CN115278779A - Rendering perception-based dynamic placement method for VR service module in MEC network - Google Patents
Rendering perception-based dynamic placement method for VR service module in MEC network Download PDFInfo
- Publication number
- CN115278779A CN115278779A CN202210900105.7A CN202210900105A CN115278779A CN 115278779 A CN115278779 A CN 115278779A CN 202210900105 A CN202210900105 A CN 202210900105A CN 115278779 A CN115278779 A CN 115278779A
- Authority
- CN
- China
- Prior art keywords
- user
- cost
- time slot
- delay
- mec server
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0925—Management thereof using policies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
- H04W28/0967—Quality of Service [QoS] parameters
- H04W28/0975—Quality of Service [QoS] parameters for reducing delays
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/08—Load balancing or load distribution
- H04W28/09—Management thereof
- H04W28/0958—Management thereof based on metrics or performance parameters
- H04W28/0967—Quality of Service [QoS] parameters
- H04W28/0983—Quality of Service [QoS] parameters for optimizing bandwidth or throughput
-
- 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
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a rendering perception-based dynamic placement method for VR service modules in an MEC network, belonging to the field of rendering of VR games; the method specifically comprises the following steps: firstly, building a cellular network scene comprising U user players, M base stations and H VR game service modules; respectively calculating placement cost, migration cost, rendering cost and communication cost aiming at the current time slot t; constructing a target function and a constraint condition based on the cost of each time slot, and enabling the total cost of the whole network to be minimum under the condition of meeting the delay constraint of each user; finally, the objective function is decomposed into two sub-problems: dynamic access and service module placement problems and quasi-static resource allocation problems; and respectively solving by using a minimum cut theory and a convex optimization method to obtain an optimal user access scheme, an optimal VR service module placement scheme and an optimal spectrum resource distribution scheme. The invention realizes a better balance between the route delay cost of the rendered VR video stream and the migration cost of the corresponding VR service module.
Description
Technical Field
The invention belongs to the field of rendering of VR games, and particularly relates to a rendering perception-based dynamic placement method for VR service modules in an MEC network.
Background
In recent years, combining multiple access edge computing (MEC) technology with wireless Virtual Reality (VR) gaming is a promising computing paradigm. The defect that the computing resources of the mobile equipment are insufficient can be overcome by unloading the rendering task to the edge node for rendering. However, to execute a rendering task on an MEC server, the MEC server needs to have a VR service module corresponding to the rendering task deployed thereon. Obviously, it is not practical for an edge server with limited storage capacity to deploy all VR service modules in the network scenario at the same time.
The prior art studies on VR video are as follows:
document [1] proposes a task offloading scheme supporting a block chain to resist malicious attacks, reduce the computational load of a virtual machine, and meet the high QoE of VR users;
document [2] proposes to minimize the long-term energy consumption of MEC systems based on terahertz wireless access by jointly optimizing viewport rendering offload and downlink transmission power control to support high-quality immersive VR video services.
Document [3] proposes a wireless VR network supporting MECs. The network predicts the field of view of each VR user in real time using a recurrent neural network and offloads the VR rendering tasks from the VR devices to the MEC server through a rendering model migration function.
Document [4] considers the provision of multi-tile wireless VR video services over MEC networks; the goal is to minimize system delay and energy consumption. First, the study converts the time-varying view popularity into a model-free markov chain to effectively capture its dynamics. Then, a hybrid strategy is adopted to coordinate dynamic cache replacement and deterministic offloading so as to fully utilize system resources.
However, the article reduces the end-to-end delay of the user by optimizing the offloading policy of the VR rendering task, and improves the QoE of the user. However, in a network environment with limited memory space of the MEC server, little attention is paid to the placement of various types of VR service modules to support corresponding rendering tasks, and the influence of user mobility on VR service module placement strategies. In addition, the placement strategy of the VR service module has little concern on the routing delay and the network resource consumption cost of the rendered VR video stream.
Edge rendering inevitably introduces edge computation delay and transmission delay caused by the return of rendered VR video stream to the mobile terminal; in particular, the increase in delay is more pronounced since the data volume of VR video streams is generally huge. Therefore, it is important to optimize the transmission path of the rendered VR video stream and to reasonably allocate edge resources including radio spectrum and computation.
It should also be noted that deploying service modules on MEC servers increases placement costs and is limited by storage capacity, and service modules for various types of VR games cannot be deployed on each MEC at the same time. However, a premise for performing rendering tasks on the MEC server is that the VR game service module that the user participates in is already deployed on the MEC server. Therefore, it is important to tightly couple service module layout optimization with computing resource allocation to optimize the delivery performance of wireless VR game video streams.
