CN-122002300-A - 6G edge collaborative computing network task unloading device based on propagation dynamics

CN122002300ACN 122002300 ACN122002300 ACN 122002300ACN-122002300-A

Abstract

The invention discloses a 6G edge cooperative computing network task unloading device based on propagation dynamics, and belongs to the technical field of computation, calculation or counting. The device considers the dynamic change problem of the task unloading and spreading of the 6G edge collaborative computing network and the problem that the excessive dependence on the D2D technology assists in computing and unloading possibly causes network paralysis and the like, and provides a device for solving the problem of the paralysis of the edge communication network and the problem of the task overload of the node equipment so as to meet the requirement of a user terminal on task unloading. The device firstly builds a 6G edge cooperative computing task offloading network model, secondly defines critical conditions and constraint conditions for preventing the occurrence of paralysis of an edge network, and finally solves the problems of task overload and network paralysis by adopting a joint task offloading propagation model and a critical condition deduction algorithm. Simulation results show that the provided model has a certain reference value for optimizing the edge communication network deployment scheme.

Inventors

ZHANG YUEXIA
ZHU CHAO

Assignees

北京信息科技大学

Dates

Publication Date: 20260508
Application Date: 20241104

Claims (4)

1. Establishing a propagation dynamics-based 6G edge collaborative computing network task offloading device, which is characterized by comprising the following steps: 1) Constructing a 6G edge collaborative computing task offloading network model; 2) Defining critical conditions and constraint conditions for preventing the occurrence of paralysis of the edge network; 3) And solving the problems of task overload and network paralysis by adopting a joint task unloading propagation model and a critical condition derivation algorithm.
2. The device for unloading the 6G edge collaborative computing network task based on propagation dynamics according to claim 1 is characterized in that a 6G edge collaborative computing task unloading network model is built in the step 1), the total number of node devices of the network model is N, the degree distribution is P (k), node devices in four states of a service state, an auxiliary state, a request state and a recovery state exist in the network, the request state user terminal can unload computing tasks to a neighbor user terminal, a neighbor edge server and a cloud server for computation, the computing resources required by each node device for processing a single data packet are the same, the computing resources required by W 0 (cycles) are used for representing that the set of computing resources which can be provided by each node device is f (cycles/s) = { f 1 ,f 2 ,...,f i ,...,f N }, the computing time delay set for processing a single data packet in the network is t= { t 1 ,t 2 ,...,t i ,...,t N }, the computing tasks generated by itself are assumed to reach a lambda i , the service rate is mu i of M/M/1/N queuing model, the service rate is the maximum queuing rate of each node device can be equal to the data packet which can be processed by the node device, and the maximum queuing rate of the data packet can be equal to the maximum queue length of the node devices per second (mu.3272), and the maximum queue length of the data packets can be respectively generated by the node devices per the node device is the maximum queue length of the node device according to the number of the data packets of the corresponding to the auxiliary state i ＝ρ/(1-ρ),L i ＜N p ,N q ,N p ; In the network model, the network is in a stable state at the beginning, node equipment in the two states of a service state and an auxiliary state exists in the network, the proportion of the service state equipment and the auxiliary state user terminal is p and q respectively, wherein p+q=1, excessive calculation tasks are distributed by the base station in the network at a certain time point 0 , so that the user terminal in the proportion of ρ 0 has a task overload condition, the situation is converted into a request state terminal, the request state user terminal can possibly transmit an unloading task to a neighbor auxiliary state user terminal or a neighbor edge server with probability alpha in each time step, the auxiliary state user terminal can be converted into a request state with probability h q (m,T q when the received data packet sequence is accumulated to exceed a threshold value, the service state equipment can be converted into the request state with probability h p (m,T p ), the edge server can generally provide sufficient calculation resources and system capacity, and therefore the service state equipment can always be kept in the service state, meanwhile, each request state user terminal can be converted into the state terminal with probability gamma, and does not participate in unloading task transmission, and the whole network state does not tend to be stable when the network state is not transmitted in the network.
