A task allocation method for medical Internet of Things based on information perception and priority determination

Abstract

The present invention relates to a method for allocating tasks of a medical Internet of Things based on information perception and priority determination, and belongs to the technical field of Internet of Things task allocation. The present invention includes the following steps: system model construction, including: information collection, cluster division, cluster head determination; equipment performance monitoring, including: query of the cluster where the task is located, query of cluster heads in the cluster, cluster head collection of node status information, cluster node level determination; feasible scheduling search, including: query of the lowest level node in the cluster, judgment of the number of query nodes, query of the node with the least number of allocations; task issuance and execution, including: transmitting the task from the receiving node to the execution node; task execution; feedback of the result to the front-end interface. The present invention can make intelligent decisions based on real-time status information, so that tasks can be optimally allocated between edge nodes, minimize task waiting time, and improve the response speed and overall efficiency of the system.

Description

Technical Field

[0001] The invention relates to a medical Internet of Things task allocation method based on information perception and priority determination, and belongs to the technical field of Internet of Things task allocation. Background Art

[0002] IoT devices for medical applications require more data storage and intensive data processing. Relying on the powerful computing power of cloud servers, intensive tasks can be processed at a lower computing cost. At the same time, cloud computing also has the advantages of high scalability, low cost and abundant resources. However, the terminal device is far away from the cloud server, which not only generates additional communication overhead, but also easily leads to high latency service response time, making it difficult to meet the needs of delay-sensitive applications, thereby reducing user experience. Faced with problems such as large-scale data transmission, centralized data storage, and inflexible response to real-time interactive applications, the traditional cloud computing processing model will face difficult bottlenecks and pressures, and edge computing has emerged. Edge computing, also known as a "micro data center", provides users with a diversified service that integrates distributed computing, storage, network, and applications by deploying edge nodes close to the data source, meeting the basic needs of large data volume and real-time health monitoring, and reducing the response time of network services. In addition, edge computing is close to the terminal, and most computing tasks do not need to be uploaded to the data center, which improves the security of user data to a certain extent. In the medical and health monitoring scenario under the edge computing environment, considering that the patient vital sign monitoring task is highly sensitive to delays, the edge task allocation center needs to schedule the unloaded tasks in real time and allocate reasonable resources. The types, priorities, data volumes and other characteristics of different tasks are highly different, and this information will only be revealed after the unloading request arrives. At present, most task allocation solutions prioritize tasks by considering their urgency. These technologies have their own limitations and are not applicable to the heterogeneity of edge devices in the medical Internet of Things.

[0003] The prior art closest to the present invention includes:

[0004] (1) Multi-stage greedy adjustment method

[0005] The multi-stage greedy adjustment method is divided into three key stages, the first is the initial schedule creation stage, the second is the network flow conflict detection stage, and the last is to resolve the network flow conflict by adjusting the task location and the bandwidth of the flow. In the first stage, the initial schedule is created by not considering the network congestion. This can use different algorithms, such as evolutionary algorithms (e.g., genetic algorithms), not limited to greedy algorithms, to generate diverse initial schedules. The second stage involves detecting network flow conflicts, that is, identifying conflicts caused by task location and flow bandwidth. This step is to find areas that need to be adjusted to resolve potential conflict issues. The last stage is to resolve network flow conflicts and optimize the schedule by adjusting task location and flow bandwidth. This process allows different strategies to be adopted to find the optimal or better solution.

[0006] Advantages: The algorithm is highly flexible and some parts of the algorithm can be replaced to generate different solutions. In addition, it provides multiple strategies for network flow conflicts, increasing the diversity of problem solving.

[0007] Disadvantages: The computational complexity is high for large-scale problems, and in some cases it may not be possible to find the global optimal solution. In addition, its performance may be affected by algorithm selection and parameter settings.

[0008] (2) Multi-objective task allocation method

[0009] The multi-objective task allocation method is divided into four steps. In the first stage, based on the NSGAII algorithm, multiple objective functions are optimized, which enables the algorithm to consider multiple objectives in task allocation and improves the flexibility of the algorithm. The second stage is to continuously improve the task allocation and continuously improve the possible task allocation to make it more in line with multiple optimization objectives. Through iteration and evolution, the algorithm can gradually approach the global optimal solution. The third stage is to reallocate tasks when necessary, such as in the case of node failure. This dynamic adjustment enables the algorithm to be more adaptable to changes in the actual network environment. The last stage is the secondary goal of improving latency, which takes improving latency as a secondary goal and provides improved latency performance throughout the network.