In addition, in a network scenario where multiple wireless VR games are concurrent, the geographic location of the user changes as the user moves, and accordingly, the access base station may also change accordingly. To ensure low routing cost for rendered VR video streams, the VR service module corresponding thereto may need to be migrated to a new base station. Such migration will undoubtedly increase migration costs, including hardware loss costs and data migration delay costs. Therefore, it becomes important to dynamically jointly optimize the routing overhead and the migration overhead.
[1]P.Lin,Q.Song,F.R.Yu,D.Wang and L.Guo,"Task Offloading for Wireless VR-Enabled Medical Treatment With Blockchain Security Using Collective Reinforcement Learning,"in IEEE Internet of Things Journal,vol.8,no.21,pp.15749-15761,1Nov.1,2021,doi:10.1109/JIOT.2021.3051419.
[2]J.Du,F.R.Yu,G.Lu,J.Wang,J.Jiang and X.Chu,"MEC-Assisted Immersive VR Video Streaming Over Terahertz Wireless Networks:A Deep Reinforcement Learning Approach,"in IEEE Internet of Things Journal,vol.7,no.10,pp.9517-9529,Oct.2020,doi:10.1109/JIOT.2020.3003449.
[3]X.Liu and Y.Deng,"A Decoupled Learning Strategy for MEC-enabled Wireless Virtual Reality(VR)Network,"2021IEEE International Conference on Communications Workshops(ICC Workshops),2021,pp.1-6,doi:10.1109/ICCWorkshops50388.2021.9473847.
[4]C.Zheng,S.Liu,Y.Huang and L.Yang,"Hybrid Policy Learning for Energy-Latency Tradeoff in MEC-Assisted VR Video Service,"in IEEE Transactions on Vehicular Technology,vol.70,no.9,pp.9006-9021,Sept.2021,doi:10.1109/TVT.2021.3099129.
Disclosure of Invention
Aiming at the problems, the invention provides a rendering perception-based dynamic placement method of a VR service module in an MEC network, which tightly combines the unloading of the rendering task of a VR game with the placement of the service module; meanwhile, in order to further reduce the end-to-end delay of VR video delivery, the routing delay of VR video stream and the migration cost of service module are optimized in a combined manner; the goal is to minimize the sum of network long time overheads while satisfying the tolerable delay constraints of each game user.
The rendering perception-based VR service module dynamic placement method in the MEC network comprises the following specific steps:
step one, building a cellular network scene comprising U user players, M base stations and H VR games.
The user set is represented asEach base station is respectively provided with an MEC server, and the base station set is represented asThe base stations are connected through wires;
there are H types of VR games, the set beingWhile H different service modules are required to support the corresponding VR games.
Step two, aiming at the current time slot t, respectively calculating the placement cost, the migration cost, the rendering cost and the communication cost of the time slot;
a) The placement cost calculation formula is as follows:
λmhthe storage resource occupation cost required for placing the service module of the VR game h on the MEC server m is represented;indicating whether the MEC server m stores an index quantity of a service module of the VR game h at the time slot t; if so, thenIf not, then,
b) The migration cost calculation formula is as follows:
wherein the content of the first and second substances,fhrepresenting the cost of reconfiguration of the service modules for VR game h.
ghRepresenting the migration delay of a service module of the VR game h, wherein upsilon represents a regulation parameter, and the reconfiguration cost is enabled by regulating the value of upsilonAnd migration delayThe magnitude of (a) is in the same order of magnitude.
c) The rendering cost calculation formula is as follows:
an index amount indicating whether VR game h selects to perform rendering on MEC server m, and if so,if not, then the mobile terminal can be switched to the normal mode, representing the computing resources required by VR game h at time slot t;representing the computing resources allocated to the VR game h by the MEC server m at the time slot t;
d) The communication cost calculation formula is as follows:
the sum of downlink transmission time delays of all users in a time slot t is represented, namely the network downlink communication cost;
representing the downlink transmission delay of a user u at a time slot t; the calculation formula is as follows: an index quantity indicating whether the user u accesses the MEC server m at the time slot t, and if so,if not, then, indicating the size of the video stream to be transmitted to user u at time slot t;represents the uplink transmission rate between the user u and the MEC server m at the time slot t;
representing the routing delay of all users at the time slot t, namely the routing overhead of the network; the calculation formula is as follows: representing the time delay of the rendered video stream to the user u to access the MEC server at the time slot t; the formula is as follows:
an index amount indicating whether the user u joins the VR game h; if the number of the data packets is more than the preset value,if not, then, an index quantity indicating whether the user u accesses the MEC server m' at the time slot t, and if so,if not, then,d (m, m ') represents a routing delay between the MEC server m to the MEC server m'.
And step three, constructing an objective function and constraint conditions based on the placement cost, the migration cost, the rendering cost and the communication cost of each time slot, and enabling the total cost of the whole network to be minimum under the delay constraint of each user.
The objective function is:
C6:Du',t≤Du
ε1,ε2,ε3and ε4For the weight coefficients, the communication costs are respectively correspondedCost of renderingCost of placementAnd migration costThe proportion in the objective function.