3. The propagation dynamics-based task offloading device for a 6G edge collaborative computing network according to claim 1, wherein in step 2), critical conditions and constraint conditions for preventing occurrence of paralysis of an edge network are defined, and a task offloading process in a 6G edge collaborative computing network model is studied centering on each terminal device and an edge server, and key influencing factors for causing occurrence of paralysis of the network are further discussed; Let θ (t) denote the probability that the neighbor i did not offload tasks to the service state device (or auxiliary state terminal) j before the time t, if the node degree of the service state device (or auxiliary state terminal) j (i.e. the neighbor node terminal or device) is k j , the probability that the node j receives m neighbor offload tasks before the time t can be obtained as follows: For the auxiliary state terminal j, the linear threshold model can know that the state of the auxiliary state terminal j is unchanged when the auxiliary state terminal j receives no more than T q unloading tasks, so that the probability that the auxiliary state user terminal j is kept in the auxiliary state at the moment T is as follows: for the auxiliary state terminal j selected randomly, since the network degree distribution is P (k), at the time t, the probability that the auxiliary state node terminal does not generate state transition in the whole network is as follows: The proportion of node user terminals in the auxiliary state in the whole network at the moment t is as follows: A(t)=(1-ρ 0 )·q·η q (4) for the service state device j, the linear threshold model can know that the state of the service state device j does not change when the service state device j receives no more than T p unloading tasks, so that at the moment T, the probability that the service state device j is kept in the service state is calculated as follows: For the service state device j selected randomly, since the network degree distribution is P (k), at time t, the probability that the service state node device maintains the original state in the whole network is as follows: The proportion of the node devices in the service state in the whole network at the moment t is as follows: S(t)=(1-ρ 0 )·p·η p (7) Because the auxiliary state user terminal or the service state equipment (edge server) is the node participating in the cooperative task unloading and providing the task unloading service in the whole edge network and the unloading propagation process, the action and the state transition mechanism of the auxiliary state node and the service state node in the network are similar and the linear propagation threshold values of the auxiliary state node and the service state node are different, so that the proportion A (t) occupied by the user terminal in the auxiliary state at the moment t and the proportion S (t) occupied by the edge server in the service state can be integrated together to obtain SA (t) for facilitating the subsequent analysis: SA(t)=S(t)+A(t)=(1-ρ 0 )·[pη p +qη q ] (8) Wherein ρ 0 is the ratio of the request state user terminals at the beginning, and SA (t) can be obtained as long as θ (t) is obtained by analysis; The neighbor node i of the auxiliary state node j (or the service state node j) may be in one of four states, θ (t) may be expressed as: θ(t)=ψ S (t)+ψ A (t)+ψ R (t)+ψ H (t) (9) the probability that the neighbor node i is in a service state, an auxiliary state, a request state, a recovery state and does not offload tasks to the auxiliary state (or request state) node j before the time t is respectively represented by phi S (t)、ψ A (t)、ψ R (t)、ψ H (t), and the specific analysis is as follows: if the neighbor i is an auxiliary state terminal, and the node i is provided with k i neighbor nodes, the node j can not offload tasks to the node i because the node j is an auxiliary state terminal or service state equipment, so the node i can only receive the computation tasks for offloading propagation from other k i -1 neighbor nodes, and the probability of receiving n offload tasks before the time t is as follows: Considering that the neighbor i is an auxiliary state node and a corresponding linear threshold model, the probability that the neighbor i receives n offloading tasks before the time t and no state transition occurs at the time t is calculated as follows: In a 6G edge collaborative computing network with a degree distribution of P (k), the probability that a node j and a node i with a node degree of k i are neighbors is k i P(k i )/< k >, wherein < k > is the average degree of the network, and for a randomly selected node i, considering all values of k i , the probability that the node i is an auxiliary state terminal and no state transition occurs at the time t is as follows: If the neighbor i is a service state device, since the node j is an auxiliary state terminal or a service state device, the node j will not offload tasks to the node i, so the node i can only receive the computation tasks of offload propagation