the characteristics of an O-RAN system structure, an O-RAN-based Internet of things system is provided. The system level intelligent management is realized by standard communication through near real-time RICs and non-real-time RICs. In order to meet the lower delay and energy consumption requirements, an energy-saving calculation unloading strategy of an O-RAN-based Internet of things system is provided. In the future, we will better utilize near real-time RICs and non-real-time RICs to predict the arriving traffic model, making the energy-saving offloading strategy better adapted to the dynamic environment.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (1)
1. The energy-saving computing unloading system based on the O-RAN internet of things system comprises an O-RAN and an O-Cloud; the O-RAN is characterized in that a near real-time intelligent controller and a non-real-time intelligent controller are deployed;
The terminal equipment of the Internet of things requests the O-RAN to calculate and unload tasks;
the non-real-time intelligent controller is used for collecting non-real-time information of the O-RAN internet of things system and carrying out data reasoning analysis, wireless resource management and energy-saving calculation unloading strategy optimization based on the non-real-time information;
the near real-time intelligent controller is used for collecting real-time information of the O-RAN (Internet of things) system, carrying out decision adjustment and decision execution by combining the optimized energy-saving calculation unloading strategy, and finally sending an execution result back to the terminal equipment of the Internet of things;
the O-RAN is further configured to: service management and traffic management framework SMO, control unit CUs, distributed units DUs and radio units RUs; all wireless unit RUs contain transceiver antennas for performing physical layer operations;
the control unit CUs is used for operation of the distributed units DUs and the wireless units RUs;
the near real-time intelligent controller is deployed in the control unit CUs, and the non real-time intelligent controller is deployed in the service management and service management framework SMO;
all the terminal devices of the Internet of things send a calculation unloading task request to the distributed units DUs through the connected wireless units RUs;
the O-RAN is further configured to: an A1 interface, an O2 interface, an F1-C interface and an F1-U interface;
The non-real-time intelligent controller provides an optimized energy-saving calculation unloading strategy for the near-real-time intelligent controller through an A1 interface;
the service management and service management framework SMO manages the control unit CUs, the distributed units DUs and the radio units RUs through the O1 interface;
the service management and service management frames SMO and O-closed are connected through an O2 interface;
the distributed units DUs are connected with a user data plane CU-UP of the control unit CUs through an F1-C interface and connected with a control plane CU-CP of the control unit CUs through an F1-U interface;
m terminal devices IOTDs of the Internet of things are arranged, and M is {1, 2., M }; in the case of full frequency multiplexing, the frequency spectrums used by the radio units RUs overlap, the interference inside the radio units RUs is ignored, and all downlink links between each radio unit RU and the terminal equipment IoTDs of the internet of things have complete channel state information CSI;
the signal-to-noise ratio SINR transmitted by the wireless unit RUr to the terminal device IoTD m of the internet of things is:
wherein ,Pm The transmitting power g of the terminal equipment m of the Internet of things m,r For the channel gain, sigma, of the wireless unit RUr to the terminal equipment m of the internet of things 2 Power being additive white gaussian noise; assuming that the distribution is compliant with Gaussian distribution, according to the formula of Shannon, the channel model is the propagation rate V of the wireless unit RUr to the terminal equipment m of the Internet of things on the wireless link channel m,r :
Wherein B is the total available bandwidth and R represents the total number of wireless units RUs;
the time delay model comprises: a queuing delay model, a transmission delay model and a processing delay model; since O-closed has rich computing resources, the queuing delay of O-closed is not considered;