Constraint C1Ensuring that a user cannot connect to multiple MEC servers simultaneously, while each user can connect to one MEC server;
constraint C2Ensuring that one VR game can be executed in one time slot and only one MEC server can be selected to execute the rendering task;
constraint C3Ensuring that the total bandwidth allocated to the access user by the MEC server m does not exceed the whole bandwidth of the wireless access network in each time slot;a spectrum bandwidth representing the time slot t allocated to the user u from the MEC server m; b istRepresents the entire bandwidth of the radio access network;
constraint C4Ensuring that the computing resources allocated by the MEC server to the VR games it serves cannot exceed its maximum computing resources; k ismIs the maximum computing power of MEC server m.
Constraint C5It is ensured that the service module size of the VR game stored in the MEC server m does not exceed the maximum storage capacity of the MEC server m. w is ahData size of service module for VR Game h
Constraint C6Ensure that the total delay of each VR game cannot exceed its maximum tolerated delay. DuThe maximum tolerated delay for VR game h. From the above description, the actual end-to-end delay of user u over time period t can be expressed as:
whereinIndicating the downlink transmission delay of the rendered video stream transmitted by the access base station of the user u to the user u at the time slot t,indicating the routing delay required for the rendered video stream to be routed from the rendering base station to the access base station of user u at time slot t,representing the rendering delay required for rendering the video stream requested by user u at time slot t,indicating migration of VR service module required by user u to corresponding rendering at time slot tThe migration delay required by the base station is dyed.
And step four, decomposing the objective function into two subproblems on the premise of meeting the constraint condition: dynamic access and service module placement problems and quasi-static resource allocation problems; and respectively solving by using a minimum cut theory and a convex optimization method to obtain an optimal user access scheme, an optimal VR service module placement scheme and an optimal spectrum resource distribution scheme.
The invention has the advantages that:
1. the rendering perception-based dynamic placement method for the VR service module in the MEC network reduces the overhead of the whole system in the VR game video stream delivery process, including the consumption cost and the time delay cost of various network resources. Simulation results show that the dynamic placement method of the rendering-aware service module can flexibly select the user access base station and the rendering task execution base station according to the real-time position of the user, and finally the cost of the whole system is minimized in a long and continuous time.
2. The rendering perception-based dynamic placement method for the VR service module in the MEC network can meet the requirements of low time delay of VR game service and real-time movement of users, so that the computing capacity of terminal equipment and the cache capacity of an edge server are no longer barriers for providing low-time-delay wireless VR service.
3. The rendering perception-based dynamic placement method for the VR service module in the MEC network realizes a better balance between the route delay cost of the rendered VR video stream and the migration cost of the corresponding VR service module.
Drawings
FIG. 1 is a flowchart of a rendering awareness-based dynamic placement method for VR service modules in an MEC network according to the present invention;
FIG. 2 is a schematic diagram of a cellular network scenario constructed in accordance with the present invention, including a player, a base station, and a VR gaming service module;
FIG. 3 is a graph comparing the effect of the total network overhead of the method and the particle swarm algorithm with the iteration time;
FIG. 4 is a graph comparing the relationship between the computing power of the method and the particle swarm algorithm MEC server and the average delay of the user;
FIG. 5 is a comparison graph of the relationship between the computing power and the total network overhead of the MEC server of the method and the particle swarm algorithm;
FIG. 6 is a graph comparing the effect of the delay constraints of the method and the particle swarm optimization on the total network overhead and the average user delay;
FIG. 7 is a comparison graph of the actual time delay of the user under the condition of no time delay constraint and tolerable time delay of the method and the particle swarm optimization.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
To meet the low latency requirements of wireless VR gaming services in a multidimensional (storage-computation-communication) constrained MEC network. The invention discloses a rendering perception-based dynamic placement method for VR service modules in an MEC network, which is used for unloading rendering tasks of VR games to an MEC server and tightly combining unloading decisions with service module placement. Meanwhile, in order to further reduce the end-to-end delay of VR video delivery, in the placement method, the routing delay of VR video stream and the migration cost of the service module are optimized in a combined manner; specifically, the scheme optimizes the cost of migrating VR service modules between different base stations between adjacent time slots while considering the reasonable allocation of bandwidth, calculation and storage resources in each time slot. The goal is to minimize the sum of the network long-time spending under the condition of satisfying the tolerable delay constraint of each game user, namely the sum of the total delay of all users and all consumed network resources in a period of time; for this purpose, the invention is modeled as a high-order, non-convex and time-varying function, and the placement problem is converted into the minimum cutting problem by constructing a series of auxiliary graphs. Finally, a two-stage iterative algorithm based on convex optimization and graph theory is provided to solve the objective function. A large number of simulation results show that compared with other baseline algorithms, the method can ensure lower end-to-end delay and lower network cost.