from other k i -1 neighbor nodes, and considering that the neighbor i is a service state node and a corresponding linear threshold model, the probability that the neighbor i receives n offload tasks before the time t and the state transition does not occur at the time t is: considering all values of k i for a randomly selected node i, for the whole network, the probability that node i is a service state device and no state transition occurs at time t is: The function and the state transition mechanism of the auxiliary state node and the service state node in the network are similar, but the linear propagation thresholds of the auxiliary state node and the service state node are different, so that the probability that the node i is in the auxiliary state and the service state and the state transition does not occur at the moment t can be analyzed together to obtain the psi SA (t): Because the auxiliary state user terminal or the service state device does not need to offload tasks, the probability that the node i is still in an auxiliary state or a service state (i.e. no state transition occurs) at the time t is the probability that the neighbor node i is the auxiliary state terminal or the service state device and does not offload tasks to the auxiliary state (or service state) node j before the time t; if the neighbor i is a request state terminal, an evolution equation of ψ R (t) needs to be analyzed, and when the request state terminal i unloads tasks to an auxiliary state (or service state) node j with a probability alpha, the following expressed equation form can be obtained: If the neighbor i is a recovery state terminal, and the request state terminal is transformed into the recovery state with the probability of γ after being unloaded and propagated (but not unloaded to the node j) in the previous time interval of the time t, the evolution of ψ H (t) can be represented by the following equation: Combining equations (18), (19) and initial conditions θ (0) =1 and ψ H (0) =0, it can be deduced that: substituting equations (17) and (20) into equation (11) can result in the equation form expressed as follows: Substituting equation (21) into equation (18), the evolution of θ (t) can be re-expressed in terms of the following equations: the density change of each state can be described as, in terms of propagation dynamics, throughout the network: Therefore, by combining and iterating equations (6), (9), (10), (3) and (23), the proportion and evolution trend occupied by the node devices in four states of S (t), a (t), R (t) and H (t) in the whole network at any moment can be calculated; According to the linear threshold model and the SARH task offloading propagation model, when t & gtto & gtinfinity, the system tends to be stable, offloading propagation is stopped, only node terminals or devices in an auxiliary state, a service state and a recovery state are needed in the network, and no user terminal in a request state exists From this it can be deduced that: By combining and iterating equations (6), (9) (10) and (24), and R (≡) =0, S (++), A (++), SA (++) and H (++) can be obtained; the required critical condition, i.e. the critical condition that the occurrence of task overload of a large number of user terminals in a short time in the network causes network paralysis, can be determined by deriving (24) a non-trivial solution, letting the function be as follows: From the analysis, the solution of the critical condition corresponds to the abscissa θ c (++) of the point of tangency of the function (25) with the horizontal axis, where θ c (++) represents the probability that after the network has stabilized, there is always no probability that the offload task propagates to the auxiliary (or service) node j at the critical offload propagation probability α c , and therefore, the critical condition is defined as: Solving equation (26) may derive a critical offload probability: Wherein, the
4. The apparatus for offloading tasks in a 6G edge collaborative computing network based on propagation dynamics according to claim 1, wherein step 3) solves the problems of task overload and network paralysis using a joint task offloading propagation model and a critical condition derivation algorithm, as follows: (1) Initializing the number of base stations, edge servers and user equipment, a task arrival processing model and the data packet capacity of node equipment; (2) Calculating unloading paths of different calculation tasks and processing results according to parameters such as the quantity proportion of the initial request state user terminals, the maximum request state user terminals which can be served by a single auxiliary state user terminal, the quantity proportion of the initial auxiliary state user terminals and the like; (3) Solving critical conditions causing network paralysis by using a critical condition deduction algorithm, a task unloading propagation model and a device state transition linear threshold model; (4) And solving an optimal edge communication network deployment scheme by using a propagation dynamics cooperative task offloading algorithm and a current critical condition.