[0010] Advantages: Multi-objective optimization, considering multiple objectives, making it more adaptable to complex task allocation problems; dynamic adaptability, the algorithm has dynamic adaptability and can reallocate tasks when the network changes, which improves robustness; overall network performance improvement, through the setting of secondary objectives, it provides improved performance in the entire network, not just focusing on a single objective.

[0011] Disadvantages: High computational complexity. The algorithm may have high computational complexity on large-scale problems. Sensitive to parameter settings. The algorithm performance may be affected by the parameter settings and requires careful adjustment of parameters to obtain the best results. Local optimal solution. Due to the complexity of multi-objective problems, the algorithm may fall into a local optimal solution and fail to find the global optimal solution.

[0012] (3) Integer Linear Programming Method

[0013] The basic steps of the integer linear programming method include setting up linear constraints and objective functions, ensuring that the variables of the problem need to take integer values, and then using a specialized optimization solver to find the optimal integer solution or the integer solution that is closest to the optimal solution.

[0014] Advantages: Handling integer constraints, the algorithm is specifically designed to handle integer variables and is suitable for problems that require integer solutions; ensuring feasibility, the integer linear programming solver can ensure that an integer solution that satisfies the constraints is found, providing a feasible solution.

[0015] Disadvantages: When solving large-scale problems, the algorithm may face challenges in computational complexity and take a long time to solve; it has high requirements for computing resources, especially when dealing with large-scale problems, the algorithm may require a lot of computing resources; the solver may find the integer solution closest to the optimal solution instead of the true optimal integer solution.

[0016] In view of this, the present invention proposes a task allocation method based on information perception and priority determination, which constructs a cluster network architecture through task information and status information, and monitors the performance level of hospital workstations in the same cluster in real time through cluster heads, and assigns different priorities to different nodes. In the constructed architecture, data generated by various sensors / devices attached to patients are transmitted to hospital workstations deployed in the hospital. The method assigns different priorities to different nodes based on the task information and status information in the cluster and considering the performance of the hospital workstations. Summary of the invention

[0017] The present invention provides a method for allocating tasks in the medical Internet of Things based on information perception and priority determination, so as to solve the problem of optimizing completion delay under the conditions of heterogeneous edge nodes and task diversity in the medical Internet of Things. The present invention can make intelligent decisions based on real-time status information, so that tasks can be optimally allocated among edge nodes, thereby minimizing task waiting time and improving the response speed and overall efficiency of the system.

[0018] The technical solution of the present invention is: a method for allocating tasks of the medical Internet of Things based on information perception and priority determination, and the specific steps of the method are as follows:

[0019] Step 1, system model construction; Step 1 includes: information collection, cluster division, cluster head determination;

[0020] Step 2, equipment performance monitoring; Step 2 includes: querying the cluster where the task is located, querying the cluster head in the cluster, collecting node status information of the cluster head, and determining the level of cluster nodes;

[0021] Step 3, feasible scheduling search; Step 3 includes: querying the lowest level node in the cluster, determining the number of query nodes, and querying the node with the least number of allocations;

[0022] Step 4: Task distribution and execution; Step 4 includes:

[0023] Step 4.1, transfer the task from the receiving node to the execution node;

[0024] Step 4.2, task execution;

[0025] Step 4.3, feed back the results to the front-end interface.

[0026] Furthermore, the cluster node level determination in Step 2.4 includes: the cluster head node collects necessary performance data by accessing the node status module database; based on the capability level evaluation formula (1), a distributed analysis is performed on each node in the cluster to calculate the capability level score of each node; further, the capability level rule formula (2) is used to clearly determine the specific capability level of each node; finally, the evaluation result is accurately recorded in the capability level module database;

[0027]

[0028]

[0029] Among them, α, β, γ and δ represent the weights of different performance parameters, ac_DT i represents the capability value of edge node i, l i Indicates the capability level of edge node i; Indicates the capacity level, including low pressure, standard, high pressure and overpressure, that is, and

[0030] The input of cluster node level determination includes the task graph of M tasks, the network of N edge devices and the input data Dti of each task i in the task graph. In the initial stage of each cluster, an edge device will be randomly selected as the manager of the cluster. i , so that the cluster M cThe managed nodes will be responsible for task access and processing; the managed nodes will calculate their CPU utilization N in real time. i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS , and this type of information is saved to the corresponding node information SE i Finally, it is sent to the cluster i , cluster i According to SE i The capability information in the node is the capability value of the node ac_DT i Make a judgment.