As shown in fig. 1, the rendering-aware-based VR service module dynamic placement method in the MEC network specifically includes the following steps:
step one, building a cellular network scene comprising U user players, M base stations and H VR game service modules.
The MEC server is a micro data center, typically deployed with a cellular base station or WiFi access point. Hardware resources in the MEC server are virtualized by adopting some lightweight virtualization technologies, and flexible sharing of the resources is achieved.
As shown in fig. 2, in a cellular network with MEC servers deployed, there are U players and M Base Stations (BSs), each deployed with one MEC server. The user set is represented asEach base station is respectively provided with an MEC server, and the base station set is represented asThe base stations are connected through wires; there are H types of VR games in the scene, the set isTherefore, H different VR service modules are needed to support VR games.
In addition, to make dynamic decisions, the problem is modeled as a slotted system and usedTo represent a contiguous set of time slots, each of which is assumed to be much larger than the delay incurred by transmission and processing.
Step two, aiming at the current time slot t, respectively calculating the placement cost, the migration cost, the rendering cost and the communication cost of the time slot;
a) Cost of placement
Assume that VR Game service Module places a set for policyTo indicate. When in useThen, the MEC server m stores a service module of the VR game h in the time slot t; otherwiseλ is used for storage resource occupation cost required for placing service module of VR game h on MEC server mmhRepresents; the total placement cost of the VR game service module is expressed by the following equation:
suppose that the storage capacity of the server m is pimThe size of the service module of the VR game h is omegahSince the total size of the VR game service modules deployed at server m should not exceed the maximum storage capacity of that server, the constraint on placement cost should be expressed as:
Πmrepresenting the storage capacity, ω, of the MEC server mhRepresents the size of a service module of the VR game h;
b) And migration cost
As the user moves, the base station transmitting the rendered data to the user may change due to changes in geographic location. At the same time, a server that originally provided rendering services for a group of users may no longer be the best choice to provide rendering services. Thus, the system may need to select a new suitable base station for the group of users to render, and may even need to re-deploy the corresponding VR game service module on the newly selected BS. That is, the data information of the traffic module may need to be migrated from the old MEC server to the new MEC server and the environment built on the new MEC. However, migration of VR game service modules incurs hardware loss costs and data migration delay costs.
The migration delay for each user belonging to the same group is equal and can be expressed as:
an index value indicating whether the user u joins the service module of the VR game h; if so, the mobile terminal can be started,if not, then, a delay representing migration of a service module of the VR game h, which was not stored in the MEC server m at the previous time slot, into the MEC server m at the time slot t;
in addition, all migration costs are expressed as:
wherein the content of the first and second substances,fhrepresenting the cost of reconfiguration of the service modules for VR game h.
ghRepresenting the migration delay of a service module of the VR game h, wherein upsilon represents a regulation parameter, and the reconfiguration cost is enabled by regulating the value of upsilonAnd migration delayThe magnitude of (a) is in the same order of magnitude.
c) Rendering cost
There may be overlapping computing tasks for the same group of users, and the computing resources on each server are allocated in groups, assuming that the MEC server performs centralized computing after collecting all the information for the same group of players. In an MEC network, when an MEC server serves only one group, the group may have more computing resources available to perform rendering, resulting in a lower processing latency experience.
However, each MEC server typically needs to serve multiple game groups simultaneously, which may result in competition for computing resources. In particular, if one MEC server provides rendering services for too many game groups, the rendering delay for all game groups performing rendering tasks at that server will increase significantly. By indicating quantitiesTo indicate whether user u joins VR game h. Since one participant can only participate in one game, the corresponding constraint is expressed as:
by usingTo represent a set of base station selection strategies being presented. When VR game h selects MEC server m to perform a rendering task,otherwiseIn addition, in order to ensure information synchronization between users in the same group, assuming that a group of users can only select one MEC server to process tasks at one time point, the constraint condition can be expressed as:
because the cost of placing the VR game service module on the server is high, the invention places the VR game service module on the base station for processing the rendering task of the game group, and obtains the following formula:
assume that the maximum computing power of MEC server m is KmThe computing resources allocated to VR game h by server m at time slot t areBy usingTo represent an allocation scheme for computing resources.Indicating the computational resources required by VR game h at time slot t. The rendering delays for the same set of players are equal. Therefore, the rendering delay of user u at time slot t is represented as:
the rendering cost may be represented by the sum of the rendering delays of all game groups:
meanwhile, the sum of the computing resources allocated by the MEC server to all game groups it serves cannot exceed its maximum computing resource. Thus, the corresponding computational resource constraint can be expressed as:
d) Cost of communication
