Mobile terminal device data processing method and system based on edge computing

Abstract

The application discloses a kind of based on edge computing's mobile terminal equipment data processing method and system, comprising: mobile terminal equipment is based on task type, equipment resource state and network environment parameter, utilizes dynamic decision model and divides local computing task and edge computing task;Based on load balancing strategy, select target edge node from edge computing network, establish secure transmission channel, and distribute edge computing task to target edge node through secure transmission channel;Local computing task is executed multi-core heterogeneous computing resource allocation strategy, and parallel processing is carried out using dynamic memory partition mechanism, and data cleaning, feature extraction and data compression operation are synchronously executed;The computing result returned by target edge node is received, and after computing result and local processing result are aligned in time and space dimension, data fusion processing is executed, and final processing result is output.The application effectively improves the efficiency, security and reliability of overall data processing.

Description

Technical Field

[0001] This invention relates to the technical field of mobile communication and edge computing integration, and in particular to a data processing method and system for mobile terminal devices based on edge computing. Background Technology

[0002] As applications become increasingly complex and computational demands grow, the computing power and resources of mobile terminal devices themselves cannot meet users' needs for real-time performance, efficiency, and low latency. To alleviate this contradiction, edge computing technology has gradually emerged and is widely used. As a distributed computing framework, edge computing can deploy computing, storage, and network service capabilities on edge nodes close to terminal devices to reduce the computational burden on mobile terminal devices, improve application response speed, and reduce network latency and data transmission costs.

[0003] In recent years, existing technologies have failed to fully consider the real-time changes in the local resource status of mobile terminal devices and the dynamic impact of complex network environment parameters. As a result, there are problems such as low efficiency in task offloading decisions, unreasonable selection of edge nodes, and insufficient allocation of local task processing resources.

[0004] Traditional edge computing-based task offloading and resource scheduling schemes typically employ static or semi-static decision-making methods, failing to effectively track dynamic changes in device resource status and network environment in real time. This can easily lead to unreasonable task partitioning, resulting in wasted local processing resources or excessive load, uneven load on edge nodes, and overall low task execution efficiency. Furthermore, existing solutions often neglect security and transmission reliability management during task transmission, posing a risk of data leakage or loss for sensitive data or critical tasks with high security requirements. In addition, current technologies lack efficient parallel processing mechanisms for local computing tasks, especially failing to achieve optimal allocation and collaborative parallel computing when facing various heterogeneous computing resources, leading to low resource utilization and increased task processing latency. Moreover, the fusion processing mechanism for computation results from edge nodes and local computation results lacks effective time and space alignment methods, easily resulting in reduced accuracy of fusion results and poor data consistency.

[0005] In summary, existing edge computing-based mobile terminal data processing technologies suffer from problems such as unreasonable task decision-making, low efficiency in local resource utilization, and low accuracy in result fusion processing. Therefore, there is an urgent need to propose a method that can dynamically decide on task partitioning in real time, accurately select edge nodes, optimize the allocation of local multi-core heterogeneous resources, and achieve efficient data fusion. By using efficient multi-core heterogeneous computing resource allocation and dynamic memory partitioning mechanisms, combined with precise time and space alignment technology for data fusion, more efficient and reasonable data processing for mobile terminal devices can be achieved. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the aforementioned existing problems, the present invention is proposed.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: The mobile terminal device divides the tasks to be processed based on the task type, device resource status and network environment parameters using a dynamic decision model to obtain local computing tasks and edge computing tasks that need to be offloaded to edge nodes.

[0009] Based on a load balancing strategy, target edge nodes are selected from the edge computing network, a secure transmission channel is established, and the edge computing tasks are distributed to the target edge nodes through the secure transmission channel.

[0010] The local computing task is processed using a multi-core heterogeneous computing resource allocation strategy, a dynamic memory partitioning mechanism for parallel processing, and data cleaning, feature extraction and data compression operations are performed simultaneously.

[0011] The system receives the calculation results returned by the target edge node, aligns the calculation results with the local processing results in the time and space dimensions, performs data fusion processing, and outputs the final processing result.

[0012] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, the task types include real-time data processing tasks, latency-sensitive tasks, and high-computational-load tasks.

