Distributed dynamic task offloading method for mobile intelligent clusters based on location prediction

Abstract

The present invention discloses a distributed dynamic task offloading method for a mobile intelligent cluster based on location prediction, which is applied to intelligent devices in an edge environment. The method comprises: each intelligent device estimates its own first location prediction sequence within a preset time period; the intelligent devices include a task node and a communication node, wherein the task node is the intelligent device that generates the task, and the communication node is other intelligent devices within the communication range of each intelligent device; after the task node sends a task offloading request to a base station or its own communication node, it receives a second location prediction sequence sent by the base station and / or the communication node, calculates the reliability value between each second location prediction sequence and the corresponding element in its own first location prediction sequence; calculates the sum of the benefits of different task offloading strategies based on the reliability value; uses the sum of the benefits as a fitness function, and uses an artificial fish school algorithm to search for the optimal offloading strategy. The present invention can ensure the reliability of offloading while reducing offloading latency, thereby improving the success rate of task offloading.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a mobile intelligent cluster distributed dynamic task offloading method based on location prediction. Background Art

[0002] During actual operation, networked intelligent devices generate significant computing power demands. However, due to their limited computing and storage resources, they cannot meet the strict time constraints of computational tasks. Therefore, they distribute the tasks to nearby resource-rich nodes for computation. After the computation is completed, the results are returned to the intelligent devices. This process is called task offloading. Task offloading technology can share the heavy computing demands of terminal nodes, reduce overall system energy consumption, shorten task computation time, and improve system utilization.

[0003] As task offloading technology is applied to scenarios such as mobile edge computing and unmanned intelligent clusters, the high mobility of smart devices has brought challenges to task offloading technology. Therefore, many studies have considered combining node location information with traditional task offloading solutions to improve the reliability of task offloading while ensuring offloading efficiency.

[0004] Currently, some existing studies have considered the impact of high vehicle maneuverability on the unloading process, but there are problems as follows:

[0005] (1) Lack of consideration of the impact of dynamic changes in position on reliability. Existing technologies calculate the reliability of links between entities based on the current motion state of smart devices. For example, by combining vehicle mobility characteristics, the connection relationship between vehicles is described as the maximum task processing capability, where the task selects the optimal unloading path through this connection relationship. However, using task processing capability to represent the workshop relationship is too complicated; or using multi-hop vehicle computing resources based on vehicle mobility analysis, specifically, using time-based inscribed link connectivity to more intuitively represent the connectivity relationship of mobile vehicles. Both of the above solutions use the reliability value calculated from the position relationship and motion status between nodes before unloading as the reliability during unloading, without considering that the real-time changing position relationship of the nodes will lead to dynamic changes in their reliability.

[0006] (2) Lack of consideration of the impact of node communication distance on reliability. In the prior art, there are also methods that use learning models to predict vehicle motion trajectories to calculate motion trajectory similarity. Specifically, a dual-LSTM-based vehicle trajectory prediction model is used to implement a spatiotemporal trajectory prediction process, and uses position deviation and trajectory overlap to describe the motion correlation between vehicles. However, the above method only describes the similarity of vehicle motion trajectories in terms of reliability, and does not consider the communication distance between vehicles. As a result, the task may be offloaded to a vehicle with the same motion direction but too far away, or nodes with rich computing resources but large motion trajectory deviations are ignored, ultimately resulting in low offloading task success rate and system operation efficiency.

[0007] (3) The above-mentioned unloading methods are applicable to simple one-way or two-way roads and cannot be applied to real complex scenarios.

[0008] In summary, existing task offloading methods improperly describe the reliability between smart devices. One type of method uses the reliability value calculated based on the position relationship and movement status between nodes before offloading as the reliability during offloading, without considering that the real-time changing position relationship of nodes will lead to dynamic changes in their reliability; another type of method only uses the position deviation and movement correlation between nodes to characterize the reliability between nodes, without considering the impact of the communication distance between nodes on offloading reliability, resulting in a low success rate of task offloading. Summary of the Invention

[0009] In order to solve the above problems existing in the prior art, the present invention provides a method for distributed dynamic task offloading of mobile intelligent clusters based on location prediction. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0010] The present invention provides a distributed dynamic task offloading method for mobile intelligent clusters based on location prediction, which is applied to intelligent devices in edge environments;

[0011] The method comprises:

[0012] Each smart device estimates and obtains its own first position prediction sequence within a preset time period; the smart device includes a task node and a communication node, the task node is the smart device that generates the task, and the communication node is other smart devices within the communication range of each smart device;

[0013] After sending a task offloading request to the base station or its own communication node, the task node receives the second position prediction sequence sent by the base station and / or the communication node, and calculates the reliability value between each second position prediction sequence and the corresponding element in its own first position prediction sequence;

[0014] Calculating the sum of benefits of different task offloading strategies based on the reliability value;

[0015] The sum of the benefits is used as a fitness function, and the optimal unloading strategy is searched using the artificial fish swarm algorithm.