The invention provides a communication model in a mobile edge computing network based on mmWave, which is mainly focused on downlink transmission. Meanwhile, route transmission delay is introduced. By usingTo indicate access policies and, furthermore, users cannot connect to multiple base stations simultaneously, there is a need to ensure that each player can connect to an appropriate base station. Thus, the following constraint equation is obtained:
the time slot t is allocated to the user u from the MEC server m with the spectrum bandwidth ofIn combination withAs a bandwidth allocation scheme. Since the total bandwidth allocated to its access subscribers by MEC server m does not exceed the entire bandwidth of the radio access network at time slot t, i.e. BtThe corresponding bandwidth constraint is expressed as:
then the uplink transmission rate between user u and MEC server m at time slot t is:
suppose that the time slot t requires the size of the video stream transmitted to user u to beThe downlink transmission delay is:
an index quantity indicating whether the user u accesses the MEC server m at the time slot t, and if so,if not, then,the sum of the downlink transmission delays of all users in the time slot t, that is, the network downlink communication cost, is:
and dividing the VR games into H groups according to the difference of the VR games participated by the user, wherein different game groups need different service modules for rendering. An appropriate MEC server needs to be selected to render for VR game h and then the rendered video stream is quickly routed to the visiting base station of user u participating in VR game h. Wherein, the selected MEC server needs to deploy a corresponding VR service module and has enough computing resources to execute the rendering task. According to the above assumption, at time slot t, the time delay of the rendered video stream to reach the MEC server of user u is represented as:
whereinAn index quantity indicating whether the user u accesses the MEC server m' at the time slot t, and if so,if not, then,d (m, m ') is the routing delay between the MEC server m to the base station MEC server m'. The routing delay of all users in time slot t, i.e. the routing overhead of the network, is:
therefore, the communication overhead of the time slot t can be expressed as the sum of the downlink transmission delay and the routing delay.
And step three, constructing an objective function and constraint conditions based on the placement cost, the migration cost, the rendering cost and the communication cost of each time slot, and enabling the total cost of the whole network to be minimum under the delay constraint of each user.
The goal of rendering aware based service module dynamic placement strategy is to minimize the total cost of the entire network over a long period of time while satisfying the delay constraints of each user. The strategy simultaneously considers a resource allocation scheme in each time slot and an inter-base station service module migration scheme between adjacent time slots.
The objective function is:
C6:Du',t≤Du
ε1,ε2,ε3and ε4For the weight coefficients, the communication costs are respectively correspondedCost of renderingCost of placementAnd migration costThe proportion in the objective function.
Constraint C1Ensuring that a user cannot connect to multiple MEC servers simultaneously, while each user can connect to one MEC server;
constraint C2Ensuring that one VR game can be executed in one time slot and only one MEC server can be selected to execute the rendering task;
constraint C3Securing MEC serversm total bandwidth allocated to its access users, not exceeding the entire bandwidth of the wireless access network in each time slot;a spectrum bandwidth representing the time slot t allocated to the user u from the MEC server m; b istRepresents the entire bandwidth of the radio access network;
constraint C4Ensuring that the computing resources allocated by the MEC server to the VR games it serves cannot exceed its maximum computing resources; k ismIs the maximum computing power of MEC server m.
Constraint C5It is ensured that the VR game service module size stored in the MEC server m does not exceed the maximum storage capacity of the MEC server m. w is ahIs the data size of the service module of VR game h.
Constraint C6Ensure that the total delay of each VR game cannot exceed its maximum tolerated delay. DuThe maximum tolerated delay for VR game h. From the above description, the actual end-to-end delay of user u in slot t can be expressed as:
whereinIndicating the downlink transmission delay of the rendered video stream transmitted by the access base station of the user u to the user u at the time slot t,indicating the routing delay required for the rendered video stream to be routed from the rendering base station to the access base station of user u at time slot t,representing the rendering delay required for rendering the video stream requested by user u at time slot t,and the migration time delay required for migrating the VR service module required by the user u to the corresponding rendering base station in the time slot t is represented.
And step four, decomposing the objective function into two subproblems on the premise of meeting the constraint condition: dynamic access and service module placement problems and quasi-static resource allocation problems; and respectively solving by using a minimum cut theory and a convex optimization method to obtain an optimal user access scheme, an optimal VR service module placement scheme and an optimal spectrum resource distribution scheme.
The method comprises the following main steps:
1) Establishing an initial user access model, a VR service module placement model and a resource distribution model;
2) Constructing an auxiliary graph and giving each edge in the graph an appropriate weight. And determining whether to replace the access base stations of some users and the placing base stations of some VR service modules by the base station alpha by determining the minimum cut set of the auxiliary graph.
3) On a given access and placement scheme, the computational and spectral resources at each base station are allocated rationally to optimize the overhead per slot and further optimize the network overhead for the entire system.