the queuing delay of each distributed unit DU is modeled as an M/M/1 queue of the distributed unit DU, wherein the first M represents that the arrival rate of the traffic obeys the Poisson distribution, and the second M represents that the service rate is exponentially and negatively distributed; queue delay calculation based on traffic and node capacity, queuing delay at distributed units DUsThe method comprises the following steps:
wherein ,Yd Representing the traffic arrival rate, X, of a distributed unit DUs d Representing the processing rate of the distributed units DUs;
the transmission delay model includes a transmission delay to a distributed unit DUsAnd transmission delay to O-closed
Setting the input data size of the mth IOT terminal equipment IOTD as D m An unloading ratio of 1-lambda m (0<λ m <1) The data amount locally executed is lambda m D m The amount of data offloaded is (1-lambda m )D m The method comprises the steps of carrying out a first treatment on the surface of the The front-end link transmission is transmitted by the wireless unit RUr to the distributed unit DUd, with the front-end link transmission capability being C r,d The method comprises the steps of carrying out a first treatment on the surface of the Then
wherein ,dm Representing the delay caused by the wired transmission to the O-closed;
The processing delay model comprises processing delay local to the terminal equipment of the Internet of thingsProcessing delay in distributed units DUs>And processing delay on O-closed +.>
Definition of the definition and />Respectively representing the computing power of IoTDs, the computing power of DUs and the computing power of O-group; is provided with->Then:
the energy consumption model comprises: transmission energy consumption to distributed units DUsTransmission energy consumption for transmission to O-closed>And local processing energy consumption->
wherein ,rm The energy consumption generated by occupying a channel after the data transmission is completed is realized, and epsilon is a chip factor;
defining two cost metrics to decide whether a computationally intensive task should be transferred to a distributed unit DU or O-closed; the option of offloading each device is defined as a m, wherein am =1 means that the task is offloaded to O-closed, a m =0 means that the task is offloaded to the distributed unit DUs;
the delay costs in the distributed units DU and O-closed are calculated as follows:
thus, the total cost and />Meaning that the task of setting the IOTD m by the terminal of the Internet of things is respectively unloaded to a distributed unit DU and an O-closed;
if the task of the mth Internet of things terminal equipment is unloaded to O-group, an unloading model of the mth Internet of things terminal equipment is unloaded The method comprises the following steps:
if the task of the mth Internet of things terminal device is offloaded to the distributed unit DU, an offloading model thereofThe method comprises the following steps:
the data size of the fixed unloading of the mth Internet of things terminal equipment is as followsPower of completed transmission->
Ignoring the delay of the front-end link, calculating and d m Related threshold d thr :
If d m <d thr Then a m =1; otherwise, a m =0;
The energy-saving calculation unloading strategy after optimization is as follows:
optimal unloading ratio lambda of mth Internet of things terminal equipment m The method comprises the following steps:
optimal transmitting power P of mth Internet of things terminal equipment m The method comprises the following steps:
in the formula tm For total delay time, B for total available bandwidth, D m Is the input data size of the mth Internet of things terminal equipment,to delay in distributed units DUs or O-groups after policy enforcement, T MAX For the maximum time delay of the tasks of the Internet of things equipment, mu is Lagrangian multiplier and is->Г m and βm Is a different equation related to the signal-to-noise ratio during the salifying process;
A. problem formula
According to the channel model, the time delay model, the energy consumption model and the unloading model, calculating the total energy consumption of the mth Internet of things terminal equipment task as e m A total delay time of t m :
Considering local calculation speedOffloading policy->Transmit power->And unloading ratio->Obtaining weighted sum of IoTDs energy consumption of all terminal devices of the Internet of things >The weighted sum is used as energy consumption balance among all the Internet of things equipment; consider F L A, P and λ, a combined strategy that minimizes energy consumption is obtained; the established problems are expressed as:
s.t:
wherein C1 For maximum local processing speed constraint, C 2 For non-negative transmit power constraints, C 3 To unload the ratio constraint, C 4 For maximum delay constraint, C 5 and C6 For offloading policy constraints; due to the problem P 1 Is not convex and will solve the problem P 1 The two sub-problems are divided: determining unloading points among the IoTDM terminal equipment, the distributed units O-DU and the O-group, and distributing unloading rate, transmission power and local processing speed according to the determined unloading points; performing convex approximation on the two sub-problems based on a continuous convex optimization framework, and solving the non-convex optimization problem by iteratively solving the two sub-problems;
B. problem analysis
For problem P 1 Firstly, adjusting local processing speed to obtain the optimal calculation speed of each terminal device of the Internet of things in a closed form; based on e m Along with itMonotonically increasing, minimize->And reduce the dimension of the original problem to solve for P 1 Expressed as:
so that
According toCalculate d thr Comparing the values of d thr and dm Searching for the value of A, defined as P 2 ;
P 2 :
s.t:C 4 ,C 5 ,C 6
wherein C7 From C 1 And (21) obtained, C 8 C for local processing speed constraint 9 Processing speed constraints for maximum distributed units DUs, C 10 and C11 Respectively restricting the maximum task number and the maximum transmitting power of the IOTDM terminal equipment;
P 3 :
s.t:C 2
C 12 and C13 Converted from (22) and (20); according toWill->Represented as delays in the distributed units DUs or O-groups after executing the policies, including queuing delays and processing delays;
C 13 expressed as:
according to (23), the current problem is still non-convex, and according to the non-convexity, it is divided into (24) and (25), expressed as:
non-convex:
convex:
wherein ,
h m ′(P m )+h m ′(λ m )≤0,(26)
let P be m =2 qm Then:
wherein At the kth th The individual pro-coding sequences are denoted->For h m ′(q m (k) ) Is required to satisfy the following three properties:
arbitrary convergence sequence q m (k) The limit of (2) is a Karush-Kuhn-Tucker (KKT) point;
kth th The iteration point is q m (k) Constructing h on the basis m ′(q m ) Is a proxy function of (a); on the kth salified sequence, according to (30), q is locally preserved m (k) Middle h m ′(q m ) And is strongly convex, expressed as:
equation (29) represents h m ′(q m (k) ) At any feasible point it need not be its own upper bound, then there is:
thus:
through k th The sequence is raised to obtain a proper h m ′ (q m ) Is a approximation of (a); will P 3 Conversion to P 4 Expressed as:
P 4 :
s.t:
C 14 transformed by (36), C 15 Similar to C 2 ;
Solving convex problem P using Lagrangian multiplier 4 Solving a closed expression of the closed expression by using a KKT condition;
the KKT condition is given by:
repeating the process of updating the resource allocation and the Lagrangian multiplier until the predetermined maximum number of iterations is converged or reached, i.e., the minimum value is reached;
the one-to-one correspondence between transmission power and unloading ratio is established;
the expressions (42) and (43) of the optimal unloading ratio are finally obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210741451.5A CN115134364B (en) | 2022-06-28 | 2022-06-28 | Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210741451.5A CN115134364B (en) | 2022-06-28 | 2022-06-28 | Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115134364A CN115134364A (en) | 2022-09-30 |
CN115134364B true CN115134364B (en) | 2023-06-16 |
Family
ID=83379790
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210741451.5A Active CN115134364B (en) | 2022-06-28 | 2022-06-28 | Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115134364B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108920279A (en) * | 2018-07-13 | 2018-11-30 | 哈尔滨工业大学 | A kind of mobile edge calculations task discharging method under multi-user scene |
CN113163365A (en) * | 2021-03-26 | 2021-07-23 | 北京工业大学 | Unmanned aerial vehicle support networking resource optimization method based on alternating direction multiplier algorithm |
CN113612843A (en) * | 2021-08-02 | 2021-11-05 | 吉林大学 | MEC task unloading and resource allocation method based on deep reinforcement learning |
