Intention-driven task transmission path selection method and device in uncertain environment

Abstract

The invention discloses an intention-driven task transmission path selection method and intention-driven task transmission path selection equipment in an uncertain environment, and the intention-driven task transmission path selection method in the uncertain environment comprises the following steps: when a source node has a task unloading demand, sending a message containing task data to a destination node, according to the message containing the task data, task transmission path selection is achieved through a task transmission path selection model, and an optimal task transmission path is obtained. The invention discloses an intention-driven task transmission path selection method and intention-driven task transmission path selection equipment in an uncertain environment. A task transmission and execution algorithm with deterministic task processing time delay constraint in an uncertain environment can solve the problem of task processing in an emergency, a scene can be modeled into a pervasive edge computing environment composed of a plurality of heterogeneous devices, and the devices can be mutually connected through wireless communication.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and device for selecting an intent-driven task transmission path in an uncertain environment. Background Art

[0002] At present, a large number of growing physical and virtual devices and the network entities they are in need to communicate with each other, making many IoT applications diverse and complex. Moreover, these applications require ubiquitous interconnection of heterogeneous devices and low latency task processing. In order for the network to achieve this application goal based on swarm intelligence, it is necessary to solve the problem of business object identification such as IP, identity, and geographic location, especially content identification, service identification, and multi-dimensional resource identification, so as to effectively support the interconnection and information exchange needs of heterogeneous entities. However, the current network structure lacks elasticity and IP single bearer, making it difficult to meet diverse addressing needs. This is a basic problem facing ubiquitous edge computing.

[0003] Pervasive edge computing is a new type of distributed computing. It has no central authorization management. If necessary, tasks can be migrated to nearby devices with more abundant resources for execution. Data storage, processing and scheduling decisions are all performed at the edge of the network, which can significantly reduce transmission delays and improve the user experience quality. However, due to the lack of a centralized management mechanism, it is difficult to obtain the global status information of the entire network. At the same time, due to the mobility of the device, the active devices in different time slots are uncertain, and the number of task migrations is also uncertain. For example, the field battlefield environment lacking infrastructure is also a working scenario of the research content. In this scenario, a marine detachment operating in the field is equipped with a variety of heterogeneous devices, such as smart wearables, tanks, infantry fighting vehicles, etc., and forms a mobile ad hoc network (MANET) with tanks as the local command center during movement. However, in an emergency, important tasks perceived by smart wearables are required to be transmitted to the intent-driven target device, such as the tank command center, within a limited time under local information conditions. However, due to the mobility of the device, the network topology is dynamic and unpredictable. At the same time, due to the lack of communication infrastructure, devices can only rely on their own equipment for short-range wireless communication, and their communication range is limited. Mobile nodes need the assistance of other nodes when forwarding data packets to find the best path to the destination node.

[0004] Traditional routing protocols select the next hop node according to the number of hops when the router forwards the data packet, so as to reduce the network delay and packet loss rate and improve the service quality. However, when the computing resources of the entire network are relatively scarce, a node will have the need to offload tasks and transmit its own task data packet to the corresponding destination node. At the same time, due to the heterogeneity of devices and the existence of the business intent of the task, the transmission of information needs to solve the object identification problem. The business intent is a specific goal that is expected to be achieved. It is the processing result of several interrelated tasks. In this case, on the one hand, it is necessary to understand the business intent of the task through object identification and select a suitable transmission path. On the other hand, the destination device must complete the execution of the task within a limited delay. Although there are many routing algorithms suitable for mobile ad hoc networks (MANETs), they are not suitable for the above situation. Therefore, it is urgent to develop a task transmission and execution algorithm with deterministic task processing delay constraints in an uncertain environment to achieve the purpose of solving task processing in emergency situations. The scenario can be modeled as a ubiquitous edge computing environment composed of multiple heterogeneous devices, and the devices can be connected to each other through wireless communication. Summary of the invention

[0005] The present invention overcomes the shortcomings of the prior art and provides an intent-driven task transmission path selection method and device in an uncertain environment; a task transmission and execution algorithm with deterministic task processing delay constraints in an uncertain environment can solve the problem of task processing in emergency situations, and can model the scene as a ubiquitous edge computing environment composed of multiple heterogeneous devices, and the devices can be connected to each other through wireless communication.

[0006] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: an intent-driven task transmission path selection method in an uncertain environment, comprising the following steps: when the source node has a task offloading demand, it sends a message containing task data to the destination node, and the source node implements task transmission path selection according to the message containing task data through a task transmission path selection model to obtain the preferred task transmission path; the acquisition of the message containing task data includes the following steps: according to the task data, an object identifier Mult-IP is established based on an extended header of the IPv6 protocol, and the object identifier Mult-IP is mapped to the IPv6 address of the network layer, and is compatible with IP; the identification and management of objects are realized at the network layer; the task transmission path selection model adopts an edge ant colony routing protocol through an ant colony algorithm with the help of the Mult-IP identification message to dynamically optimize the transmission of task data packets, and find the best route to the destination as the preferred task transmission path.

[0007] In a preferred solution of the present invention, the fields of the object identifier Mult-IP include a Function field set and an M-data field. The Function field set is used to illustrate the specific meaning of the M-data field set, and the M-data field set includes parameter values.

[0008] In a preferred embodiment of the present invention, the object identifier Mult-IP implements the configuration classification of various network functions according to the semantic information of different types of requirements of mobile devices, including: task identifier T-uid message, device identifier D-uid message, service identifier S-uid message and content identifier C-uid message.

[0009] In a preferred solution of the present invention, the task identification T-uid message is in an environment without infrastructure. When a mobile device generates a task and cannot complete the execution of the task due to insufficient resources, it will request the surrounding devices to unload the task. At this time, the task message transmitted by the requesting device will contain the corresponding task identification T-uid message. The task identification of the task identification T-uid message records the information related to the computing task; the T-uid message includes , where tid is the unique identifier of each task, At is the input size of each task, Ct is the number of clock cycles of the task, St is the type of task, TDDL is the deadline of the task, Ft is the computing power required for the task, Bt is the required bandwidth for the task, Mt is the memory required for the task, Timestamp is the timestamp generated by the task, and the task is abstractly represented as a nine-tuple in the system: ; and / or, the device identification D-uid message is the relevant information of the node in the network, which is used to allow the network to perceive the surrounding environment information in a fine-grained manner for resource sharing; the device identification D-uid adopts the Mult-IP extended header structure, including , where nid is the unique identification number of the node, Xt is the longitude of the node, Yt is the latitude of the node, Vt is the speed of the node, Ft is the calculation rate of the node, Et is the remaining energy of the node, Et is the remaining memory of the node, Mt is the remaining memory of the node, and Lt is the type of the node; and / or, the service identification S-uid message is used by mobile devices to express their service needs and quickly locate the target node that can provide services. The service identification S-uid message includes , where tid is the unique identifier of each task, operation is the operation of the service, and service is the service information body; and / or, the content identifier C-uid message is that when the source node is performing task unloading or executing a service, it needs to preload some static resources, then the static resource information required by the source node is added to the Mult-IP extension header to form a content identifier C-uid. The content identifier C-uid message includes tid, cid and K1, K2, K3, K4, K5, where tid is the unique identifier of each task, cid is the type of content, K1 is the keyword 1 corresponding to the content, K2 is the keyword 2 corresponding to the content, K3 is the keyword 3 corresponding to the content, K4 is the keyword 4 corresponding to the content, and K5 is the keyword 5 corresponding to the content.

[0010] In a preferred embodiment of the present invention, the ant colony algorithm includes: determining the execution location of the computing task before the task data starts to be transmitted; for a vehicle node with more resources, periodically broadcasting its device information to surrounding nodes; the broadcasted task data is used as a task data packet, and the task data packet contains a D-uid header related to the device information; the D-uid header contains the node identification number nid, the node location information , the current speed of the node V t And its device hardware information ; When the source node receives the task data packet, it will parse the D-uid information from the task data packet and select a suitable node from the D-uid information as the unloading destination node.