The two subproblems are respectively expressed by the following formulas:
wherein, the first and the second end of the pipe are connected with each other,represents a base station providing rendering service for VR game h at time slot t whenWhen the temperature of the water is higher than the set temperature,in the same way, the method for preparing the composite material,represents a base station providing rendering service for VR game h at time slot t-1 whenWhen the temperature of the water is higher than the set temperature, indicates that user u accesses the base station in time slot t whenWhen the temperature of the water is higher than the set temperature,
and the number of the first and second electrodes,indicating that at time slot t, the base stationThe size of the bandwidth resource allocated to user u;indicating that at time slot t, the base stationThe size of the computing resources allocated to VR game h.Indicates that in time slot t, user u and its access base stationThe signal to noise ratio therebetween.Indicating the computational resources required by VR game h at time slot t.
(1) Aiming at the first subproblem, optimizing the dynamic access and service module dynamic placement problem through a minimal cut theory;
firstly, introducing an alpha expansion algorithm; then, an auxiliary graph is constructed, and each edge in the graph is given a proper weight to optimize the objective function. The minimum value of the objective function is equal to the sum of the weights of all edges in the minimum cut set of the auxiliary graph;
the alpha extension is defined as binary optimization and reflects the trend of the service module moving from the current base station to the base station alpha and the trend of the user accessing the current base station to the base station alpha. For example when base station alpha is selected as the extension of the access base station.There is a binary choice, when user u continues to maintain access to the current base station,when user u changes the access base station to base station alpha,for convenience of calculation, the expanded result can also be an index vector with binary decision variablesAnd (4) showing. When the temperature is higher than the set temperatureWhen the temperature of the water is higher than the set temperature,otherwiseWhen in useTime of flightOtherwiseAfter base station alpha spreading, using binary variableReconstructed CostIIs composed ofConverting a sub-problem into a binary optimization problem with a variable of 0 or 1At the same time, define
in the auxiliary graph, there are T points and U points corresponding to the player as the creditGame group notation with T x H verticesA source node S and a destination node T. Edges are added to the graph and each edge is given an appropriate weight.
Through the assignment mode, the minimum value of the target function is equal to the sum of the weights of all edges in the minimum cut set of the auxiliary graph.
Dividing the vertexes in the graph into two sets by using a minimal cut algorithm, assigning the vertexes which are positioned in the same set with the source node S as 0, and assigning the vertexes which are positioned in the same set with the terminal node T as 1, thereby determining the index variableAndcan determine the iteration process by back-steppingAndthe value of (a). Through multiple iterations, a user access scheme and a VR service module placement scheme can be finally determined.
(2) Resource allocation scheme based on convex optimization
Cost is to be paid attention to for the second sub-problemIIBy properly allocating computational and spectral resources, the total transmission and editing delay in each time interval is minimized.
wherein Λ is a penalty function, and the optimization problem is a convex problem. The ADMM algorithm and the CVX tool box are used for solving the problems to obtain a distribution scheme of computing resources and spectrum resources, namelyAndthe value of (a).
(3) Two-stage iterative algorithm based on alpha expansion
Because the original problem has more optimization variables, the algorithm complexity is higher. In order to reduce the complexity of the algorithm and obtain the optimal solution of the original problem, the original problem is solved in two steps. And obtaining an optimal user access scheme and a VR service module placement scheme after one round of alpha expansion according to the algorithm 1. Substituting the results of the two optimal schemes into an algorithm 2 as fixed values, and solving the allocation scheme of the calculation sum spectrum resources in each time slot according to the algorithm 2. And calculating an objective function through a user access scheme obtained by the iteration of the round, a VR service module placement scheme, and a calculation and spectrum resource allocation scheme. And comparing the objective function value calculated in the current round with the historical optimal result, and keeping the smaller value of the objective function value and the historical optimal result and the corresponding problem solution. And carrying out the next iteration and repeating the process.
The invention considers that the buffer capacity of the edge base station is limited, when the user moves, new access base stations and working base stations for executing rendering tasks are allocated to the current user again according to the current environment, such as channel state, the size of calculation tasks unloaded by each user, the current placement position of a VR service module and the like, and meanwhile, required working data are transferred from the original working base station to the new working base stations so as to meet the requirement of low time delay of the wireless VR game service.
The present invention uses MATLAB for simulation. Consider a 100m x 100m cellular network scenario with 100 users and 10 base stations. The data used for the simulation are shown in table 1 below.
TABLE 1
Total bandwidth | 1GHz |
Kind of |
40 |
Coefficient of downlink path loss | [2.75-4.75] |
Video stream size | [5-20]Mb |
Server storage space | [400-900]G |
Server computing power | [30-80]GHz |
As shown in fig. 3, the minimum value of the total network overhead is found through iteration under the condition that the maximum computing capacity of each MEC server is 60ghz and each server can place data with the size of 600G at most. The total network overhead of the method and the particle swarm algorithm is rapidly reduced along with the increase of the iteration times at the beginning, and then the total network overhead is converged and kept at an almost constant value. As can be seen from the iteration chart, the convergence time of the method is about 18 generations, and the convergence time of the particle swarm algorithm is about 25 generations. Therefore, compared with other schemes, the method converges faster in the iterative process and keeps the total network overhead at the lowest.