[0013] The device resource status includes CPU utilization, memory utilization, and remaining battery power.

[0014] The network environment parameters include the current network bandwidth, network latency, and packet loss rate.

[0015] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, a dynamic decision model is used to divide local computing tasks into edge computing tasks that need to be offloaded to edge nodes, including:

[0016] Obtain the task type of the task to be processed, and assign a numerical label to the task type;

[0017] Obtain the device resource status of the mobile terminal device, including CPU utilization, memory utilization, and remaining battery power;

[0018] Detect network environment parameters, including current network bandwidth, network latency, and packet loss rate;

[0019] The task type, the device resource status, and the network environment parameters are preprocessed and normalized.

[0020] An interpolation function is established to calculate the difference in overall cost between executing a task locally and executing it on an edge node. When the difference is greater than zero, the task is classified as an edge computing task, and when the difference is less than or equal to zero, the task is classified as a local computing task.

[0021] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, the method selects a target edge node from the edge computing network based on a load balancing strategy and establishes a secure transmission channel, including:

[0022] Call the edge computing network status monitoring interface to obtain the load information and network environment parameters of each candidate edge node;

[0023] Based on preset weighting factors, the CPU utilization, memory utilization, available bandwidth, network latency and packet loss rate of the candidate edge nodes are comprehensively scored.

[0024] Select several target edge nodes whose scores meet the preset threshold, and initiate a secure handshake process between the mobile terminal device and the target edge nodes to negotiate the encryption suite and key;

[0025] After completing the digital certificate verification, the system enters a secure session state and establishes an encrypted data transmission channel.

[0026] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, the edge computing tasks to be offloaded are distributed to the target edge nodes, including:

[0027] Based on the task type and data size, the edge computing tasks to be distributed are packaged and divided into blocks, and the block data is digitally signed.

[0028] Each data block is sent to the target edge node through the secure transmission channel. If no acknowledgment response is received within a preset time, a retransmission operation is performed.

[0029] Once the target edge node confirms that the reception is correct, it adds the edge computing task to its execution queue for computation processing.

[0030] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, a multi-core heterogeneous computing resource allocation strategy is executed on the local computing task, and a dynamic memory partitioning mechanism is used for parallel processing, including:

[0031] Scan the CPU multi-core, GPU, and NPU computing units within the mobile terminal device to generate a resource description list;

[0032] The local computing task is decomposed into multiple subtasks. Based on the computing characteristics and priorities of the subtasks, the execution cost of the subtasks on different computing units is calculated using a resource scheduling function.

[0033] Based on the execution cost, a scheduling algorithm is used to map the subtasks to the corresponding computing units. If a load spike is detected during execution, an online update mechanism is triggered.

[0034] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, the dynamic memory partitioning mechanism includes:

[0035] Initially divide the memory into multiple partitions, and reserve a partition for video memory;

[0036] When the usage of a memory partition is detected to be close to the limit, check the free status of other partitions and merge some free address space into the partition that currently needs it.

[0037] When the overall memory space is insufficient, a data compression operation is triggered to alleviate memory pressure;

[0038] After the subtask is completed, the corresponding memory partition is released and the memory usage table is updated.

[0039] As a preferred embodiment of the mobile terminal device data processing method based on edge computing described in this invention, the data fusion processing includes:

[0040] Receive the calculation results returned by the target edge node under a secure transmission channel, and verify the checksum and digital signature;

[0041] If the verification is successful, record the timestamp and spatial index of the edge node results, and align them with the local processing results in a unified coordinate system in terms of time and space.

[0042] Perform a data fusion operation on the aligned local processing results and the edge node results, and output the final fused result.