Description

6G edge collaborative computing network task unloading device based on propagation dynamics Technical Field The invention relates to a communication technology, in particular to a 6G edge cooperative computing network task unloading device based on propagation dynamics, and belongs to the technical field of computation, calculation or counting. Background With the continued development of sixth generation (6G) mobile communication networks and emerging internet of things (Internet of Thing, ioT), smart terminals are rapidly expanding in scale and computing-intensive and time-delay sensitive applications (e.g., immersive cloud augmented reality, holographic communication, autopilot, digital twinning, etc.) are rapidly increasing, resulting in explosive growth in data traffic of the mobile internet, and increasing demands for high-reliability network performance and ultra-low latency. In order to meet the increasingly stringent requirements of emerging intelligent applications, D2D-assisted MEC technology is considered as one of the key technologies to solve the problems faced by 6G edge networks, representing an important component of distributed architecture and ubiquitous edge intelligence in 6G landscape. By unloading complex computing tasks to nearby edge servers or nearby intelligent terminals for processing, computing resources in the idle intelligent terminals can be fully utilized, computing task processing efficiency of the intelligent terminals is effectively improved, task processing time delay and energy consumption of the terminals are greatly reduced, and data flow pressure of a return link is relieved. Thus, the study of D2D-assisted MEC technology is of great significance for 6G edge collaborative networks. However, in 6G edge networks, more and more user terminals choose to offload computing tasks preferentially to surrounding intelligent terminals. When an intelligent terminal node receives excessive offloading tasks, the node may overload the tasks and lose the capability of providing offloading services, and the task processing delay may increase significantly, even in the event of a large area network paralysis. And because 6G edge network equipment is of a plurality of types and large in number scale, the network topology structure is complex, the dynamic difference of service demands is large, and most of researches at present also lack of analysis on the overall condition of the network system. The complex network propagation dynamics theory is used as a tool for researching the propagation rule of information among groups, and can reveal the characteristics and dynamic mechanism of information propagation in the network and cognize and analyze the overall condition of the network. Therefore, from the system perspective, the application of complex network propagation dynamics theory and model to analyze the task offloading dynamics change rule in the 6G edge network is a new approach and trend. Disclosure of Invention Aiming at the problems of dynamic change of the 6G edge cooperative computing network task unloading and transmission, network paralysis possibly caused by over-dependence on D2D technology auxiliary computing and unloading, the invention designs a 6G edge cooperative computing network task unloading device based on transmission dynamics. First, a 6G edge collaborative computing task offload network model is constructed. And secondly, defining critical conditions and constraint conditions for preventing the occurrence of paralysis of the edge network. And finally, solving the problems of task overload and network paralysis by adopting a joint task unloading propagation model and a critical condition derivation algorithm. The invention relates to a propagation dynamics-based 6G edge collaborative computing network task unloading device, which comprises the following three steps: 1) The method comprises the steps of constructing a 6G edge collaborative computing task unloading network model, wherein the total number of node devices of the network model is N, the degree distribution is P (k), the network comprises four node devices of a service state, an auxiliary state, a request state and a recovery state, the request state user terminal can unload computing tasks to a neighbor user terminal, a neighbor edge server and a cloud server to perform computation, the computing resources required by each node device to process a single data packet are the same, the computing resources are represented by W 0 (cycles), the set of computing resources which can be provided by each node device is f (cycles/s) = { f 1,f2,...,fi,...,fN }, the computing time delay set for processing the single data packet in the network is t= { t 1,t2,...,ti,...,tN }, t i＝W0/fi is assumed to be the M/M/1/N queuing model with the arrival rate of lambda i and the service rate of mu i for the computing tasks generated by the user terminal. The service rate is defined as the number of packets that can be pr