As shown in fig. 4, which is the relationship between the computing power of the MEC server and the average delay of the users; as shown in fig. 5, is the relationship between the computing power of the MEC server and the total network overhead. In the above two figures, as the computing power of the MEC server increases, the average delay of the user and the total network overhead are greatly reduced. This is mainly because the more computing resources an MEC server can provide to a user, the less delay is required to perform rendering. Meanwhile, the richer the computing resources on the MEC server, the more MEC servers are available for the system to select to provide rendering services for each game group, thus saving the network cost of routing to some extent.
As shown in fig. 6, the average delay and total network overhead of users without delay constraint are compared with the average delay and total network overhead of users with delay constraint. The network parameters are that the maximum computing capacity of each MEC server is 60GHz, and the storage capacity is 600g. When the delay constraint of each user is not required to be considered, the feasible domain of the target problem becomes large, and compared with the time delay constraint, the total network overhead is reduced, but the average delay of the users is increased. At the same time, some users cannot complete the corresponding video processing task within a tolerable delay, as shown in fig. 7.
Claims (4)
- A rendering perception-based VR service module dynamic placement method in an MEC network is characterized by comprising the following specific steps:firstly, building a cellular network scene comprising U user players, M base stations and H VR game service modules;the set of U users is represented as:the set of M base stations is represented as:there are H types of VR games, the set beingH different VR service modules are needed to support corresponding VR games;then, aiming at the current time slot t, respectively calculating the placement cost, the migration cost, the rendering cost and the communication cost of the time slot;Establishing a target function and a constraint condition based on the placement cost, the migration cost, the rendering cost and the communication cost of each time slot, and enabling the total cost of the whole network to be minimum under the delay constraint of each user;the objective function is:C6:Du',t≤Duit is meant that the access policy,as a scheme for the allocation of the bandwidth,representing an allocation scheme of the computing resources,indicating the presentation of a set of base station selection strategies,representing a VR gaming service module placement strategy;ε1,ε2,ε3and ε4For the weight coefficients, the communication costs are respectively correspondedCost of renderingCost of placementAnd migration costThe proportion in the objective function;constraint C1Ensuring that a user cannot connect to multiple MEC servers simultaneously, while each user can connect to one MEC server;an index quantity indicating whether the user u accesses the MEC server m at the time slot t, and if so,if not, then the mobile terminal can be switched to the normal mode,constraint C2Ensuring that one VR game can be executed in one time slot and only one MEC server can be selected to execute the rendering task;an index amount indicating whether the service module of VR game h selects MEC server m, and if so,if not, then,constraint C3Ensuring that the total bandwidth allocated to its access user by the MEC server m does not exceed the entire bandwidth of the radio access network in each timeslot;a spectrum bandwidth representing the time slot t allocated to the user u from the MEC server m; b istRepresents the entire bandwidth of the radio access network;constraint C4Ensuring that the computing resources allocated by the MEC server to the VR games it serves cannot exceed its maximum computing resources;representing the computing resources allocated to the VR game h by the MEC server m at the time slot t; kmIs the maximum computing power of MEC server m;constraint C5Ensuring that the size of VR game service module stored in MEC server m does not exceed the maximum storage capacity Π of MEC server mm;Indicating whether the MEC server m stores an index quantity of a service module of the VR game h at the time slot t; if so, thenIf not, then,whthe data size of a service module for the VR game h;constraint C6Ensuring that the total delay of each VR game cannot exceed its maximum tolerated delay; d'u,tActual end-to-end delay of user u at time period t; duMaximum tolerated delay for VR game h;and finally, decomposing the objective function into two subproblems on the premise of meeting the constraint condition: dynamic access and service module placement problems and quasi-static resource allocation problems; and respectively solving by using a minimum cut theory and a convex optimization method to obtain an optimal user access scheme, an optimal VR service module placement scheme and an optimal spectrum resource distribution scheme.
- 2. The method of claim 1, wherein in the network scenario, each base station is respectively deployed with an MEC server, and the base stations are connected by a wire.
- 3. The method of claim 1, wherein the placement cost calculation formula is:λmhthe storage resource occupation cost required for placing the service module of the VR game h on the MEC server m is represented;the migration cost calculation formula is as follows:wherein the content of the first and second substances,fhrepresenting a cost of reconfiguration of the service modules of the VR game h;ghrepresenting the migration delay of a service module of the VR game h, wherein upsilon represents a regulation parameter, and the reconfiguration cost is enabled by regulating the value of upsilonAnd migration delayThe numerical value of (A) is in the same order of magnitude;the rendering cost calculation formula is as follows:the communication cost calculation formula is as follows:the sum of the downlink transmission time delay of all users in the time slot t is represented, namely the network downlink communication cost;representing the downlink transmission delay of the user u at the time slot t; the calculation formula is as follows: indicating the size of the video stream to be transmitted to user u at time slot t;represents the uplink transmission rate between the user u and the MEC server m at the time slot t;representing the routing delay of all users at the time slot t, namely the routing overhead of the network; the calculation formula is as follows:representing the time delay of the rendered video stream to the user u to access the MEC server at the time slot t; the formula is as follows:indicates that user u isWhether the index amount of the VR game h is added or not; if so, the mobile terminal can be started,if not, then, an index quantity indicating whether the user u accesses the MEC server m' at the time slot t, and if so,if not, then,d (m, m ') represents a routing delay between the MEC server m to the MEC server m'.