[0011] In a preferred embodiment of the present invention, the ant colony algorithm includes: setting the calculation rate of vehicle d to , the remaining memory is , the remaining energy is , then node d should satisfy the condition: ; ; ; At this time, the source node s will randomly select one of the vehicle nodes that meet the conditions as the destination node d for unloading; when the source node s is ready to send a data packet to the destination node d, it will look for the current corresponding routing table entry. If the routing table entry of the destination node d is not found in the routing table, the route discovery process will be started; in the route discovery process, it is necessary to find a transmission path to help the task load the required static resources; when the route discovery starts, the source node s will generate a forward ant packet, and a service identifier S-uid will be attached to the header of the forward ant packet; after the intermediate node receives the header of the forward ant packet, it will parse the content of the service identifier S-uid on the header of the forward ant packet, and match it with the C-uid contained in the C-uid information table, filter out the neighboring nodes with the corresponding content, and forward the packet according to the pheromone concentration on the link of the filtered neighboring nodes; use represents the pheromone concentration on the link from intermediate node i to destination node d via neighbor node n. The probability of intermediate node i selecting next hop node n is expressed as: ;in, It is the set of neighbor nodes that reach the destination node d and contain corresponding static resources recorded in the routing table of node i; is the exploration parameter value used to control the exploration behavior of the ant packet header; for the header of the ant packet from the same source node s, the intermediate node i will only process the header of the forward ant packet received for the first time, and the headers of the ant packets received later will be killed.

[0012] In a preferred solution of the present invention, the ant colony algorithm includes: when the destination node d receives the header of the forward ant packet from different neighboring nodes, it will unicast the header of the backward ant packet to different neighboring nodes in turn; after the intermediate node i receives the header of the backward ant packet, it will calculate the estimated value of the delay to the destination node d by increment: ; Where n is the number of nodes on the path from intermediate node i to destination node d; The estimated value of the delay for sending a data packet can be expressed as: ;in, is the number of packets in the current MAC layer queue, It represents the transmission time required for node i to successfully send a message, which is the average time from when the data packet reaches the MAC layer to when the transmission is successfully completed. It is expressed as: ;in, ; In order to reflect the changes in network topology, the intermediate node i will update or create pheromones in the routing entry; the value of the pheromone is determined by two parameters: the update operator and the volatility factor. The update operator It is related to the number of hops of the path and the estimated delay, and is defined as: ;in is the average delay of each hop, is the bandwidth between the intermediate node i and the neighbor node j; using the update operator and volatility factor , the update rules of the pheromone value in the routing entry are as follows: ; The source node s will observe whether the backward ant returns to the source node. If no backward ant arrives after a certain period of time, the route discovery process will be restarted; when the number of restarts of the route discovery process exceeds the predetermined maximum number of times, the source node discards the cached data packet.

[0013] In a preferred solution of the present invention, different types of services correspond to different types of service intentions. For computing-intensive services, a deadline is set and the services are required to be completed within the specified deadline. represents the total deadline of task t; when the total delay to complete task t exceeds the deadline , the task offloading is considered failed; the total delay of task t includes the calculation delay and transmission delay, where the calculation delay is expressed as: Among them, C t is the clock cycle number F of task t t is the computation rate of the target node d; the transmission deadline of task t is defined as: ;when, , the task cannot be completed within the deadline, the source node s will cancel the task offloading request and execute task t locally; otherwise, the source node s will abstract the task information into a T-uid and attach it to the IPv6 header, and set the deadline field of T-uid to , start the transmission of the task data packet; after receiving the task data packet, the intermediate node will parse the T-uid content in the header and obtain the transmission deadline of the task And update it using the following formula: ;in For tasks that need to load static resources, the intermediate node j will match the task identifier tid field of the previously received S-uid message with the T-uid message. If the match is successful, the corresponding static resource will be loaded for the computing task, and the node will continue to forward the task data packet; if there are multiple next hops to the destination node d, the node will follow The probability obtained by the formula is randomly forwarded. At this time, a higher value, making the forwarding strategy more greedy; when , then the task transmission time at this time has exceeded the total deadline , task offloading fails; then the node will discard the task data packet and send a task offloading failure message to the source node; after receiving the notification, the source node s will cancel the task offloading request and execute task t locally.

[0014] A preferred solution of the present invention also includes that in a field battlefield environment lacking infrastructure, soldiers and various vehicles form a self-organizing network to communicate with each other and jointly complete combat tasks; there are differences in resources between different devices, and edge computing is introduced into the self-organizing network to solve the problem of resource sharing and content collaboration in the battlefield environment; The realization of resource sharing and content collaboration in battlefield environments through edge computing includes the following stages: the first stage is the information collection stage, which collects the D-uid information broadcast periodically by nodes and uses these nodes as candidate nodes for future unloading; the second stage is the task unloading stage. When the wearable device generates a computing task t that it cannot complete by itself, the device will abstract the task information into a T-uid to prepare for unloading; the third stage is the service execution stage. After receiving the result data of the computing task, it is also necessary to request the corresponding service from the adjacent vehicle node. At this time, the wearable device abstracts the service into an S-uid data packet, which contains the task requested in the previous stage. , the operation of the service , the target of the operation .

[0015] In a preferred embodiment of the present invention, a task transmission path selection device based on object identification and intent-driven in an uncertain environment includes: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to implement a task transmission path selection method driven by intent in an uncertain environment.

[0016] The present invention solves the defects existing in the background technology: The present invention discloses an intent-driven task transmission path selection method and device in an uncertain environment; a task transmission and execution algorithm with deterministic task processing delay constraints in an uncertain environment can solve the problem of task processing in emergency situations, and can model the scene as a ubiquitous edge computing environment composed of multiple heterogeneous devices, and the devices can be interconnected through wireless communication.

[0017] The present invention adopts a universal object identification method (Mult-IP) to identify multiple objects; and, based on the object identification mechanism to understand the intention of the task, a path selection algorithm based on edge ant colonies (AntEdge) is proposed; and, based on the Mult-IP object identification mechanism, multiple demand information is classified and tasks are collaboratively processed, which simplifies the process of task offloading and completes the task execution process within a limited delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0019] FIG. 1 is a Mult-IP structure diagram of an intent-driven task transmission path selection method in an uncertain environment according to a preferred embodiment of the present invention.

[0020] FIG. 2 is a T-uid parameter value table of an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0021] FIG. 3 is a business type table of an intention-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0022] FIG. 4 is a T-uid example diagram of an intent-driven task transmission path selection method in an uncertain environment in a preferred embodiment of the present invention.

[0023] FIG. 5 is a D-uid parameter value table of an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0024] FIG. 6 is a table of S-uid parameter values ​​of an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0025] Figure 7 It is a C-uid parameter value table of an intent-driven task transmission path selection method in an uncertain environment in a preferred embodiment of the present invention.

[0026] Figure 8 It is a task data routing algorithm diagram of an intention-driven task transmission path selection method in an uncertain environment in a preferred embodiment of the present invention.

[0027] FIG. 9 is an unloading flow chart of an intention-driven task transmission path selection method in an uncertain environment according to a preferred embodiment of the present invention.

[0028] FIG. 10 is a task offloading process diagram of an intention-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0029] FIG. 11 is a simulation experiment parameter setting table of an intention-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0030] FIG. 12 is a graph showing task completion rates at different speeds in an intention-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0031] FIG. 13 is a graph showing average end-to-end delays at different speeds in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0032] FIG. 14 is a diagram of routing overheads at different speeds of an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0033] FIG. 15 is a graph of message delivery rates at different speeds in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0034] FIG. 16 is a graph showing task completion rates at different deadlines in an intention-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0035] FIG. 17 is a graph showing message delivery rates at different pause times in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0036] FIG. 18 is a graph of end-to-end delays of different pause times in an intent-driven task transmission path selection method under an uncertain environment according to a preferred embodiment of the present invention.