[0043] The data fusion operation is performed through a constructed data fusion model, the mathematical expression of which is:

[0044]

[0045] Where A is the start time of the actual data collection, which is the lower limit of the external integration, and B is the end time of the actual data collection, which is the upper limit of the external integration. To sum the results and use kernel indices from one to δ, where δ is the upper limit of the summation and the fusion depth threshold, Let Λ be the variable phase offset factor for the aligned data, Λ be the reduction coefficient in the denominator of the fusion model, μ be the starting spatial scale of the data filtering corresponding to the lower bound of the embedded integral, ν be the ending spatial scale of the data filtering corresponding to the upper bound of the embedded integral, Φ be the variable-time-space mapping of the embedded integral, X be the coordinates of the variables of the outer integral and the overall iteration process, ξ be the lower bound of the alignment function R to define the alignment start interval, η be the upper bound of the alignment function R to define the alignment end interval, L be the local processing result function mapping, and E be the edge node return result function mapping. For time-aligned indexing The resulting time independent variable For spatial alignment indexing The resulting spatial independent variable, ∥L∥∥E∥, represents the norm product of the local and marginal results in vector form, used to scale the alignment difference; erf(·) is the error function used to filter out abnormal distributions; and Z is the final fused output value.

[0046] If all integrals and summations converge, then Z takes the real number in the interval (0, +∞).

[0047] The larger Z is, the higher the degree of superposition between the local processing result and the edge node result after fusion, the more observable the superposition is.

[0048] Conversely, if Z approaches 0, then there is a data anomaly.

[0049] As a preferred embodiment of the mobile terminal device data processing system based on edge computing according to the present invention, it includes: one or more processors;

[0050] The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the aforementioned edge computing-based mobile terminal device data processing method.

[0051] As a preferred embodiment of the computer-readable medium for storing software according to the present invention, the software includes instructions executable by one or more computers, the instructions causing the one or more computers to perform operations, the operations including the flow of the aforementioned mobile terminal device data processing method based on edge computing.

[0052] The beneficial effects of this invention are:

[0053] 1. Based on the nature of real-time tasks, the current resource occupancy of devices, and dynamic parameters of the network environment, this invention enables intelligent dynamic decision-making between local processing on mobile terminals and offloading processing at edge nodes, thereby making resource allocation more precise and flexible, avoiding overload or idleness of device resources, improving overall task processing efficiency, and reducing task response latency.

[0054] 2. By selecting appropriate nodes and ensuring secure and reliable data transmission, the problem of overloading or idle resources of a single edge node is effectively avoided, maintaining the overall load balance of the edge network. At the same time, the establishment and management of secure transmission channels ensure the security and reliability of task data transmission, making it suitable for processing security-sensitive data, reducing the risk of data leakage or tampering, and improving data security and edge node computing efficiency.

[0055] 3. By efficiently utilizing and optimizing local resources on the mobile terminal, heterogeneous computing resources such as CPU multi-core, GPU, and NPU are rationally allocated according to the nature of the task, and the memory partition size is dynamically adjusted according to the real-time memory usage of the task. This effectively avoids excessive load on a single computing core or memory bottlenecks, ensuring full utilization of each computing resource. At the same time, data cleaning, feature extraction, and data compression operations are performed synchronously, further reducing redundancy in the data processing process, improving the parallel processing efficiency of local computing tasks, reducing processing latency, and achieving the beneficial effects of improved computing efficiency and reduced device power consumption.

[0056] 4. Through the high precision and consistency of multi-source data fusion processing, the results returned by edge nodes and the local calculation results are synchronized and aligned in the time dimension and the coordinates are unified in the spatial dimension. This ensures the accuracy and reliability of multi-source heterogeneous data fusion, effectively avoids possible conflicts and errors in the fusion process, and greatly improves the quality and credibility of the fused data. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0058] Figure 1 This is a flowchart illustrating the data processing method for mobile terminal devices based on edge computing as shown in this invention. Detailed Implementation

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0060] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0061] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0062] According to an embodiment of the present invention, in combination Figure 1 The flowchart shown illustrates a data processing method for mobile terminal devices based on edge computing, which specifically includes the following steps:

[0063] S1. Based on task type, device resource status, and network environment parameters, the mobile terminal device uses a dynamic decision model to divide the tasks to be processed, resulting in local computing tasks and edge computing tasks that need to be offloaded to edge nodes. It should be noted that in this step:

[0064] Mobile terminal devices identify the task type of the task to be processed. The identification results include, but are not limited to, real-time data processing tasks, latency-sensitive tasks, and high-computation tasks.