CN114189892A (en) * | 2021-12-15 | 2022-03-15 | 北京工业大学 | Cloud-edge collaborative Internet of things system resource allocation method based on block chain and collective reinforcement learning |
WO2022121097A1 (en) * | 2020-12-07 | 2022-06-16 | 南京邮电大学 | Method for offloading computing task of mobile user |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11146455B2 (en) * | 2019-12-20 | 2021-10-12 | Intel Corporation | End-to-end quality of service in edge computing environments |
US11330511B2 (en) * | 2020-06-22 | 2022-05-10 | Verizon Patent And Licensing Inc. | Method and system for multi-access edge computing (MEC) selection and load balancing |
-
2022
- 2022-06-28 CN CN202210741451.5A patent/CN115134364B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108920279A (en) * | 2018-07-13 | 2018-11-30 | 哈尔滨工业大学 | A kind of mobile edge calculations task discharging method under multi-user scene |
WO2022121097A1 (en) * | 2020-12-07 | 2022-06-16 | 南京邮电大学 | Method for offloading computing task of mobile user |
CN113163365A (en) * | 2021-03-26 | 2021-07-23 | 北京工业大学 | Unmanned aerial vehicle support networking resource optimization method based on alternating direction multiplier algorithm |
CN113612843A (en) * | 2021-08-02 | 2021-11-05 | 吉林大学 | MEC task unloading and resource allocation method based on deep reinforcement learning |
CN114189892A (en) * | 2021-12-15 | 2022-03-15 | 北京工业大学 | Cloud-edge collaborative Internet of things system resource allocation method based on block chain and collective reinforcement learning |
Non-Patent Citations (2)
Title |
---|
移动边缘计算卸载技术综述;谢人超;廉晓飞;贾庆民;黄韬;刘韵洁;;通信学报(第11期);全文 * |
移动边缘计算系统中的卸载和计算联合优化;李霆;;电脑知识与技术(第25期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN115134364A (en) | 2022-09-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109413724B (en) | MEC-based task unloading and resource allocation scheme | |
WO2022121097A1 (en) | Method for offloading computing task of mobile user | |
CN110493360B (en) | Mobile edge computing unloading method for reducing system energy consumption under multiple servers | |
CN111930436B (en) | Random task queuing unloading optimization method based on edge calculation | |
CN111586696B (en) | Resource allocation and unloading decision method based on multi-agent architecture reinforcement learning | |
CN107919986B (en) | VM migration optimization method among MEC nodes in ultra-dense network | |
CN111132191B (en) | Method for unloading, caching and resource allocation of joint tasks of mobile edge computing server | |
CN113612843A (en) | MEC task unloading and resource allocation method based on deep reinforcement learning | |
Labidi et al. | Joint multi-user resource scheduling and computation offloading in small cell networks | |
CN110928654A (en) | Distributed online task unloading scheduling method in edge computing system | |
CN109756912B (en) | Multi-user multi-base station joint task unloading and resource allocation method | |
CN110545584A (en) | Communication processing method of full-duplex mobile edge computing communication system | |
CN112860429A (en) | Cost-efficiency optimization system and method for task unloading in mobile edge computing system | |
Wang et al. | Task allocation mechanism of power internet of things based on cooperative edge computing | |
El Haber et al. | Computational cost and energy efficient task offloading in hierarchical edge-clouds | |
CN114885420A (en) | User grouping and resource allocation method and device in NOMA-MEC system | |
CN113573363A (en) | MEC calculation unloading and resource allocation method based on deep reinforcement learning | |
Guo et al. | Dynamic computation offloading in multi-server MEC systems: An online learning approach | |
Wang et al. | Energy conserved computation offloading for O-RAN based IoT systems | |
Li et al. | A trade-off task-offloading scheme in multi-user multi-task mobile edge computing | |
CN116828534B (en) | Intensive network large-scale terminal access and resource allocation method based on reinforcement learning | |
Mu et al. | Latency constrained partial offloading and subcarrier allocations in small cell networks | |
Chen et al. | Dynamic task caching and computation offloading for mobile edge computing | |
Chen et al. | Joint optimization of task caching, computation offloading and resource allocation for mobile edge computing | |
CN115134364B (en) | Energy-saving computing and unloading system and method based on O-RAN (O-radio Access network) Internet of things 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 |