[0037] FIG. 19 is a chart of routing overheads of different pause times for an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0038] Figure 20 is a chart 1 of message delivery rates for different numbers of nodes in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0039] Fig.21 This is a chart 1 of end-to-end delays of different numbers of nodes in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0040] 22 is a second chart of message delivery rates for different numbers of nodes in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention.

[0041] Fig.23 This is a second diagram of the end-to-end delay of different numbers of nodes in an intent-driven task transmission path selection method under an uncertain environment in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention will now be further described in detail with reference to the accompanying drawings and embodiments. These drawings are simplified schematic diagrams that only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0043] In MANET, developing efficient routing protocols has always been a challenging task. Most researchers have been committed to improving the performance of routing protocols, but there are relatively few studies on the routing problem of task transmission in ubiquitous edge computing environments. For example, based on the mobility analysis of vehicles, the prior art proposes a multi-hop vehicle computing resource task routing scheme to control the connectivity of the link and the computing power of the vehicle to meet the given requirements, and proposes a semi-definite relaxation method with an adaptive adjustment program to obtain the corresponding task transmission path. Furthermore, in order to provide edge computing services to users under the cooperation of drones, in order to assign a suitable drone transmission path to each task, the literature uses a software-defined network (SDN) controller to study the problem. The controller obtains a partial solution to the problem through a greedy algorithm, and by adding task arrival order constraints, the original problem is approximated as a multi-path multi-hop task transmission problem. When the environment is static, the literature proposes a traversal method to obtain the optimal solution of the approximate problem. When the environment is dynamic, a heuristic solution based on average rate is proposed to obtain a suboptimal solution to the problem. For example, a multi-hop task routing framework based on multi-agent deep deterministic policy gradient is proposed in the prior art for vehicle-assisted collaborative edge computing, which effectively solves the dimensionality problem caused by vehicle mobility and channel variability. Furthermore, the multi-hop task routing problem in the IoT system is studied, and a new two-stage method with incomplete information (TMII) is proposed in the prior art to minimize the overall task execution cost of the actual IoT system. In the first stage, the literature estimates the offloading cost using a hierarchical estimation model to determine the execution location of the task. In the second stage, the literature schedules the effective transmission path of the computing task by constructing a distributed tree structure to reduce the transmission load and offloading delay of each link. In the development of the prior art, a greedy heuristic-based method has been extended to solve the problem of transmission path selection for task offloading in the edge computing environment. The greedy heuristic-based scheme takes into account dynamic network conditions such as latency, energy consumption, and link utilization. The above task transmission schemes all require centralized control, and cannot optimize the transmission of task data packets for the purpose of business intent. When the network state changes dynamically, it is difficult to meet the offloading needs of all users.

[0044] Recently, mobile edge computing [13,14] has become a research hotspot. Resource-constrained mobile devices can offload computing tasks to nearby edge nodes or remote cloud servers to reduce the computing pressure of the devices. There has been considerable research on the task offloading problem in edge computing. In the prior art, a joint online offloading and resource allocation (JORA) framework under MEC energy constraints is disclosed. This framework uses Lyapunov optimization to maximize the QoE of end users in the long term. The algorithm controls energy consumption by constructing energy loss queues and can solve the problem in real time. Considering the joint task offloading and resource allocation problem in multi-cellular mobile edge computing (MEC) networks, the prior art describes the problem as a mixed integer nonlinear programming (MINLP) and decouples the problem into two independent sub-problems. Finally, convex optimization techniques and heuristic algorithms are used to obtain the optimal solution and suboptimal solution of the two problems respectively. Based on the multi-user computing offloading problem of mobile edge cloud computing in a multi-channel wireless interference environment, the distributed computing offloading decision problem between mobile device users is formulated as a multi-user computing offloading game, and the structural properties of the game are analyzed. Then a distributed computing offloading algorithm is proposed, which can achieve Nash equilibrium. Based on the problem of offloading computing tasks with inseparable and delay-sensitive tasks, a distributed offloading algorithm based on the long short-term memory (LSTM) model and the deep Q network (DQN) is studied to enable mobile devices to make offloading decisions in a decentralized manner, thereby solving the load balancing problem between edge nodes and reducing the average latency of tasks. Based on the vehicle-assisted MEC paradigm, in which tasks can be offloaded to edge nodes and nearby vehicles (VENS), a Stackelberg game is established between users and MEC providers, and a gradient-based resource allocation iterative algorithm (GRAIA) is designed to obtain the optimal offloading strategy and pricing scheme. The above task offloading schemes generally rely on edge computing environments with infrastructure, which are not suitable for resource-scarce MANET environments.

[0045] The heterogeneity of devices is one of the important characteristics of MANET. Considering the service migration problem between various heterogeneous devices with different access media, mobility and connectivity. In order to continue to provide services after device failure, a reliable service exchange protocol called SFUSP is proposed. This scheme aims to use the Pervasive Discovery Protocol (PDP) to discover suitable services in a pervasive environment according to the request to find a reliable path consisting of efficient nodes with services.

[0046] In order to manage the information of heterogeneous devices in a unified manner, the present invention proposes a classification and identification scheme based on Mult-IP, which is a data label that can be stored in the IPv6 extension header and whose length can be dynamically adjusted. Mult-IP can extract and compress the demand information of the node, thereby realizing the underlying message classification and identification. Based on the intent of the task, the present invention proposes an edge ant colony routing protocol (AntEdge). The edge ant colony routing protocol (AntEdge) of the present invention dynamically optimizes the transmission of data packets by sensing the status of tasks in the network, and cooperates with Mult-IP to effectively reduce end-to-end delay and improve resource utilization.

[0047] Embodiment 1 An intent-driven task transmission path selection method in an uncertain environment comprises the following steps: when a source node has a demand for task offloading, a message containing task data is sent to a destination node, and task transmission path selection is implemented through a task transmission path selection model according to the message containing the task data to obtain a preferred task transmission path; The acquisition of the task data message includes the following steps: Figure 1 As shown in the Mult-IP structure diagram, an object identifier Mult-IP is established based on the extended header of the IPv6 protocol according to the task data, and the object identifier Mult-IP is mapped to the IPv6 address of the network layer and is compatible with IP; the identification and management of objects at the network layer are realized; the task transmission path selection model adopts the edge ant colony routing protocol through the ant colony algorithm with the help of the Mult-IP identification message to dynamically optimize the transmission of task data packets, and find the best route to the destination as the preferred task transmission path.

[0048] Specifically, the object identifier Mult-IP is an extended header structure Mult-IP based on the IPv6 protocol, which supports physical or virtual network objects, such as content, services, IoT terminals, network devices, application services, etc., and maps the object identifier to the IPv6 address of the network layer, which is compatible with IP and implements the identification and management of objects at the network layer. Since the size of the Function field is fixed, the number of Function field sets in the extended header can be inferred from the first Function field. Further, if Figure 2As shown in the T-uid parameter value table, the fields of the object identifier Mult-IP include the Function field set and the M-data field. The Function field set is used to explain the specific meaning of the M-data field set, and the M-data field set contains specific parameter values. In addition, each Function field is fixedly composed of three parts, with a total length of 16 bits. The type (Type) occupies 4 bits and describes the type corresponding to the M-data. The length (Length) occupies 4 bits and indicates the length of the M-data in bytes. The offset (Offset) describes the character offset of the M-data in the entire Mult-IP extension header to quickly locate the position of the M-data.

[0049] Specifically, Figure 3 As shown in the service type table, the object identifier Mult-IP implements the configuration of various network functions according to the semantic information of different types of requirements of mobile devices, including: task identifier T-uid message, device identifier D-uid message, service identifier S-uid message and content identifier C-uid message.