- 4. The method of claim 1, wherein the actual end-to-end delay D of the user u at time period tu',tExpressed as:whereinIndicating the downlink transmission delay of the rendered video stream transmitted by the access base station of the user u to the user u at the time slot t,indicating the routing delay required for the rendered video stream to be routed from the rendering base station to the access base station of user u at time slot t,representing the rendering delay required for rendering the video stream requested by user u at time slot t,and the migration time delay required for migrating the VR service module required by the user u to the corresponding rendering base station in the time slot t is represented.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210900105.7A CN115278779A (en) | 2022-07-28 | 2022-07-28 | Rendering perception-based dynamic placement method for VR service module in MEC network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210900105.7A CN115278779A (en) | 2022-07-28 | 2022-07-28 | Rendering perception-based dynamic placement method for VR service module in MEC network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115278779A true CN115278779A (en) | 2022-11-01 |
Family
ID=83770802
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210900105.7A Pending CN115278779A (en) | 2022-07-28 | 2022-07-28 | Rendering perception-based dynamic placement method for VR service module in MEC network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115278779A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117689096A (en) * | 2024-01-25 | 2024-03-12 | 武汉科技大学 | Mobile charging scheduling method with obstacle avoidance function |
CN117689096B (en) * | 2024-01-25 | 2024-04-19 | 武汉科技大学 | Mobile charging scheduling method with obstacle avoidance function |
-
2022
- 2022-07-28 CN CN202210900105.7A patent/CN115278779A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117689096A (en) * | 2024-01-25 | 2024-03-12 | 武汉科技大学 | Mobile charging scheduling method with obstacle avoidance function |
CN117689096B (en) * | 2024-01-25 | 2024-04-19 | 武汉科技大学 | Mobile charging scheduling method with obstacle avoidance function |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109413724B (en) | MEC-based task unloading and resource allocation scheme | |
CN109684075B (en) | Method for unloading computing tasks based on edge computing and cloud computing cooperation | |
CN107995660B (en) | Joint task scheduling and resource allocation method supporting D2D-edge server unloading | |
WO2022121097A1 (en) | Method for offloading computing task of mobile user | |
CN107919986B (en) | VM migration optimization method among MEC nodes in ultra-dense network | |
CN109547555B (en) | Non-equilibrium edge cloud network access and resource allocation method based on fairness criterion | |
CN112020103B (en) | Content cache deployment method in mobile edge cloud | |
CN110098969B (en) | Fog computing task unloading method for Internet of things | |
CN109194763B (en) | Caching method based on small base station self-organizing cooperation in ultra-dense network | |
CN109548031B (en) | Unbalanced edge cloud network access and resource allocation method | |
WO2023024219A1 (en) | Joint optimization method and system for delay and spectrum occupancy in cloud-edge collaborative network | |
CN111800812B (en) | Design method of user access scheme applied to mobile edge computing network of non-orthogonal multiple access | |
Li et al. | Distributed task offloading strategy to low load base stations in mobile edge computing environment | |
CN112512065B (en) | Method for unloading and migrating under mobile awareness in small cell network supporting MEC | |
Zhang et al. | DMRA: A decentralized resource allocation scheme for multi-SP mobile edge computing | |
CN114363984B (en) | Cloud edge collaborative optical carrier network spectrum resource allocation method and system | |
Xia et al. | Joint resource allocation at edge cloud based on ant colony optimization and genetic algorithm | |
Wu et al. | A mobile edge computing-based applications execution framework for Internet of Vehicles | |
CN115396953A (en) | Calculation unloading method based on improved particle swarm optimization algorithm in mobile edge calculation | |
El Haber et al. | Computational cost and energy efficient task offloading in hierarchical edge-clouds | |
CN115801091A (en) | Large-scale constellation network resource scheduling method for satellite-ground cooperative computing | |
CN113010317A (en) | Method, device, computer equipment and medium for joint service deployment and task unloading | |
CN115665258B (en) | Priority perception deployment method of multi-target service function chain based on deep reinforcement learning | |
CN115499875B (en) | Satellite internet task unloading method, system and readable storage medium | |
CN115278779A (en) | Rendering perception-based dynamic placement method for VR service module in MEC network |
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 |