[0016] In one embodiment of the present invention, the step of each smart device estimating its own first position prediction sequence within a preset time period includes:

[0017] Each smart device predicts its own prior state at each moment within the preset time period to obtain a first position prediction sequence within the preset time period.

[0018] In one embodiment of the present invention, each smart device estimates its own prior state at each moment within the preset time window according to the following steps:

[0019] Each smart device is based on its own state vector x at time k k =[x,y,v x ,v y ] T , estimate its own prior state vector at time k+1:

[0020]

[0021] Among them, g is a nonlinear function, u k is the preset control vector, x and y represent the horizontal and vertical coordinates of the smart device's own position respectively, and v x 、v y Respectively represent the horizontal and vertical speeds of the smart device itself, a x 、a y They represent the horizontal and vertical accelerations of the smart device, respectively, and Δt represents the time interval between time k and time k+1;

[0022] Each smart device estimates the prior noise covariance matrix at time k+1 based on the noise covariance matrix at time k:

[0023]

[0024] Among them, J A represents the Jacobian matrix of the nonlinear function g, T represents the transpose, and Q represents the preset process noise matrix;

[0025] The Kalman gain at time k is calculated according to the prior noise covariance matrix:

[0026]

[0027] Among them, J H represents the Jacobian matrix of the observation function H, and R represents the preset observation noise covariance matrix;

[0028] The position prediction result at time k+1 is calculated based on the Kalman gain:

[0029] x k+1 =x k +K k (z k -H(x k ))

[0030] Among them, z k Represents the observation values ​​of the smart device itself obtained through sensors, including the abscissa, ordinate, abscissa velocity, ordinate velocity, abscissa acceleration, and ordinate acceleration of its own position;

[0031] According to the Kalman gain at the k moment, the prior noise covariance matrix at the k+1 moment is updated:

[0032] P k+1 =(IK k J H )P' k+1 ;

[0033] Where I represents the identity matrix.

[0034] In one embodiment of the present invention, after a task node sends a task offloading request to a base station or its own communication node, the task node receives a second position prediction sequence sent by the base station and / or the communication node, and calculates a reliability value between each second position prediction sequence and a corresponding element in its own first position prediction sequence, including:

[0035] The task node sends a task offloading request to a base station or a communication node within its communication range, and receives a second position prediction sequence sent by the base station and / or the communication node;

[0036] The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence, and quantifies the difference based on its own maximum communication distance;

[0037] The reliability values ​​between the corresponding elements of the second position prediction sequence and the first position prediction sequence itself are calculated based on the difference sequence obtained by quantization.

[0038] In one embodiment of the present invention, the task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence, and quantizes the difference based on its own maximum communication distance, including:

[0039] The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence;

[0040] Comparing the difference with its own maximum communication distance;

[0041] If the difference is less than or equal to the maximum communication distance of the self, the difference is quantized to 1; otherwise, the difference is quantized to 0.

[0042] In one embodiment of the present invention, the reliability value between the corresponding elements of the second position prediction sequence and the first position prediction sequence itself is calculated according to the following formula:

[0043]

[0044] Where, Represents task node V i Its own first position prediction sequence and the communication node or base station V j The rth element of the difference sequence between the second position prediction sequence, r = 1, 2, ..., q.