[0050] Among them, Figure 4 As shown in the T-uid example diagram, the task identification T-uid message is in an environment without infrastructure. When a mobile device generates a task and cannot complete the execution of the task due to insufficient resources, it will request the surrounding devices to unload the task. At this time, the task message transmitted by the requesting device will contain the corresponding task identification T-uid message. The task identification of the task identification T-uid message records the information related to the computing task; the T-uid message includes , where tid is the unique identifier of each task, At is the input size of each task, Ct is the number of clock cycles of the task, St is the type of task, TDDL is the deadline of the task, Ft is the computing power required for the task, Bt is the required bandwidth for the task, Mt is the memory required for the task, Timestamp is the timestamp generated by the task, and the task is abstractly represented as a nine-tuple in the system: Furthermore, the total length of the T-uid extension header is 256 bits, of which the Function field set occupies 144 bits, including There are 9 fields in total, and the M-data field set occupies 112 bits. The function of the M-data field set is to record the corresponding parameter values. For fields without units, such as It can accurately record parameter values. For fields with units, a combined counting method of values ​​and units can be used, with the first 4 digits representing the value and the last 4 digits representing the unit. Taking bandwidth as an example, the last four digits "0000" represent 3kpbs, "0001" represent 30kpbs, "0011" represent 300kpbs, and so on. The timestamp generated by the task uses the total number of seconds from 00:00:00 Greenwich time on January 1, 1970 to the present. The task types are divided into 5 types, and the field St records the corresponding binary value of each task type, such as Figure 2 A T-uid example is shown in Figure 4 As shown, the 9 Function fields correspond to 9 M-data fields respectively, and M-data records the complete information of a computing task.

[0051] Since the Mult-IP extended header structure can dynamically increase or decrease the number of fields as needed, it can effectively improve the flexibility of identifying messages and prevent the generation of redundant fields. Tasks are abstractly represented as nine tuples in the system: , where each element has the following meaning: Figure 2 As shown.

[0052] The information of the message can be abstracted as {34, 800, 232, 2, 200, 4, 300, 1000, 2023-06-0120:41:55}, which means that the message is a task message and the task is in the request state. The task identification number is 34, the input size is 800KB, the number of clock cycles calculated is 232Mcycle, the task type is 2 (environmental monitoring), the deadline is 200ms, the required computing power is 4GHZ, the required transmission bandwidth is 300kbps, the required memory is 1MB, and the task generation time is 2023-06-01 20:41:55.

[0053] The device identification D-uid message is the relevant information of the nodes in the network, which is used to allow the network to perceive the surrounding environment information in a fine-grained manner for resource sharing; the device identification D-uid adopts the Mult-IP extended header structure, including , where nid is the unique identification number of the node, X t is the longitude of the node, Y t is the latitude of the node, V t is the velocity of the node, F t is the computation rate of the node, Et is the remaining energy of the node, E t is the remaining memory of the node, M t is the remaining memory of the node, L t is the type of the node. Further, Figure 5As shown in the D-uid parameter value table, the Function field set in the D-uid extension header occupies 128 bits, including There are 8 fields in total, the M-data field set occupies 88 bits, and the total length of the D-uid extension header is 216 bits. The design of D-uid not only plays a certain role in assisting the source node to select the target node for unloading, but also enables the nodes in the network to better understand the surrounding environment information and reduce the probability of communication interruption. For example, a D-uid consists of the field set {41, 4580, 3300, 6, 2, 100, 3, 2}, representing the node identification number 41, the node's longitude is 76°20′, the latitude is 55°0′, the speed is 6m / s, the computing rate is 2GHZ, the remaining energy is 100kJ, the remaining memory is 2GB, and the node type is 1 (wearable device).

[0054] The service identifier S-uid message is used by mobile devices to express their service needs and quickly locate the target node that can provide services. The service identifier S-uid message includes , where tid is the unique identifier of each task, operation is the service operation, and service is the service information entity. Figure 6 As shown in the figure, all parameter values ​​of S-uid are displayed. In the extended header structure of S-uid, the Function field set occupies 24 bits, including Three fields, M-data field concentration Occupies 24 bits, It is a variable length field, so the total length of the M-data field set of S-uid is also not fixed. The service in the network is abstracted into two fields. and Fields, The fields represent the operations that the service needs to perform. The fields describe the information body that the service is going to operate on. Since the information of the S-uid is attached to the IPv6 data packet, the mobile device can request services from the network layer protocol, such as adjusting the forwarding behavior of the router. For example, an S-uid consists of the field set {69, "cl", "car", 14}. Among them, "cl" represents the cleaning operation, and "car" indicates the file name on which the operation is performed. Combining the two, it can be seen that the service intention is to perform data cleaning on the file named "car".

[0055] Among them, the content identification C-uid message is that when the source node is performing task unloading or executing services, it needs to preload some static resources, then the static resource information required by the source node is added to the Mult-IP extension header to form the content identification C-uid. The content identification C-uid message includes tid, cid and K1, K2, K3, K4, K5, where tid is the unique identifier of each task, cid is the type of content, K1 is the keyword 1 corresponding to the content, K2 is the keyword 2 corresponding to the content, K3 is the keyword 3 corresponding to the content, K4 is the keyword 4 corresponding to the content, and K5 is the keyword 5 corresponding to the content. Further, the parameters of C-uid are as follows Figure 7 As shown. The Function field set in the content identifier C-uid extension header occupies 112 bits, including tid, cid and 5 K1, K2, K3, K4, K5 content keyword fields. In the M-data field set, tid and cid occupy 24 bits and 8 bits respectively, and the length of the keyword corresponding to the content is indefinite, so the total length of the M-data field set is not fixed. Since the field set of Mult-IP supports dynamic expansion, the number of content keywords in C-uid is not fixed and can be increased or decreased according to specific needs, which does not affect the recognition of the extended header by the underlying network. For example, a field set abstracted into {34, 1, "tank", "soldier", "weapon", "landform", "uav"} satisfies the composition of the content identifier C-uid. The content identifier indicates that task 34 of the source node requires type 1 content resources, and the content contains the above five keywords. The design of C-uid can increase the resource sharing capability between network nodes and improve the efficiency of node task offloading.

[0056] The ant colony algorithm is a heuristic algorithm that simulates the foraging behavior of ants and is widely used to solve optimization problems. The routing discovery problem in a random connection network (MANET network) is very similar to the ant pathfinding process, both of which are to find the best route to the destination. It has been proven that the ant colony algorithm has the advantages of distribution, positive feedback, and global convergence in solving routing problems.

[0057] Specifically, Figure 8 The task data routing algorithm is shown in the figure; the ant colony algorithm includes: before the task data starts to be transmitted, determine the execution location of the computing task; for vehicle nodes with more resources, periodically broadcast their device information to surrounding nodes; the broadcasted task data is used as a task data packet, and the task data packet contains a D-uid header related to the device information; the D-uid header contains the node identification number nid, the node location information , the current speed of the node V tAnd its device hardware information When the source node receives the task data packet, it will parse the D-uid information from the task data packet and select a suitable node from the D-uid information as the unloading destination node.

[0058] Specifically, the ant colony algorithm includes: setting the calculation rate of vehicle d to , the remaining memory is , the remaining energy is , then node d should satisfy the condition: ; ; ; At this time, the source node s will randomly select one of the vehicle nodes that meet the conditions as the destination node d for unloading; when the source node s is ready to send a data packet to the destination node d, it will look for the current corresponding routing table entry. If the routing table entry of the destination node d is not found in the routing table, the routing discovery process will be started. For some computing tasks, in addition to the computing data of the task itself, corresponding static resources are also required. Static resources are corresponding data (such as a picture or a text) found based on resource identifiers, which require almost no computing power, but some special static resources are only available to specific nodes.