[0065] At the same time, the current device resource status is monitored in real time, including at least CPU utilization, memory utilization and remaining power, which correspond to the sufficiency of resources available for computing on the mobile terminal device at the current moment.

[0066] And by providing real-time feedback on network connection status, network environment parameters are obtained, which include at least the current network bandwidth, network latency, and packet loss rate.

[0067] Furthermore, the three types of parameters obtained above (i.e., task type, device resource status, and network environment parameters) are preprocessed and normalized; among which:

[0068] Task type: Maintain a task type mapping table within the mobile terminal device. For example, real-time data processing task → label value "1", latency-sensitive task → label value "2", high computational task → label value "3", thereby converting the task type into a discrete label that can be used for numerical operations.

[0069] Equipment resource status: Keep its value range (e.g., [0,1]) in the same dimension so that it can be calculated in the same model;

[0070] Network environment parameters: mapped according to their physical meaning. For example, higher network bandwidth is more advantageous, while higher network latency or packet loss rate is more disadvantageous.

[0071] In this embodiment, the dynamic decision model comprehensively considers the overall cost of executing the current task locally versus offloading it to an edge node, and makes the optimal decision based on the comparison results. For example, the decision model can be formalized as a difference function D:

[0072] D(T)=[C local (T)]-[C edge (T)]

[0073] Where D(T) represents the difference between the total cost of executing task T locally and the total cost of offloading it to the edge node, and C local (T) represents the total cost of task T when it is executed locally, C edge (T) represents the total cost of offloading task T to an edge node for execution;

[0074] When D(T)>0, it means that the cost of local execution is higher than that of offloading execution, and the task is then classified as an edge computing task.

[0075] When D(T)≤0, the task is classified as a local computing task;

[0076] For example, the local execution cost function C local (T):

[0077] C local (t)=α1·f(T type )+α2·CPUutil+α3·MEMutil+α4·(1-Battery)

[0078] Where, f(T) type ) is the computational load factor set according to the task type, CPUutil represents the CPU utilization at the current moment, MEMutil represents the memory utilization at the current moment, (1-Battery) indicates that the lower the remaining power, the greater the negative impact on local execution, and α1, α2, α3, α4 are the weight factors of local execution.

[0079] For example, unloading the execution cost function C edge (T):

[0080] C edge(T)=β1·(1 / Bandwidth)+β2·Delay+β3·LossRate+β4·f(T type )

[0081] Where Bandwidth represents the current network bandwidth value. The smaller the bandwidth, the larger 1 / Bandwidth, which increases the offloading cost. Delay represents network latency, LossRate represents network packet loss rate, and β1, β2, β3, and β4 are weighting factors related to offloading cost.

[0082] but:

[0083] D(T)

[0084] =α1f(T type )+α2CPUutilα3MEMutilα4(1-Battery)-[β1(1 / Bandwidth)+β2Delayβ3LossRateβ4f(T type )]

[0085] If D(T)≤0, then task T is divided into local computing tasks;

[0086] If D(T)>0, then task T is classified as an edge computing task;

[0087] If it is a local computing task, it will directly enter the local multi-core heterogeneous resource scheduling and execution process (step S3);

[0088] If the task is an edge computing task that needs to be offloaded to an edge node, it will be handled by the load balancing and transmission channel establishment process in step S2.

[0089] It is easy to understand that local computing tasks include critical tasks that have high real-time requirements but small data volumes and are difficult to rely on remotely when the network environment is poor, as well as ordinary tasks that can be completed locally by mobile terminal devices when resources are sufficient and the network is unstable; while edge computing tasks are those that have high computational intensity requirements, require long execution time, and whose local resource status is difficult to meet.

[0090] S2. Based on a load balancing strategy, select target edge nodes from the edge computing network, establish a secure transmission channel, and distribute edge computing tasks to the target edge nodes through the secure transmission channel. Note that the following points should be noted in this step:

[0091] The mobile terminal device calls the edge computing network's status monitoring interface to obtain the resource usage (i.e., load information) of each accessible edge node, including CPU utilization, memory utilization, current task queue length, and remaining available computing power. Simultaneously, it collects network environment parameters with each candidate edge node. Based on the obtained node load information, network environment parameters, and preset weighting factors, each edge node is scored.