[0045] In one embodiment of the present invention, the step of calculating the sum of benefits of different task offloading strategies based on the reliability value includes:

[0046] When the task offloading strategy is: at task node V i When executing locally, the task node V i The first unloading benefit is:

[0047]

[0048] Where p iR Represents task node V i Offload tasks to V R Uninstallation strategy, V R represents the set of smart devices and base stations in the edge environment M = {V1, V1, ..., V m} any smart device or base station, when V R For task node V i Time iR =1, otherwise p iR =0, m represents the number of smart devices and base stations in the edge environment, Represents the task offloading strategy p iR The task execution time under , α represents the preset delay factor, and β represents the preset reliability factor;

[0049] When the task offloading strategy is: Task node V i Offload tasks to V R When the task node V i The second unloading benefit is:

[0050]

[0051] Where p iR Represents task node V i Offload tasks to V R Uninstall strategy, when V R For M={V1,V1,...,V m} except task node V i When any other smart device or base station other than iR =1, otherwise p iR =0, Represents task node V i Transfer the task to V R The task transmission time, Re iR Indicates V i With V R Reliability value between

[0052] Based on the first offloading benefit and the second offloading benefit, determine the task set N = {task1, task1, ..., task n The offloading benefit of each task in};

[0053] Calculate the total revenue based on the offload revenue of each task:

[0054]

[0055] Where, f R1 Represents the task set N = {task1,task1,...,task n}The offloading benefit of the R1th task in.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] (1) In the task offloading method provided by the present invention, each smart device predicts its future trajectory by constructing a smart device location prediction model. Compared with other methods based on simple or specific mobility models, this method can be applied to real complex scenarios without the need to train models or general model formulas based on a large amount of data. In addition, it can only predict the location under normal circumstances, and cannot perform exception processing based on normal data in abnormal circumstances, thereby ensuring the robustness of task offloading during operation.

[0058] (2) The present invention calculates the relationship between the two position prediction sequences and the difference and the maximum communication distance as a reliability value by combining the position prediction sequence of the smart device, which can ensure that the positions of the task node and the communication node are close, and greatly reduces the probability of the communication node leaving the communication range of the task node during the unloading process, resulting in the failure of the task unloading. Compared with the method in the prior art that ensures the efficiency of task unloading by calculating the link reliability or the position deviation and the mobility correlation, the present invention can increase the possibility of the potential communication node in the cluster being selected, and the calculation of the reliability value is simple and the efficiency can be guaranteed.

[0059] (3) The present invention constructs an optimization equation by combining the three elements of predicted position, computing resources and unloading delay. In the process of task unloading, it not only considers the influence of two important factors, resources and delay, in the traditional scheme, but also integrates the predicted position to calculate the reliability value. It ensures the reliability of unloading while reducing the unloading delay, thereby improving the success rate of task unloading.

[0060] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of a method for distributed dynamic task offloading of mobile intelligent clusters based on location prediction provided by an embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of an application scenario provided by an embodiment of the present invention;

[0063] Figure 3 This is a flow chart of an artificial fish swarm algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0065] Figure 1 This is a flow chart of a method for distributed dynamic task offloading of mobile intelligent clusters based on location prediction provided by an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a distributed dynamic task offloading method for a mobile intelligent cluster based on location prediction, which is applied to intelligent devices in an edge environment;

[0066] The above methods include:

[0067] S1. Each smart device estimates and obtains its own first position prediction sequence within a preset time period. Smart devices include task nodes and communication nodes. Task nodes are smart devices that generate tasks, and communication nodes are other smart devices within the communication range of each smart device.

[0068] S2. After the task node sends a task offloading request to the base station or its own communication node, it receives the second position prediction sequence sent by the base station and / or the communication node, and calculates the reliability value between each second position prediction sequence and the corresponding element in its own first position prediction sequence;

[0069] S3. Calculate the sum of benefits of different task offloading strategies based on the reliability value;

[0070] S4. Take the total revenue as the fitness function and use the artificial fish swarm algorithm to search for the optimal unloading strategy.

[0071] It should be noted that the task offloading method provided by this invention is applicable to collaborative computing scenarios involving intelligent devices. These devices can typically move freely in complex environments or form clusters, moving from different paths to the same location to perform the same task. These intelligent devices communicate with base stations or other intelligent devices to perform resource scheduling and task offloading. Figure 2 is a schematic diagram of an application scenario provided by an embodiment of the present invention, such as Figure 2 As shown in the figure, a task node is a node that can generate tasks, such as smart device A (a vehicle). The dotted circle around smart device A represents the communication range of the task node. Within this communication range, two smart devices, smart device B and smart device C, can directly communicate with the task node. These are called communication nodes. However, only one of these communication nodes can be the target of a task's offload and become the offload node. Other smart devices outside the task node's communication range are called external nodes and cannot directly communicate with the task node.