[0059] In the process of route discovery, it is necessary to find a transmission path to help the task load the required static resources; when route discovery starts, the source node s will generate a forward ant packet (Forward Ant), and a service identifier S-uid will be attached to the header of the forward ant packet. The content of the service identifier is {task identification value, "load resources" (service operation), "resource identifier" (service information ontology)}. After the intermediate node receives the header of the forward ant packet, it will parse the content of the service identifier S-uid in the header of the forward ant packet, match it with the C-uid contained in the C-uid information table, filter out the neighbor nodes with the corresponding content, and forward the packet according to the pheromone concentration on the link of the filtered neighbor nodes. For example, the header of a forward ant contains an S-uid of {23, "load resources", "location_1.png"}, which means that task 23 needs a static resource named "location_1.png", and the intermediate node i contains two content identifier (C-uid) messages, namely C-uid-1: {0, "location_1.png", "tank_1.png"} from neighbor node 1 and C-uid-2: {0, "location_2.png"} from neighbor node 2. It can be seen that the keyword 1 of C-uid-1 can match the service information field of S-uid, so the intermediate i forwards the forward ant to neighbor node 1. When the content identifiers (C-uid) of multiple neighbor nodes can match the service identifier (S-uid), the forward ant is forwarded according to the pheromone concentration on the path.

[0060] use represents the pheromone concentration on the link from intermediate node i to destination node d via neighbor node n. The probability of intermediate node i selecting next hop node n is expressed as: ;in, It is the set of neighbor nodes that reach the destination node d and contain corresponding static resources recorded in the routing table of node i; is the exploration parameter value used to control the exploration behavior of the ant packet header; for the header of the ant packet from the same source node s, the intermediate node i will only process the header of the forward ant packet received for the first time, and the headers of the ant packets received later will be killed.

[0061] When the destination node d receives the headers of the forward ant packets from different neighboring nodes, it will unicast the headers of the backward ant packets to different neighboring nodes in turn; after the intermediate node i receives the header of the backward ant packet, it will calculate the estimated value of the delay to the destination node d through increments: ; Where n is the number of nodes on the path from intermediate node i to destination node d; The estimated value of the delay for sending a data packet can be expressed as: ;in, is the number of packets in the current MAC layer queue, It represents the transmission time required for node i to successfully send a message, which is the average time from when the data packet reaches the MAC layer to when the transmission is successfully completed. It is expressed as: ;in, ; In order to reflect the changes in network topology, the intermediate node i will update or create pheromones in the routing entry; the value of the pheromone is determined by two parameters: the update operator and the volatility factor. The update operator It is related to the number of hops of the path and the estimated delay, and is defined as: ;in is the average delay of each hop, is the bandwidth between the intermediate node i and the neighbor node j; using the update algorithm and volatility factor , the update rules of the pheromone value in the routing entry are as follows: ; The source node s will observe whether the backward ant returns to the source node. If no backward ant arrives after a certain period of time, the route discovery process will be restarted; when the number of restarts of the route discovery process exceeds the predetermined maximum number of times, the source node discards the cached data packet.

[0062] Different types of services correspond to different types of service intentions. For computing-intensive services, a deadline is set and the services must be completed within the specified deadline. represents the total deadline of task t; when the total delay to complete task t exceeds the deadline , the task offloading is considered failed; the total delay of task t includes the calculation delay and transmission delay, where the calculation delay is expressed as: Among them, C t is the clock cycle number F of task t t is the computation rate of the target node d; the transmission deadline of task t is defined as: ;when, , the task cannot be completed within the deadline, the source node s will cancel the task offloading request and execute task t locally; otherwise, the source node s will abstract the task information into a T-uid and attach it to the IPv6 header, and set the deadline field of T-uid to , start the transmission of the task data packet; after receiving the task data packet, the intermediate node will parse the T-uid content in the header and obtain the transmission deadline of the task And update it using the following formula: Updated ;in For tasks that need to load static resources, the intermediate node j will match the task identifier tid field of the previously received S-uid message with the T-uid message. If the match is successful, the corresponding static resource will be loaded for the computing task, and the node will continue to forward the task data packet; if there are multiple next hops to the destination node d, the node will follow The probability obtained by the formula is randomly forwarded. At this time, a higher value, making the forwarding strategy more greedy; when , then the task transmission time at this time has exceeded the total deadline , task offloading fails; then the node will discard the task data packet and send a task offloading failure message to the source node; after receiving the notification, the source node s will cancel the task offloading request and execute task t locally.

[0063] Specifically, according to this strategy, all computing tasks can be responded to within a limited time, avoiding the source node from initiating multiple task offloading requests, which would cause network congestion. After receiving the task data packet, the destination node will parse the task information from the T-uid, and return the result data to the source node after executing the task.

[0064] From the time complexity of the algorithm, it can be seen that the program execution time of the algorithm increases with the increase of input scale. The time complexity of the task data routing algorithm mainly depends on the number of tasks. When the number of tasks increases, the number of T-uid messages that the router needs to parse increases, so the time complexity of the algorithm is O(n).

[0065] Working principle: like Fig. 9 The uninstallation flowchart, and Fig.10 As shown in the task offloading process diagram, when the source node s requests task offloading from the destination node d, the source node s will periodically release routing maintenance ants to the destination node s. The forwarding rules of the ants are still based on the formula At this time, the ants will reach the destination node d along the existing path and obtain the latest quality estimate of the current path. In addition, the intermediate nodes broadcast with a lower probability after receiving the forward ants, so that the ants may obtain a path that they have not traveled before, which increases the ants' exploration. However, the disadvantage of doing so is that after the neighboring nodes receive the broadcasting ants, they cannot find the pheromone to the destination node, which causes the ants to continue broadcasting, thereby rapidly increasing the number of ants in the network. In order to avoid the above situation, the number of ant broadcasts is limited. If the ant does not find the routing information within the specified number of hops, the ant will be killed. The algorithm complexity of the route maintenance operation is mainly related to the number of ants released. Therefore, the algorithm time complexity of the route maintenance algorithm is O(n). The more ants are released, the greater the overhead required for route maintenance.

[0066] The nodes in MANET are highly mobile. In order to solve the problem of link disconnection caused by node movement, the nodes in the network are required to broadcast a hello message at a fixed time (1s in the experiment). In order to enable the network to share resources, a C-uid extension header is added to the hello message. C-uid records the static resource information it owns. After receiving the hello message, the neighboring node will parse the C-uid from the hello message and store it in the C-uid information table. In addition, the pheromone of the neighboring node will be added to the routing table and the timing will start. If the hello message is not received again within the specified time, it is considered that the current link has been disconnected. At this time, the node will perform a series of routing maintenance operations. First, the node will delete all routing table entries related to the neighboring node and broadcast a link interruption notification message. The content of the message is the destination node list of the best path related to the neighboring node and the new routing information to the destination node (including the number of hops and the estimated end-to-end delay). If the neighboring node is the only routing table entry to the destination node, the message content is set to empty. After receiving the link interruption message, the upstream node will use the routing information in the message to re-estimate the pheromone value to the destination node. If it finds that its best path has changed, it will continue to broadcast the message until all relevant nodes have received the message.

[0067] Embodiment 2 On the basis of Example 1, Fig. 9 The uninstallation flowchart, and Fig.10 As shown in the task offloading process diagram, in a field battlefield environment lacking infrastructure, soldiers and various vehicles form a self-organizing network to communicate with each other and complete combat missions together.

[0068] For example, soldiers in battlefield environments usually carry a variety of wearable devices to collect information to perform combat missions. However, due to the limited computing resources and energy of the devices, once a high-complexity computing task occurs, it is difficult for the soldiers' devices to complete it within the expected time. At this time, the battlefield self-organizing network can be used to migrate the computing tasks to various tanks or rescue vehicles nearby with sufficient computing resources. In this way, not only can the computing pressure of wearable devices be alleviated, but also the energy consumption of the equipment can be reduced, and the resource utilization of the equipment can be improved. Generally speaking, the unloading of soldiers and various vehicles in battlefield environments involves complex processes, which usually consumes a lot of network traffic. The Mult-IP message mechanism is used to simplify the communication process between devices and improve the communication efficiency of the network.