[0031] Furthermore, the specific steps of Step 1 are as follows:

[0032] Step 1.1, Information collection: Store the device information and status information of medical devices and edge nodes into the semantic information module database by reading JSON text;

[0033] Step 1.2, cluster division: Based on the information and quantity relationship of medical devices and edge nodes in the system database, with the goal of maximizing average performance, the K-means clustering algorithm is used to perform cluster division, and the cluster information is stored in the cluster information module database;

[0034] Step 1.3, cluster head determination: The cluster head is determined by random sampling, that is, by generating a random index within the cluster range, then determining the corresponding cluster head from the cluster, and storing the cluster head information in the cluster information module database; the role of the cluster head is to manage the remaining nodes in the cluster.

[0035] Furthermore, the specific steps of Step 2 are as follows:

[0036] Step 2.1, task cluster query: when a task is generated at terminal device i, the semantic information module database is searched through the index to obtain the edge node information connected to terminal device i, and the cluster information of the edge node is retrieved from the cluster information module database;

[0037] Step 2.2, cluster head query in the cluster: use the cluster information module database to obtain the cluster head information of the cluster;

[0038] Step 2.3, cluster head collects node status information: according to the cluster head information, it sends monitoring instructions to the cluster head, collects the status information of other nodes in the cluster in real time, and stores it in the node status module database; this status information includes CPU, memory, network and disk status; the managed node calculates its CPU utilization N in real time i,CPU、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS ;

[0039] Step 2.4, cluster node level determination.

[0040] Furthermore, the specific steps of Step 3 are as follows:

[0041] Step 3.1, query the lowest level node in the cluster: the cluster head uses the index query method to identify the node information with the lowest capability level through the capability level module database;

[0042] Step 3.2, query the number of nodes to determine; when multiple nodes have the same capability level, the cluster head will execute Step 3.3;

[0043] Step 3.3, query the node with the least number of assignments; the cluster head will further retrieve the task assignment frequency of the node from the task assignment frequency module database, and give priority to the node with the least number of assignments for task assignment. If the number of task assignments of these nodes is still the same, the node closest to the cluster head will be further selected to ensure the efficiency and response speed of task assignment.

[0044] The beneficial effects of the present invention are:

[0045] 1. This invention builds a system model based on information perception to solve the problem of low response time in the medical Internet of Things scenario. In this technical framework, by dividing clusters and adjusting the location of task execution and other key steps, the number of potential scheduling is successfully reduced, thereby significantly improving the speed of algorithm response;

[0046] 2. The present invention establishes a priority standard for heterogeneous environments. The benefit of this priority allocation is that, in view of the heterogeneity of different nodes, reasonable priority allocation can more effectively schedule tasks and improve overall system performance. By considering the performance of hospital workstations, the algorithm can make intelligent decisions based on real-time status information, so that tasks can be optimally allocated among edge nodes, minimize task waiting time, and improve the response speed and overall efficiency of the system. Therefore, setting priorities can better adapt to the characteristics of heterogeneous edge nodes, realize the intelligence and personalization of task allocation, and provide an effective means for the efficient operation of the medical Internet of Things system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a system model diagram in the present invention;

[0048] Figure 2 It is a flow chart of the present invention;

[0049] Figure 3 This is a performance result diagram of the present invention. DETAILED DESCRIPTION

[0050] Example 1: Figure 1-Figure 3 As shown, a method for allocating tasks of the medical Internet of Things based on information perception and priority determination is provided, and the specific steps of the method are as follows:

[0051] Step 1, system model construction; the specific steps of Step 1 are as follows:

[0052] Step 1.1, Information collection: Store the device information and status information of medical devices and edge nodes into the semantic information module database by reading JSON text;

[0053] Step 1.2, cluster division: Based on the information and quantity relationship of medical devices and edge nodes in the system database, with the goal of maximizing average performance, the K-means clustering algorithm is used to perform cluster division, and the cluster information is stored in the cluster information module database;

[0054] Step 1.3, cluster head determination: The cluster head is determined by random sampling, that is, by generating a random index within the cluster range, then determining the corresponding cluster head from the cluster, and storing the cluster head information in the cluster information module database; the role of the cluster head is to manage the remaining nodes in the cluster.