[0072] Optionally, the step of each smart device estimating and obtaining its own first position prediction sequence within a preset time period includes:

[0073] Each smart device predicts its own prior state at each moment within a preset time period to obtain a first position prediction sequence within the preset time period.

[0074] In this embodiment, to predict the first position prediction sequence of smart devices within a preset time period, a node position prediction model must be constructed. Each smart device then predicts its own motion trajectory in real time based on this node position prediction model. For example, this embodiment employs a position prediction model based on the extended Kalman filter algorithm. This algorithm uses the current state vector to predict the prior state vector for the next moment, and then updates the prior state vector for the next moment based on the current observation value to obtain the position prediction result for the next moment.

[0075] Optionally, in step S1 above, each smart device estimates its own prior state at each moment within a preset time window according to the following steps:

[0076] S101, each smart device calculates its own state vector x at time k. k =[x,y,v x ,v y ] T , estimate its own prior state vector at time k+1:

[0077]

[0078] Among them, g is a nonlinear function, u k is the preset control vector, x and y represent the horizontal and vertical coordinates of the smart device's own position respectively, and v x 、v y Respectively represent the horizontal and vertical speeds of the smart device itself, a x 、a y They represent the horizontal and vertical accelerations of the smart device, respectively, and Δt represents the time interval between time k and time k+1;

[0079] S102. Each smart device estimates the prior noise covariance matrix at time k+1 based on the noise covariance matrix at time k:

[0080]

[0081] Among them, J A represents the Jacobian matrix of the nonlinear function g, T represents the transpose, and Q represents the preset process noise matrix;

[0082] S103, calculating the Kalman gain at time k according to the prior noise covariance matrix:

[0083]

[0084] Among them, J H represents the Jacobian matrix of the observation function H, and R represents the preset observation noise covariance matrix;

[0085] S104. Calculate the position prediction result at time k+1 according to the Kalman gain:

[0086] x k+1 =x k +K k (z k -H(x k ))

[0087] Among them, z k Represents the observation values ​​of the smart device obtained by the sensor. The observation values ​​include the horizontal coordinate, vertical coordinate, horizontal coordinate speed, vertical coordinate speed, horizontal coordinate acceleration, and vertical coordinate acceleration of the smart device.

[0088] S105. Update the prior noise covariance matrix at time k+1 according to the Kalman gain at time k:

[0089] P k+1 =(IK k J H )P' k+1 ;

[0090] Where I represents the identity matrix.

[0091] Optionally, in step S2, after the task node sends a task offloading request to the base station or its own communication node, the task node receives a second position prediction sequence sent by the base station and / or the communication node, and respectively calculates a reliability value between each second position prediction sequence and a corresponding element in its own first position prediction sequence, including:

[0092] S201: The task node sends a task offloading request to a base station or a communication node within its communication range, and receives a second position prediction sequence sent by the base station and / or the communication node;

[0093] S202: The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence, and quantifies the difference based on its own maximum communication distance;

[0094] S203: Calculate reliability values ​​between corresponding elements of the second position prediction sequence and the first position prediction sequence itself according to the difference sequence obtained by quantization.

[0095] Specifically, the task node V0 sends an unloading request to the communication node or base station within its communication range at a certain moment. The communication node or base station V j After receiving the offloading request, it calculates the time to execute the task and its second position prediction sequence Set j =[p1,p2,...,p q ] is sent to the task node V0. Since the position of the base station is fixed, each element in the second position prediction sequence of the base station is the same fixed constant.

[0096] In step S202, the task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence, and quantifies the difference based on its own maximum communication distance, including:

[0097] The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence;

[0098] Comparing the difference with its own maximum communication distance;

[0099] If the difference is less than or equal to the maximum communication distance of the self, the difference is quantized to 1; otherwise, the difference is quantized to 0, that is:

[0100]

[0101] In the above formula, L0 represents the maximum communication distance of the task node V0 itself, Set0[r] represents the rth element in the first position prediction sequence of the task node V0 itself, and Set j [r] represents a communication node or base station V j The second position of the predicted sequence is the rth element, is a difference sequence.