[0069] There are differences in resources between different devices. Introducing edge computing into self-organizing networks solves the problems of resource sharing and content collaboration in battlefield environments. Realizing resource sharing and content collaboration in battlefield environments through edge computing includes the following stages: The first stage is the information collection stage, which collects the D-uid information periodically broadcast by nodes and uses these nodes as candidate nodes for future unloading.

[0070] The second stage is the task offloading stage. When the wearable device generates a computing task t that it cannot complete by itself, the device will abstract the task information into a T-uid and prepare for offloading.

[0071] First, the wearable device will filter out the appropriate target node responsible for task offloading from the device information of the vehicle obtained in the previous stage. The screening process has been discussed in detail in route discovery. Then, the wearable device will send a data packet containing a T-uid header to the vehicle node that meets the relevant conditions. The message content is the computing task data. After receiving the task data, the vehicle node will put the computing task into its own task queue and wait for task processing. After the task processing is completed, the result data will be returned to the source node. If the wearable device receives a task offloading failure notification message after a certain period of time, it means that the task cannot be completed within the specified time and the task offloading fails. At this time, the device will execute task t locally or continue to request offloading after waiting for a period of time.

[0072] The third stage is the service execution stage. After receiving the result data of the computing task, it is necessary to request the corresponding service from the adjacent vehicle node. At this time, the wearable device abstracts the service into an S-uid data packet, which contains the task requested in the previous stage. , the operation of the service , the target of the operation .

[0073] The target of the service request is the unloading destination node in the previous stage. After receiving the service request from the wearable device node, the vehicle node will not immediately perform the service. Instead, the node will continue to collect content information related to the task from the surrounding area. At this time, the vehicle node will broadcast more content information related to the service to the surrounding area to obtain the parameters required for the service. Here, the data packet broadcast by the vehicle device contains C-uid, and the content of the message is empty. The S-uid header contains the task identifier , the type of content , the keywords contained in the content ( , , After receiving the data packet containing C-uid, the surrounding nodes will retrieve the content information contained in the C-uid from the keyword information in the C-uid and try to send the content to the source vehicle node, while the nodes without corresponding content will not respond. At this time, the C-uid only contains the request task Finally, the vehicle device node will comprehensively determine the parameters required for the service from the content information obtained from the surrounding nodes and perform specific services.

[0074] For example, a tactical team consists of 10 soldiers and 2 vehicles. Fig.10 As shown, soldiers and vehicles can form a self-organizing network to communicate with each other. At a certain moment, the soldiers in the unit Received 2 calls from surrounding vehicles and D-uid message from the vehicle The content of the D-uid message is {10,9923,3244, 30,8,5000,2000,2}, indicating that the vehicle's identification number is 10, the location is 165°23′ east longitude, 54°4′ north latitude, the speed is 30m / s, the CPU computing power is 8GHZ, the remaining energy is 5000kJ, the remaining memory is 2GB, and the vehicle type is 2 (combat tank). From the vehicle The D-uid message content is {11,9938,3251,10,5,10000,4000,2}, and its specific information is the same as above. The wearable device suddenly receives an interference signal from the enemy in front. Due to the complexity of the signal, the device is difficult to analyze in time. The device tries to offload the task to the nearby vehicle for execution. The task information is abstracted into a T-uid message. The message content is {120,200,40,2,1000,4,500,300,1705661129}, indicating that the task ID is 120, the input size is 200kb, the clock cycle number is 40Mcycle, the type is 2, the deadline is 1000ms, the required computing power is 4GHZ, the required bandwidth is 500kbps, the memory occupied is 300kb, and the timestamp of the generated task is 2024-01-19 18:45:29. Soldier The transmission equipment of the vehicle was analyzed and found Satisfy the conditions for task offloading and transmit task data to the vehicle After waiting for the processing to be completed, the vehicle Return mission result data to soldiers After receiving detailed information about the enemy, the soldiers wanted to launch a military strike at the target location, so they continued to attack the vehicle. Send a data packet containing S-uid. The content of the data packet is {120, "hit", 9918, 3230, 41}, indicating that the computing task ID is 120, the service operation is 'hit', the service target is 165°18' east longitude, 53°50' north latitude (the location of the hit), and the vehicle ID involved in the hit is 10. Vehicle After receiving the S-uid data packet, a D-uid data packet will be broadcast immediately to obtain more information about the target location (pictures, videos, etc.). The content of the data packet is {120,0,9918,3230, "tank"}, indicating that the identification number of the computing task is 120, the content type is 0 (enemy intelligence), and the content keywords are composed of 165°18' east longitude, 53°50' north latitude, and 'tank', which indicates the enemy's location and the target to be hit is the enemy tank. and After receiving the D-uid data packet, it will actively retrieve or continue to detect the enemy information at the corresponding location. If there is relevant content, it will immediately transmit the content data to the source vehicle node. Finally, the vehicle node The received enemy intelligence will be integrated to correct the parameters required to strike the target and complete the attack on the enemy.

[0075] Embodiment 3 The protocol was simulated and tested. The simulation platform is based on ns-3 (Network Simulation Version 3), which is a discrete event simulator that can simulate wireless networks. In the same simulation environment, the AntEdge protocol is compared with AODV, DSDV, MCLMR

[26] and AODV-MEC to analyze the performance differences between them. MCLMR is a multi-path optimized link state routing protocol. When selecting intermediate nodes, the protocol considers the mobility of nodes, contention window and link quality, and uses multi-criteria decision technology to provide weights for each component to obtain the optimal routing path. AODV-MEC is an improved routing protocol based on AODV in an edge computing environment. The protocol uses a node clustering strategy to enable nodes in the same cluster to use point-to-point communication, while nodes in different clusters establish routing paths through communication between cluster nodes. In addition, AODV-MEC also considers factors such as the mobility and energy consumption of intermediate nodes to improve routing efficiency.

[0076] like Fig.11As shown in the figure, the topology of the simulation network is composed of a rectangular area of ​​500m×1500m. In the first scenario, 60 mobile nodes are evenly distributed in the area, and 10 pairs of nodes are used to simulate the task collaboration process. Each source node sends 4 task data packets per second, and the size of the data packet is 64KB. The clock cycle number of the calculation task is [100,200]Mcycle, and the calculation frequency of the node is [5,10]GHZ. The total simulation time is 900s. After the simulation starts, all source nodes will randomly start sending task data packets between 0 and 180 seconds. The node mobility model adopts the random waypoint (RWM) mobility model, in which the node's moving speed and residence time are changed. The MAC layer adopts the IEEE802.11n protocol, which is a common practice in MANET research. The wireless interface bandwidth is set to 11Mbps, and the radio transmission range is 200m. In the second scenario, the impact of the number of nodes on routing performance is mainly tested. The number of nodes varies from 30 to 120. The node speed is set to 20m / s and the dwell time is 30ms. Four indicators will be used to evaluate the performance of several routing protocols, namely task completion ratio, packet delivery ratio, end-to-end delay, and routing control overhead.

[0077] Task completion rate: the number of tasks completed within the deadline The total number of tasks completed with the destination node Ratio: ; Message delivery rate: the number of data packets sent by the source node The number of packets received by the destination node Ratio: .

[0078] End-to-end delay: The total time it takes for the source node to send data and the destination node to receive the data packet: .

[0079] Routing cost: the ratio of the total number of control messages sent to the number of messages successfully sent: .