[0092] Score i =γ1·(1-CPUutil i )+γ2·(1-MEMutil i )+γ3·(Bandwidth i )-γ4·(Delay i )-γ5·(LossRate i )

[0093] Among them, Score i CPUutil represents the comprehensive score of the i-th candidate edge node. i MEMutil i These represent the current CPU and memory utilization of the node, respectively, and the Bandwidth. i Delay i LossRate i These represent the available bandwidth, latency, and packet loss rate of the mobile terminal device, respectively. γ1, γ2, γ3, γ4, and γ5 are all preset weighting factors.

[0094] Based on the calculated comprehensive score, the candidate nodes are sorted from high to low, and nodes with low scores are removed. The nodes with the highest scores are then selected as the target edge nodes.

[0095] In an optional implementation, the mobile terminal device initiates a secure handshake process with the target edge node, employs encryption suite negotiation and key exchange mechanisms to ensure the confidentiality and integrity of the data channel, and uses digital certificates for two-way authentication between the terminal and the edge node. If any step fails to authenticate, the connection is immediately terminated and other candidate nodes are selected.

[0096] After completing the negotiation of encryption algorithm parameters and keys, the mobile terminal device exchanges confirmation messages with the target edge node and enters a secure session state, providing a reliable channel for the distribution and reception of subsequent task data. The mobile terminal device packages and blocks the data according to the specific type and data scale of the edge computing task divided in step S1, and digitally signs the data to ensure its confidentiality and tamper resistance.

[0097] Before sending a task, the mobile terminal device queries the target edge node for its current task queue information. If there are already high-priority tasks in the queue within the node, the urgency of the local task needs to be assessed to determine whether the local task needs to be prioritized.

[0098] It is easy to understand that, under the established secure transmission channel, packaged data is sent in blocks, each block carrying corresponding verification information. If no acknowledgment response is received from the target edge node within a preset time, a retransmission operation is immediately performed.

[0099] Once the target edge node successfully receives all chunks and completes integrity verification, it returns a confirmation message through a secure channel, thus informing the mobile terminal device that the task has been queued for execution at the target edge node.

[0100] S3. A multi-core heterogeneous computing resource allocation strategy is implemented for the local computing task, employing a dynamic memory partitioning mechanism for parallel processing, and simultaneously performing data cleaning, feature extraction, and data compression operations. It should be noted that the following points are important in this step:

[0101] The mobile terminal device starts the local resource management module, scans the available computing units inside, including CPU multi-core (such as CPU core 1, CPU core 2, CPU core n), GPU and NPU (neural network processing unit), detects the current available load capacity, instruction set characteristics and peak execution capability of each computing unit, and forms a resource description list.

[0102] The data to be processed, which has been divided into local computing tasks in step S1, is segmented and analyzed to identify the sub-task types (such as matrix operations, convolution operations, and general computing). If a sub-task is suitable for batch processing on the GPU, it is specially marked in the decomposition stage so as to enable efficient scheduling during resource allocation.

[0103] Based on the real-time requirements, computational complexity, and data size of each subtask, assign a priority label (e.g., urgent priority, high priority, normal priority); for high-priority tasks, idle processing cores can be preempted when computing unit resources are limited.

[0104] Establish a resource scheduling function Π(j,k) to describe the execution cost of mapping the j-th subtask to the k-th computing unit. The execution cost considers a combination of factors, including the peak execution speed of the task on different cores, data transfer overhead, and the current load. An example formula is shown below:

[0105] Π(j,k)=δ1·CompCost(j,k)+δ2·DataTrans(j,k)+δ3·CurLoad(k)

[0106] Where CompCost(j,k) represents the theoretical computational cost of executing subtask j on the k-th processing unit, DataTrans(j,k) represents the time or bandwidth loss required to load the subtask data into the processing unit, CurLoad(k) represents the current load level of processing unit k, and δ1, δ2, δ3 are adjustable weighting factors.

[0107] The execution cost between each subtask and each processing unit is calculated based on the scheduling strategy model. A comprehensive minimization scheduling algorithm (such as a greedy algorithm) is used to globally allocate all subtasks. If a surge in the load of a computing unit is detected during execution, the online update mechanism of the allocation strategy is triggered, and some unfinished subtasks are migrated to other idle processing units.