[0102] Furthermore, the reliability value between the corresponding elements of the second position prediction sequence and its own first position prediction sequence is calculated according to the following formula:

[0103]

[0104] in, Represents task node V i Its own first position prediction sequence and the communication node or base station V j The rth element of the difference sequence between the second position prediction sequence, r = 1, 2, ..., q.

[0105] It should be understood that the total benefit of task execution is determined by the task offloading strategy. In order to avoid a large difference between delay and reliability values, this embodiment introduces a delay normalization process in the benefit calculation.

[0106] When the task offloading strategy is: at task node V i When executing locally, the task node V i The first unloading benefit is:

[0107]

[0108] Where p iR Represents task node V i Offload tasks to V R Uninstallation strategy, V R represents the set of smart devices and base stations in the edge environment M = {V1, V1, ..., V m} any smart device or base station, when V R For task node V i Time iR =1, otherwise p iR =0, m represents the number of smart devices and base stations in the edge environment, Represents the task offloading strategy p iRThe task execution time under , α represents the preset delay factor, and β represents the preset reliability factor;

[0109] When the task offloading strategy is: Task node V i Offload tasks to V R When the task node V i The second unloading benefit is:

[0110]

[0111] Where p iR Represents task node V i Offload tasks to V R Uninstall strategy, when V R For M={V1,V1,...,V m} except task node V i When any other smart device or base station other than iR =1, otherwise p iR =0, Represents task node V i Transfer the task to V R The task transmission time, Re iR Indicates V i With V R Reliability value between

[0112] Based on the first offloading benefit and the second offloading benefit, determine the task set N = {task1, task1, ..., task n The offloading benefit of each task in};

[0113] Calculate the total revenue based on the offload revenue of each task:

[0114]

[0115] Where, f R1 Represents the task set N = {task1,task1,...,task n}The offloading benefit of the R1th task in.

[0116] Furthermore, this embodiment uses the above-mentioned sum of benefits as the fitness function of the artificial fish school algorithm, and uses the meta-heuristic algorithm "artificial fish school algorithm" to solve the above-mentioned optimization problem. Figure 3 This is a flow chart of the artificial fish swarm algorithm provided by an embodiment of the present invention. Figure 3As shown, the artificial fish swarm algorithm simulates the predation, flocking, and following behaviors of a fish population by constructing artificial fish. During predation, the artificial fish swims in random directions within its field of view; during flocking, it swims toward the center of its fitness; and during following, it swims toward the direction of maximum fitness. Each artificial fish continues to repeat predation, flocking, and following behaviors in parallel during each iteration. The fitness corresponding to the current state is compared with the fitness obtained from these three behaviors, and the state corresponding to the maximum fitness is set as the current optimal unloading strategy.

[0117] It can be seen from the above embodiments that the beneficial effects of the present invention are:

[0118] (1) In the task offloading method provided by the present invention, each smart device predicts the future trajectory by constructing a smart device location prediction model. Compared with other methods based on simple or specific mobility models, this method can be applied to real complex scenarios without the need to train models or general model formulas based on a large amount of data. In addition, the location can only be predicted under normal circumstances, and in abnormal circumstances, it cannot be processed based on normal data, thereby ensuring the robustness of task offloading during operation.

[0119] (2) The present invention calculates the relationship between the two position prediction sequences and the difference and the maximum communication distance as a reliability value by combining the position prediction sequence of the smart device, which can ensure that the positions of the task node and the communication node are close, and greatly reduces the probability of the communication node leaving the communication range of the task node during the unloading process, resulting in the failure of the task unloading. Compared with the method in the prior art that ensures the efficiency of task unloading by calculating the link reliability or the position deviation and the mobility correlation, the present invention can increase the possibility of the potential communication node in the cluster being selected, and the calculation of the reliability value is simple and the efficiency can be guaranteed.

[0120] (3) The present invention constructs an optimization equation by combining the three elements of predicted position, computing resources and unloading delay. In the process of task unloading, it not only considers the influence of two important factors, resources and delay, in the traditional scheme, but also integrates the predicted position to calculate the reliability value. It ensures the reliability of unloading while reducing the unloading delay, thereby improving the success rate of task unloading.

[0121] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0122] Descriptions with reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0123] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art can understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims.