[0055] Step 2, equipment performance monitoring; the specific steps of Step 2 are as follows:

[0056] Step 2.1, task cluster query: when a task is generated at terminal device i, the semantic information module database is searched through the index to obtain the edge node information connected to terminal device i, and the cluster information of the edge node is retrieved from the cluster information module database;

[0057] Step 2.2, cluster head query in the cluster: use the cluster information module database to obtain the cluster head information of the cluster;

[0058] Step 2.3, cluster head collects node status information: according to the cluster head information, it sends monitoring instructions to the cluster head, collects the status information of other nodes in the cluster in real time, and stores it in the node status module database; this status information includes CPU, memory, network and disk status; the managed node calculates its CPU utilization N in real time i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS ;

[0059] Step 2.4, cluster node level determination. The cluster node level determination in Step 2.4 includes: the cluster head node collects necessary performance data by accessing the node status module database; based on the capability level evaluation formula (1), a distributed analysis is performed on each node in the cluster to calculate the capability level score of each node; further, the capability level rule formula (2) is used to clearly determine the specific capability level of each node; finally, the evaluation results are accurately recorded in the capability level module database;

[0060]

[0061]

[0062] Among them, α, β, γ and δ represent the weights of different performance parameters, ac_DT i represents the capability value of edge node i, l i Indicates the capability level of edge node i; Indicates the capacity level, including low pressure, standard, high pressure and overpressure, that is,

[0063]

[0064] The input of cluster node level determination includes the task graph of M tasks, the network of N edge devices and the input data Dti of each task i in the task graph. In the initial stage of each cluster, an edge device will be randomly selected as the manager of the cluster. i , so that the cluster M c The managed nodes will be responsible for task access and processing; the managed nodes will calculate their CPU utilization N in real time. i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS , and this type of information is saved to the corresponding node information SE i Finally, it is sent to the cluster i , cluster i According to SE i The capability information in the node is the capability value of the node ac_DT i Make a judgment.

[0065] Step 3: Feasible scheduling search; the specific steps of Step 3 are as follows:

[0066] Step 3.1, query the lowest level node in the cluster: the cluster head uses the index query method to identify the node information with the lowest capability level through the capability level module database;

[0067] Step 3.2, query the number of nodes to determine; when multiple nodes have the same capability level, the cluster head will execute Step 3.3;

[0068] Step 3.3, query the node with the least number of assignments; the cluster head will further retrieve the task assignment frequency of the node from the task assignment frequency module database, and give priority to the node with the least number of assignments for task assignment. If the number of task assignments of these nodes is still the same, the node closest to the cluster head will be further selected to ensure the efficiency and response speed of task assignment.

[0069] Step 4: Task distribution and execution; Step 4 includes:

[0070] Step 4.1, transfer the task from the receiving node to the execution node;

[0071] Step 4.2, task execution;

[0072] Step 4.3, feed back the results to the front-end interface.

[0073] The system model proposed by the present invention is shown in the attached Figure 1 As shown in the figure, it includes a three-layer architecture: sensor node layer (terminal device), gateway layer (edge ​​device) and cloud layer. Assume that the terminal device layer is composed of N IoT devices, each of which has a computing task to be processed. Different tasks have the same urgency and the same maximum service tolerance delay. Due to the limited computing resources, the computing power of the local terminal device may not be able to meet the basic requirements of the task. In order to improve the task processing efficiency and user service quality, the task is migrated to the edge node through the wireless transmission channel for processing.

[0074] Different terminal devices are used to monitor and measure the patient's vital signs. Once the required data is measured, it sends the data to the connected edge node. The system model comprehensively considers the location distribution and attribute types of terminal devices and edge devices;

[0075] When cluster division and cluster head determination: the gateway layer is divided into multiple clusters, where each cluster has an edge device as a manager, responsible for collecting context information and executing corresponding decision allocations; other edge devices can act as task initiators or executors to interact with data and execute tasks.

[0076] During feasible scheduling search, task issuance and execution: the gateway layer deploys multiple workstations (edge ​​nodes) to provide services for the received tasks. The edge nodes make the best migration decisions based on the request information sent by the IoT device i. For tasks that need to be migrated, the edge nodes upload the task information to the cluster head node. The cluster head node migrates the task to the best edge node for execution, so that it allocates the best bandwidth resources and computing resources to minimize the task completion time.