[0108] Furthermore, the available physical memory of the device is initially divided, for example, into several memory partitions, each partition corresponding to a fixed-size contiguous address space. For dedicated processing units such as GPU and NPU, a separate video memory partition is reserved as the core buffer for subsequent parallel processing.

[0109] If the usage of a partition approaches its limit, check the free status of other low-load partitions and perform real-time expansion, merging some free address space into the partition in need.

[0110] If the overall memory space is insufficient, a data compression operation is triggered to alleviate memory pressure;

[0111] After the subtask is completed, immediately release the corresponding partition and update the memory usage table;

[0112] For subtasks with normal priority, if they occupy a large amount of memory partitions when high-priority tasks arrive, then a forced reduction in storage is performed to ensure that high-priority tasks can be executed smoothly.

[0113] Once all subtasks have been mapped to their respective processing units and sufficient memory has been allocated, the subtasks will start in parallel.

[0114] For subtasks that require interaction, a lightweight synchronization and communication mechanism (such as a shared memory queue) is established between their corresponding processing units to ensure that the task execution order and data exchange are orderly.

[0115] If a computational anomaly occurs, a failover mechanism is triggered, and the corresponding subtask is rescheduled to other available processing units.

[0116] It should be noted that during parallel processing, the input data needs to be checked for integrity and anomalies removed. If non-compliant values, null values, or format errors are found, a removal mechanism is implemented to reduce interference with subsequent feature extraction. At the same time, depending on the task type, discrete transformation is used to perform feature extraction operations on the input data, making full use of the computing power of the GPU unit. After data cleaning and feature extraction are completed, the data is formatted and compressed to reduce storage space and network transmission bandwidth pressure.

[0117] S4. Receive the calculation results returned by the target edge node, align the calculation results with the local processing results in the time and spatial dimensions, perform data fusion processing, and output the final processing result. Note the following in this step:

[0118] After the target edge node completes the task, it returns the calculation result through a secure transmission channel. When the mobile terminal device receives the result data packet, it decrypts and verifies the integrity of the packet's checksum and digital signature.

[0119] If a packet is detected as corrupted, has an invalid signature, or is incomplete, a retransmission is triggered.

[0120] If the result data is normal, it will be temporarily stored according to the predetermined data format and marked with the corresponding timestamp and spatial index (including spatial or geographical location data);

[0121] For both the local processing results and the results returned by the edge nodes, obtain the timestamps recorded at the start of the task.

[0122] When the data involves coordinates, geographic location, and 3D spatial point cloud information, a unified coordinate transformation model is used to map the local results and edge node results to correct the differences between them in coordinate system, resolution, and geographic projection method.

[0123] If there is relative displacement, registration is performed according to the pre-set reference attitude to ensure that the data can be accurately stitched in the same spatial domain.

[0124] After alignment is completed, the local processing results and the results returned by the edge nodes are used as inputs and imported into the constructed data fusion model for fusion calculation. Specifically, the fusion model comprehensively considers the credibility, value distribution and differential characteristics of the local results and the edge results. After the model calculation is completed, the fusion result is output and stored in the mobile terminal device.

[0125] In an optional implementation, after the fusion is completed, outlier detection is performed on the result data to ensure that the output result is within a reasonable range. If it exceeds the set threshold range, an anomaly information is reported.

[0126] As an example, if the aligned local results and edge results are combined and the final fused value Z is output, then the mathematical expression of the data fusion model is:

[0127]

[0128] Where A is the start time of the actual data collection, which is the lower limit of the external integration, and B is the end time of the actual data collection, which is the upper limit of the external integration. To sum the results and use kernel indices from one to δ, where δ is the upper limit of the summation and the fusion depth threshold, Let Λ be the variable phase offset factor for the aligned data, Λ be the reduction coefficient in the denominator of the fusion model, μ be the starting spatial scale of the data filtering corresponding to the lower bound of the embedded integral, ν be the ending spatial scale of the data filtering corresponding to the upper bound of the embedded integral, Φ be the variable-time-space mapping of the embedded integral, X be the coordinates of the variables of the outer integral and the overall iteration process, ξ be the lower bound of the alignment function R to define the alignment start interval, η be the upper bound of the alignment function R to define the alignment end interval, L be the local processing result function mapping, and E be the edge node return result function mapping. For time-aligned indexing The resulting time independent variable For spatial alignment indexing The resulting spatial independent variable, ∥L∥∥E∥, represents the norm product of the local and marginal results in vector form, used to scale the alignment difference; erf(·) is the error function used to filter out abnormal distributions; and Z is the final fused output value.