[0124] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A distributed dynamic task offloading method for mobile intelligent clusters based on location prediction, characterized in that: Applied to smart devices in edge environments; The method comprises: Each smart device estimates and obtains its own first position prediction sequence within a preset time period; the smart device includes a task node and a communication node, the task node is the smart device that generates the task, and the communication node is other smart devices within the communication range of each smart device; After sending a task offloading request to the base station or its own communication node, the task node receives the second position prediction sequence sent by the base station and / or the communication node, and calculates the reliability value between each second position prediction sequence and the corresponding element in its own first position prediction sequence; Calculating the sum of benefits of different task offloading strategies based on the reliability value; The sum of the benefits is used as a fitness function, and an artificial fish swarm algorithm is used to search for an optimal unloading strategy; After the task node sends a task offloading request to the base station or its own communication node, the task node receives a second position prediction sequence sent by the base station and / or the communication node, and respectively calculates a reliability value between each second position prediction sequence and a corresponding element in its own first position prediction sequence, including: The task node sends a task offloading request to a base station or communication node within its communication range, and receives a second position prediction sequence sent by the base station and / or communication node; the task node calculates the difference between each second position prediction sequence and the corresponding element in its first position prediction sequence, and quantizes the difference based on its maximum communication distance; and calculates the reliability value between the corresponding elements in the second position prediction sequence and its first position prediction sequence based on the quantized difference sequence; The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence, and quantifies the difference based on its own maximum communication distance, including: The task node calculates the difference between each second position prediction sequence and the corresponding element in its own first position prediction sequence; compares the difference with its own maximum communication distance; if the difference is less than or equal to its own maximum communication distance, quantizes the difference to 1; otherwise, quantizes the difference to 0; The reliability value between the corresponding elements of the second position prediction sequence and its own first position prediction sequence is calculated according to the following formula: ； Where, Represents a task node Its own first position prediction sequence and the communication node or base station The first difference sequence between the second position prediction sequence elements, ; The step of calculating the sum of benefits of different task offloading strategies based on the reliability value includes: When the task offloading strategy is: When executing locally, the task node The first unloading benefit is: ； Where, Represents a task node Offload tasks to Uninstall strategy, Represents the collection of smart devices and base stations in the edge environment When any smart device or base station For task nodes hour ,otherwise , represents the number of smart devices and base stations in the edge environment, Indicates the task offloading strategy The task execution time is Indicates the preset delay factor, represents the preset reliability factor; When the task offloading strategy is: Task node Offload tasks to When the task node The second unloading benefit is: ； Where, Represents a task node Offload tasks to Uninstall strategy, when for Remove task nodes When any other smart device or base station other than ,otherwise , Represents a task node Transfer tasks to The task transfer time, express and Reliability value between Determine a task set based on the first uninstallation benefit and the second uninstallation benefit The offloading benefit of each task in ; Calculate the total revenue based on the offload revenue of each task: Where, Represents a task set Middle The uninstall revenue of a task.

2. The method for distributed dynamic task offloading of mobile intelligent clusters based on location prediction according to claim 1 is characterized in that: The step of each smart device estimating and obtaining its own first position prediction sequence within a preset time period includes: Each smart device predicts its own prior state at each moment within the preset time period to obtain a first position prediction sequence within the preset time period.

3. The method for distributed dynamic task offloading of mobile intelligent clusters based on location prediction according to claim 2 is characterized in that: Each smart device estimates its own prior state at each moment within the preset time window according to the following steps: Each smart device The state vector at time , I estimate myself in The prior state vector at time: ； in, is a nonlinear function, is the preset control vector, 、 Respectively represent the horizontal and vertical coordinates of the smart device's own position, 、 Respectively represent the horizontal and vertical speeds of the smart device itself, 、 Respectively represent the horizontal and vertical acceleration of the smart device itself, express Moment and the time interval between moments; Each smart device The noise covariance matrix at time t is estimated The prior noise covariance matrix at time : ； in, Represents a nonlinear function The Jacobian matrix of represents transpose, represents the preset process noise matrix; Calculate the prior noise covariance matrix The Kalman gain at time : in, Represents the observation function H The Jacobian matrix of Represents the preset observation noise covariance matrix; According to the Kalman gain calculation The position prediction result at the moment: in, Represents the observation values ​​of the smart device itself obtained through sensors, including the abscissa, ordinate, abscissa velocity, ordinate velocity, abscissa acceleration, and ordinate acceleration of its own position; According to the The Kalman gain at the moment The prior noise covariance matrix at the moment is updated: ； in, Represents the identity matrix.

Images