[0080] Experimental Results Fig.12 and Fig.13The simulation results of the task completion rate and average end-to-end delay of nodes at different speeds. At various movement speeds, the task completion rate of the AntEdge protocol is higher than that of other protocols, and the end-to-end delay of the message is lower than that of other protocols. The main reason is that the AntEdge protocol takes into account the deadline of each task data packet, effectively avoids the execution of expired tasks, and prevents the source node from repeatedly requesting task unloading, which leads to network congestion. When the node movement speed reaches the maximum, the task completion rate of the AntEdge protocol is 9.5% higher than that of the MCLMR protocol, 14% higher than that of the AODV-MEC protocol, 17% higher than that of the DSDV protocol, and 26% higher than that of the AODV protocol. In addition, when the node speed increases, the end-to-end delay of AntEdge and MCLMR is relatively stable, while the delay of AODV and DSDV is higher. This is because AntEdge and MCLMR are multi-path routing protocols. When the node movement causes the link to be interrupted, it can quickly switch to other paths without re-routing discovery. The AODV-MEC protocol considers the duration of the link and discards some unstable links, so the latency is lower than that of AODV. Fig.14 and Fig.15 The simulation results of the routing overhead and message delivery rate of nodes at different speeds. It can be seen that the routing overhead of the AntEdge protocol is significantly lower than that of other protocols. This is because AntEdge actively limits the blind increment of forward ants during the routing maintenance process, effectively reducing the number of control messages. In terms of the message delivery rate indicator, AntEdge also performs better than other routing protocols. When the node speed increases, the message delivery rates of the two protocols DSDV and MCLMR drop rapidly, while the delivery rates of AntEdge, AODV and AODV-MEC are relatively stable and better than DSDV and MCLMR. This is because DSDV and MCLMR are active routing protocols. In an environment where the network topology changes rapidly, the cost of establishing new available routes is large, while the on-demand routing protocols AntEdge and AODV avoid this situation. In addition, AntEdge is a hybrid routing protocol that combines the advantages of the two protocols and can achieve better performance. Deadline of the task Will directly determine the effect of task offloading, Fig.16 It shows the task completion rate under different task deadlines.

[0081] It can be seen that when the task deadline is large, all routing protocols can have a high task offloading rate. This is because a higher task deadline allows more computing tasks to be completed within the specified time, thus avoiding the failure of task offloading. When the task deadline is set to a smaller time, the AntEdge routing protocol can still maintain a high offloading rate, while other protocols have a rapid drop in offloading rate due to a large number of tasks expired. Fig.17 and Fig.18 The packet delivery rate and average end-to-end delay of nodes at different pause times are shown respectively. When the node pause time is long, the network topology is relatively stable, the packet delivery rate of all routing protocols is slightly improved, and the end-to-end delay is slightly reduced. The performance of the AntEdge protocol is better than other protocols. The performance of the AODV-MEC and MCLMR protocols is lower than AntEdge and better than the DSDV and AODV protocols. This is because DSDV and AODV are single-path routing based on the shortest path. When the link is interrupted and the path fails, AODV and DSDV must rediscover and repair the route, and there is no backup route for transmission, resulting in a decrease in the packet delivery rate and an increase in the end-to-end delay. AntEdge and MCLMR can quickly switch to the backup route, effectively ensuring the transmission of the message. In the process of path selection, the AODV-MEC protocol tends to select some more stable links, so it can reduce the probability of link interruption, and its performance is better than the AODV protocol. Fig.19 The figure shows the routing control overhead of different routing protocols as the pause time changes. It can be seen that the routing control overhead of these five protocols decreases accordingly as the node pause time increases. The generation of control packets is mainly in the routing discovery and routing maintenance stages. When the node pause increases, the probability of link interruption decreases, and there is no need to frequently execute routing discovery and maintenance algorithms, and the corresponding control overhead is reduced. In addition, AntEdge has the lowest routing overhead among the four protocols. This is because AntEdge can effectively avoid paths with poor link quality, reduce the number of routing discoveries, and have lower overhead.

[0082] In the second scenario, the number of nodes in the region was changed to compare the performance indicators of different routing protocols. Fig. 20 This is the message delivery rate of each routing protocol under different numbers of nodes. In terms of message delivery rate, it can be seen that when the number of nodes increases, the message delivery rate of the five protocols decreases slightly, and the performance index of the AntEdge protocol is the best, which is better than the other four protocols. This is because AntEdge can dynamically select different routing paths according to the link quality, dynamically adjust the traffic size on each path, and achieve load balancing between different paths. Fig.21 The end-to-end delay of the five protocols under different numbers of nodes is shown. AntEdge's end-to-end delay is smaller than that of other protocols. When the number of nodes increases, the end-to-end delay of the five protocols increases, and the increase of AODV and DSDV is larger. This is because the increase in the number of nodes will lead to an increase in the density of nodes in the area, making node routing discovery and maintenance more frequent, which leads to an increase in end-to-end delay. Fig. 22The figure below shows the routing overhead of each protocol under different numbers of nodes. As the number of nodes increases, more control messages will be transmitted between nodes, so the control message overhead is on the rise. When the number of nodes is the same, AntEdge has the smallest message overhead, which is lower than the other four protocols. Fig.23 The task completion rate under different numbers of nodes is shown in the figure. It can be seen that AntEdge is better than the other four protocols. When the number of nodes increases, AntEdge's task completion rate can also be maintained at more than 90%, while AODV's task completion rate drops seriously. In order to find out the reason for the drop, the trace file of AODV is analyzed and it is found that the jitter of AODV data packets is significantly larger than that of other protocols. The jitter of data packets causes the delay of a large number of task data packets to exceed the deadline of the task. , resulting in task offloading failure and a decrease in task completion rate.

[0083] Embodiment 4 In a preferred embodiment of the present invention, a task transmission path selection device based on object identification and intent-driven in an uncertain environment includes: a memory; a processor; and a computer program; wherein the computer program is stored in the memory and is configured to be executed by the processor to implement a task transmission path selection method driven by intent in an uncertain environment.

[0084] Working principle: like Figure 1-Figure 23 As shown, the present invention provides an intent-driven task transmission path selection method and device in an uncertain environment; an edge computing task routing solution in a mobile self-organizing network (MANET) environment is proposed to solve the problem of task transmission path selection and offloading in the absence of infrastructure. , and the Mult-IP extended header mechanism is added, four different types of extended messages are designed, and the edge ant colony routing protocol (AntEdge) is implemented in combination with these four extended headers, and the edge computing task offloading process is implemented using these four messages. Simulation experiments show that the edge ant colony routing protocol designed by the present invention improves the task completion rate by 26% compared with the traditional AODV routing protocol, and reduces the average end-to-end delay by 45.2% compared with the MCLMR protocol.

[0085] It can be understood that the various embodiments described in this specification can be implemented individually or in combination, and the embodiments of this specification are not limited to this.

[0086] Unless otherwise specified, all technical and scientific terms used in the embodiments of this specification are the same as those commonly understood by those skilled in the art of the technical field of this specification. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the scope of this specification.

[0087] The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items. The singular forms "a", "above", and "the" used in the embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0088] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the disclosed embodiments can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this specification.

[0089] In the several embodiments provided in this specification, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0090] Based on the ideal embodiment of the present invention, through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. An intention-driven task transmission path selection method in an uncertain environment, characterized in that: The following steps are involved: When the source node has a demand for task offloading, it sends a message containing task data to the destination node, and the source node selects the task transmission path according to the message containing the task data through the task transmission path selection model to obtain the preferred task transmission path; The acquisition of the message containing the task data comprises the following steps: Establishing an object identifier Mult-IP based on an extended header of the IPv6 protocol according to the task data, and mapping the object identifier Mult-IP to an IPv6 address of the network layer, and being compatible with IP; realizing identification and management of objects at the network layer; The task transmission path selection model adopts the edge ant colony routing protocol to dynamically optimize the transmission of task data packets through the ant colony algorithm with the help of Mult-IP identification messages to find the best route to the destination as the preferred task transmission path.

2. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: The fields of the object identifier Mult-IP include a Function field set and an M-data field. The Function field set is used to explain the specific meaning of the M-data field set, and the M-data field set includes parameter values.

3. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: The object identifier Mult-IP implements the configuration of various network functions according to the semantic information of different types of requirements of mobile devices, including: task identifier T-uid message, device identifier D-uid message, service identifier S-uid message and content identifier C-uid message.

4. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: The task identification T-uid message is in an environment without infrastructure. When a mobile device generates a task and cannot complete the execution of the task due to insufficient resources, it will request the surrounding devices to unload the task. At this time, the task message transmitted by the requesting device will contain the corresponding task identification T-uid message. The task identification of the task identification T-uid message records the information related to the computing task; the T-uid message includes , where tid is the unique identifier of each task, At is the input size of each task, Ct is the number of clock cycles of the task, St is the type of task, TDDL is the deadline of the task, Ft is the computing power required for the task, Bt is the required bandwidth for the task, Mt is the memory required for the task, Timestamp is the timestamp generated by the task, and the task is abstractly represented as a nine-tuple in the system: ; and / or, the device identification D-uid message is the relevant information of the node in the network, which is used to allow the network to perceive the surrounding environment information in a fine-grained manner for resource sharing; the device identification D-uid adopts the Mult-IP extended header structure, including , where nid is the unique identification number of the node, Xt is the longitude of the node, Yt is the latitude of the node, Vt is the speed of the node, Ft is the calculation rate of the node, Et is the remaining energy of the node, Et is the remaining memory of the node, Mt is the remaining memory of the node, and Lt is the type of the node; And / or, the service identification S-uid message is used by the mobile device to express its service requirements and quickly locate the target node that can provide the service. The service identification S-uid message includes , where tid is the unique identifier of each task, operation is the operation of the service, and service is the service information entity; And / or, the content identification C-uid message is that when the source node is performing task unloading or executing services, it needs to preload some static resources, then the static resource information required by the source node is added to the Mult-IP extension header to form a content identification C-uid, and the content identification C-uid message includes tid, cid and K1, K2, K3, K4, K5, wherein tid is the unique identifier of each task, cid is the type of content, K1 is the keyword 1 corresponding to the content, K2 is the keyword 2 corresponding to the content, K3 is the keyword 3 corresponding to the content, K4 is the keyword 4 corresponding to the content, and K5 is the keyword 5 corresponding to the content.

5. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: The ant colony algorithm comprises: Before the task data starts to be transmitted, the execution location of the computing task is determined; for vehicle nodes with more resources, their device information will be broadcasted to surrounding nodes periodically; the broadcasted task data is used as a task data packet, and the task data packet contains a D-uid header related to the device information; the D-uid header contains the node identification number nid and the node location information , the current speed of the node V t And its device hardware information When the source node receives the task data packet, it will parse the D-uid information from the task data packet and select a suitable node from the D-uid information as the unloading destination node.

6. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: The ant colony algorithm comprises: Assume the calculation rate of vehicle d is , the remaining memory is , the remaining energy is , then node d should satisfy the condition: ; ; ; At this time, the source node s will randomly select one of the vehicle nodes that meet the conditions as the destination node d for unloading; When the source node s is ready to send a data packet to the destination node d, it will search for the current corresponding routing table entry. If the routing table entry of the destination node d is not found in the routing table, the routing discovery process will be started; During the route discovery process, a transmission path needs to be found to help the task load the required static resources; When route discovery begins, the source node s generates a forward ant packet, and a service identifier S-uid is attached to the header of the forward ant packet; After receiving the header of the forward ant packet, the intermediate node will parse the content of the service identifier S-uid on the header of the forward ant packet, match it with the C-uid contained in the C-uid information table, filter out the neighboring nodes with the corresponding content, and forward the packet according to the pheromone concentration on the link of the filtered neighboring nodes; use represents the pheromone concentration on the link from intermediate node i to destination node d via neighbor node n. The probability of intermediate node i selecting next hop node n is expressed as: ;in, It is the set of neighbor nodes that reach the destination node d and contain corresponding static resources recorded in the routing table of node i; is the exploration parameter value used to control the exploration behavior of the ant packet header; for the header of the ant packet from the same source node s, the intermediate node i will only process the header of the forward ant packet received for the first time, and the headers of the ant packets received later will be killed.

7. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 6, characterized in that: The ant colony algorithm includes: when the destination node d receives the header of the forward ant packet from different neighboring nodes, it will unicast the header of the backward ant packet to different neighboring nodes in turn; after the intermediate node i receives the header of the backward ant packet, it will calculate the estimated value of the delay to the destination node d by increment: ; Where n is the number of nodes on the path from intermediate node i to destination node d; The estimated value of the delay for sending a data packet can be expressed as: ;in, is the number of packets in the current MAC layer queue, It represents the transmission time required for node i to successfully send a message, which is the average time from when the data packet reaches the MAC layer to when the transmission is successfully completed. It is expressed as: ;in, ; In order to reflect the changes in network topology, the intermediate node i will update or create pheromones in the routing entry; the value of the pheromone is determined by two parameters: the update operator and the volatility factor. The update operator It is related to the number of hops of the path and the estimated delay, and is defined as: ;in is the average delay of each hop, is the bandwidth between the intermediate node i and the neighbor node j; Using the update algorithm and volatility factor , the update rules of the pheromone value in the routing entry are as follows: ； The source node s will observe whether the backward ants return to the source node. If no backward ants arrive after a certain period of time, the route discovery process will be restarted. When the number of restarts of the route discovery process exceeds the predetermined maximum number, the source node discards the cached data packets.

8. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 7, characterized in that: Different types of services correspond to different types of service intentions. For computing-intensive services, a deadline is set and the services must be completed within the specified deadline. represents the total deadline of task t; When the total delay to complete task t exceeds the deadline , the task offloading is considered failed; the total delay of task t includes the calculation delay and transmission delay, where the calculation delay is expressed as: Among them, C t is the clock cycle number F of task t t is the computation rate of the target node d; the transmission deadline of task t is defined as: ;when, , the task cannot be completed within the deadline, the source node s will cancel the task offloading request and execute task t locally; otherwise, the source node s will abstract the task information into a T-uid and attach it to the IPv6 header, and set the deadline field of T-uid to , start the transmission of the task data packet; after receiving the task data packet, the intermediate node will parse the T-uid content in the header and obtain the transmission deadline of the task And update it using the following formula: ;in For tasks that need to load static resources, the intermediate node j will match the task identifier tid field of the previously received S-uid message with the T-uid message. If the match is successful, the corresponding static resource will be loaded for the computing task, and the node will continue to forward the task data packet; if there are multiple next hops to the destination node d, the node will follow The probability obtained by the formula is randomly forwarded. At this time, a higher value, making the forwarding strategy more greedy; when , then the task transmission time at this time has exceeded the total deadline , task offloading fails; then the node will discard the task data packet and send a task offloading failure message to the source node; after receiving the notification, the source node s will cancel the task offloading request and execute task t locally.

9. The method for selecting an intention-driven task transmission path in an uncertain environment according to claim 1, characterized in that: It also includes the situation where soldiers and various vehicles form a self-organizing network to communicate with each other and complete combat missions in a field battlefield environment lacking infrastructure; There are differences in resources between different devices. Introducing edge computing into self-organizing networks can solve the problem of resource sharing and content collaboration in battlefield environments. Realizing resource sharing and content collaboration in battlefield environments through edge computing includes the following stages: The first phase is the information collection phase, which collects the D-uid information broadcast periodically by nodes and uses these nodes as candidate nodes for future offloading; The second stage is the task offloading stage. When the wearable device generates a computing task t that it cannot complete by itself, the device will abstract the task information into a T-uid to prepare for offloading. The third stage is the service execution stage. After receiving the result data of the computing task, it is necessary to request the corresponding service from the adjacent vehicle node. At this time, the wearable device abstracts the service into an S-uid data packet, which contains the task requested in the previous stage. , the operation of the service , the target of the operation .

10. A task transmission path selection device based on object identification and intention-driven in an uncertain environment, characterized in that: include: Memory; processor; as well as Computer programs; Wherein, the computer program is stored in the memory and is configured to be executed by the processor to implement the intention-driven task transmission path selection method in an uncertain environment as described in any one of claims 1 to 9.

Images