[0129] If all integrals and summations converge, then Z takes the real number in the interval (0, +∞).

[0130] The larger Z is, the higher the degree of superposition between the local processing results and the edge node results after fusion, the more observable the superposition is.

[0131] Conversely, if Z approaches 0, then there is a data anomaly.

[0132] The aforementioned methods for data cleaning, feature extraction, and data compression in executing local tasks can be carried out using existing technologies and methods, and will not be elaborated further in this example.

[0133] In addition to the above embodiments, other aspects of the present invention also propose a mobile terminal device data processing system based on edge computing, including: one or more processors and a memory.

[0134] The memory is used to store operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the edge computing-based mobile terminal device data processing method of the foregoing embodiments, in particular... Figure 1 The flowchart of the method is shown.

[0135] Other aspects disclosed in the embodiments of the present invention also propose a computer-readable medium for storing software including instructions executable by one or more computers, which, upon execution, cause the one or more computers to perform operations including the flow of the data processing method for a mobile terminal device based on edge computing of the foregoing embodiments, particularly... Figure 1 The flowchart of the method is shown.

[0136] It should be recognized that embodiments of the present invention may be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium.

[0137] The method can be implemented using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program in the computer program, wherein the storage medium is configured such that the computer operates in a specific and predefined manner.

[0138] Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system; however, if required, the program can be implemented in assembly or machine language.

[0139] In any case, the language can be either compiled or interpreted.

[0140] Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit.

[0141] The processes described herein (or variations and / or combinations thereof) can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0142] Furthermore, the method can be implemented in any suitable computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices.

[0143] Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether portable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein.

[0144] Furthermore, machine-readable code, or parts thereof, can be transmitted via wired or wireless networks.

[0145] When such media includes instructions or programs that combine with a microprocessor or other data processor to implement the steps described above, the invention described herein includes these and other different types of non-transitory computer-readable storage media.