[0077] In a real medical-patient monitoring scenario, the method proposed in the present invention is compared with several existing methods to verify the feasibility of the present invention. A large number of experimental results show that compared with traditional solutions, the method proposed in the present invention shortens the task completion time by an average of 37%. Figure 3 All evaluations in the present invention use binary text data of hundreds of patients' vital signs collected by medical equipment including monitors, anesthesia machines, ventilators, etc. as real data sets, and simulate and expand the binary text data according to the data protocol structure of the medical equipment to make it a synthetic data set;

[0078] Table 1 shows the performance of different system model schemes when 60 medical terminal devices generate 9,000 tasks. It is worth noting that when the system model scheme proposed in this study is used, the average performance of the algorithm is significantly improved, which is 22.3% higher than that of the random initialization system;

[0079] Table 1: Effect comparison table Unit: seconds

[0080]

[0081] The specific implementation modes of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above implementation modes, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A method for allocating tasks in the medical Internet of Things based on information perception and priority determination, characterized in that: The specific steps of the method are as follows: Step 1, system model construction; Step 1 includes: information collection, cluster division, cluster head determination; Step 2, equipment performance monitoring; Step 2 includes: querying the cluster where the task is located, querying the cluster head in the cluster, collecting node status information of the cluster head, and determining the level of cluster nodes; Step 3, feasible scheduling search; Step 3 includes: querying the lowest level node in the cluster, determining the number of query nodes, and querying the node with the least number of allocations; Step 4: Task distribution and execution; Step 4 includes: Step 4.1, transfer the task from the receiving node to the execution node; Step 4.2, task execution; Step 4.3, Feedback the results to the front-end interface; The specific steps of Step 3 are as follows: Step 3.1, query the lowest level node in the cluster: the cluster head uses the index query method to identify the node information with the lowest capability level through the capability level module database; Step 3.2, query the number of nodes to determine; when multiple nodes have the same capability level, the cluster head will execute Step 3.3; Step 3.3, query the node with the least number of assignments; the cluster head will further retrieve the task assignment frequency of the node from the task assignment frequency module database, and give priority to the node with the least number of assignments for task assignment. If the number of task assignments of these nodes is still the same, the node closest to the cluster head will be further selected to ensure the efficiency and response speed of task assignment.

2. The method for allocating tasks in the medical Internet of Things based on information perception and priority determination according to claim 1 is characterized in that: The cluster node level determination in Step 2 includes: the cluster head node collects necessary performance data by accessing the node status module database; based on the capability level evaluation formula (1), a distributed analysis is performed on each node in the cluster to calculate the capability level score of each node; further, the capability level rule formula (2) is used to clearly determine the specific capability level of each node; finally, the evaluation result is accurately recorded in the capability level module database; Among them, α, β, γ and δ represent the weights of different performance parameters, ac_DT i represents the capability value of edge node i, l i Indicates the capability level of edge node i; Indicates the capacity level, including low pressure, standard, high pressure and overpressure, that is, and The input of cluster node level determination includes the task graph of M tasks, the network of N edge devices and the input data Dti of each task i in the task graph. In the initial stage of each cluster, an edge device will be randomly selected as the manager of the cluster. i , so that the cluster M c The managed nodes will be responsible for task access and processing; the managed nodes will calculate their CPU utilization N in real time. i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS , and the CPU utilization N i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS Save to the corresponding node information SE i Finally, it is sent to the cluster i , cluster i According to SE i The capability information in the node is the capability value of the node ac_DT i Make a judgment.

3. The method for allocating tasks in the medical Internet of Things based on information perception and priority determination according to claim 1 is characterized in that: The specific steps of Step 1 are as follows: Step 1.1, Information collection: Store the device information and status information of medical devices and edge nodes into the semantic information module database by reading JSON text; Step 1.2, cluster division: Based on the information and quantity relationship of medical devices and edge nodes in the system database, with the goal of maximizing average performance, the K-means clustering algorithm is used to perform cluster division, and the cluster information is stored in the cluster information module database; Step 1.3, cluster head determination: The cluster head is determined by random sampling, that is, by generating a random index within the cluster range, then determining the corresponding cluster head from the cluster, and storing the cluster head information in the cluster information module database; the role of the cluster head is to manage the remaining nodes in the cluster.

4. The method for allocating tasks in the medical Internet of Things based on information perception and priority determination according to claim 1 is characterized in that: The specific steps of Step 2 are as follows: Step 2.1, task cluster query: when a task is generated at terminal device i, the semantic information module database is searched through the index to obtain the edge node information connected to terminal device i, and the cluster information of the edge node is retrieved from the cluster information module database; Step 2.2, cluster head query in the cluster: use the cluster information module database to obtain the cluster head information of the cluster; Step 2.3, cluster head collects node status information: according to the cluster head information, it sends monitoring instructions to the cluster head, collects the status information of other nodes in the cluster in real time, and stores it in the node status module database; this status information includes CPU, memory, network and disk status; the managed node calculates its CPU utilization N in real time i,CPU 、Memory utilization N i,RAM , Network utilization N i,NIC and disk utilization N i,DIS ; Step 2.4, cluster node level determination.

Images