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data processing method for mobile terminal devices based on edge computing, characterized in that, include: Based on task type, device resource status and network environment parameters, the mobile terminal device uses a dynamic decision model to divide the tasks to be processed, resulting in local computing tasks and edge computing tasks that need to be offloaded to edge nodes. Based on a load balancing strategy, target edge nodes are selected from the edge computing network, and a secure transmission channel is established to distribute the edge computing tasks to the target edge nodes through the secure transmission channel. Specifically, the edge computing network's status monitoring interface is invoked to obtain load information and network environment parameters for each candidate edge node. Based on a preset weighting factor, the CPU utilization, memory utilization, available bandwidth, network latency, and packet loss rate of the candidate edge nodes are comprehensively scored. Several target edge nodes whose scores meet preset thresholds are selected, and a secure handshake process is initiated between the mobile terminal device and the target edge nodes to negotiate encryption suites and keys. After completing digital certificate verification, a secure session state is entered, and an encrypted data transmission channel is established. According to the task type and data size, the edge computing tasks to be distributed are packaged and divided into blocks, and the block data is digitally signed. Each block data is sent to the target edge node through the secure transmission channel. If no confirmation response is received within a preset time, a retransmission operation is performed. When the target edge node confirms that the reception is correct, the edge computing task is added to its execution queue for computation processing. A multi-core heterogeneous computing resource allocation strategy is implemented for the local computing task, and a dynamic memory partitioning mechanism is used for parallel processing, while data cleaning, feature extraction, and data compression operations are performed simultaneously. Specifically, the CPU multi-core, GPU, and NPU computing units within the mobile terminal device are scanned to form a resource description list. The local computing task is decomposed into multiple subtasks, and the execution cost of the subtasks on different computing units is calculated using a resource scheduling function based on the computational characteristics and priorities of the subtasks. Based on the execution cost, a scheduling algorithm is used to map the subtasks to the corresponding computing units. If a load spike is detected during execution, an online update mechanism is triggered. The dynamic memory partitioning mechanism includes: initially dividing the memory into multiple partitions and reserving a video memory partition; when the usage of a certain memory partition is close to the upper limit, checking the free status of other partitions and merging some free address space into the currently needed partition; when the overall memory space is insufficient, triggering a data compression operation to alleviate memory pressure; after the subtask is completed, releasing the corresponding memory partition and updating the memory usage table. The system receives the calculation results returned by the target edge node, aligns the calculation results with the local processing results in the time and space dimensions, performs data fusion processing, and outputs the final processing result. The data fusion processing includes: receiving the calculation results returned by the target edge node under a secure transmission channel, and verifying the checksum and digital signature; if the verification is successful, recording the timestamp and spatial index of the edge node result, and aligning it with the local processing result in a unified coordinate system in terms of time and space; performing data fusion operations on the aligned local processing result and the edge node result, and outputting the final fused result; the data fusion operation is implemented through a constructed data fusion model, the mathematical expression of which is: Where A is the start time of the actual data collection, which is the lower limit of the external integration, and B is the end time of the actual data collection, which is the upper limit of the external integration. To sum the results and use kernel indices from one to δ, where δ is the upper limit of the summation and the fusion depth threshold, Let Λ be the variable phase offset factor for the aligned data, Λ be the reduction coefficient in the denominator of the fusion model, μ be the starting spatial scale of the data filtering corresponding to the lower bound of the embedded integral, ν be the ending spatial scale of the data filtering corresponding to the upper bound of the embedded integral, Φ be the variable-time-space mapping of the embedded integral, X be the coordinates of the variables of the outer integral and the overall iteration process, ξ be the lower bound of the alignment function R to define the alignment start interval, η be the upper bound of the alignment function R to define the alignment end interval, L be the local processing result function mapping, and E be the edge node return result function mapping. For time-aligned indexing The resulting time independent variable For spatial alignment indexing The resulting spatial independent variables The norm product of the local and edge results in vector form is used to scale the alignment difference, erf(·) is the error function used to filter out abnormal distributions, and Z is the final fused output value; If all integrals and summations converge, then Z takes the real number in the interval (0, +∞). The larger Z is, the higher the degree of superposition between the local processing result and the edge node result after fusion, the more observable the superposition is. Conversely, if Z approaches 0, then there is a data anomaly.

2. The mobile terminal device data processing method based on edge computing according to claim 1, characterized in that, The task types include real-time data processing tasks, latency-sensitive tasks, and high-computational-load tasks. The device resource status includes CPU utilization, memory utilization, and remaining battery power. The network environment parameters include the current network bandwidth, network latency, and packet loss rate.

3. The mobile terminal device data processing method based on edge computing according to claim 2, characterized in that, A dynamic decision model is used to divide local computing tasks into edge computing tasks that need to be offloaded to edge nodes, including: Obtain the task type of the task to be processed, and assign a numerical label to the task type; Obtain the device resource status of the mobile terminal device, including CPU utilization, memory utilization, and remaining battery power; Detect network environment parameters, including current network bandwidth, network latency, and packet loss rate; The task type, the device resource status, and the network environment parameters are preprocessed and normalized. An interpolation function is established to calculate the difference in overall cost between executing a task locally and executing it on an edge node. When the difference is greater than zero, the task is classified as an edge computing task, and when the difference is less than or equal to zero, the task is classified as a local computing task.

4. A mobile terminal device data processing system based on edge computing, characterized in that, include: One or more processors; The memory stores operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, including the flow of the edge computing-based mobile terminal device data processing method as described in any one of claims 1 to 3.

5. A computer-readable medium for storing software, characterized in that: The software includes instructions executable by one or more computers, which cause the one or more computers to perform operations, including the flow of the edge computing-based mobile terminal device data processing method as described in any one of claims 1 to 3.

Images