Distributed data transmission optimization method, device and equipment

Abstract

The invention provides a distributed data transmission optimization method, device and equipment, and belongs to the technical field of wireless communication, and the method comprises the steps: carrying out the strategy optimization of each node of a Mesh network based on a near-end strategy optimization algorithm, and obtaining an optimization strategy result, the optimization strategy result comprises a route selection optimization strategy and a node dormancy and wake-up optimization strategy of the Mesh network; performing fragmentation and routing planning on the to-be-transmitted data according to an optimization strategy result to obtain fragmentation information and a routing path; performing data transmission on the to-be-transmitted data according to the fragment information and the routing path; and in the data transmission process, optimizing the routing path based on a preset routing optimization strategy until the data transmission is finished. According to the method, intelligent decisions of routing selection, node dormancy and wakeup can be realized, the fragment size and the transmission path are dynamically adjusted, uniform distribution and efficient scheduling of data fragments in the network are ensured, single-point overload is avoided, and the data storage and transmission efficiency of the whole network is improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a distributed data transmission optimization method, device and equipment. Background Art

[0002] As a multi-hop, self-organizing, and self-healing network architecture, wireless mesh networks have been widely used in data communications in the Internet of Things, wireless sensor networks, and smart cities due to their high flexibility and scalability. Each node in the mesh network not only has the function of sending and receiving data, but also can act as a relay node to help other nodes forward data, so it has strong network robustness. Compared with traditional centralized networks, the decentralized architecture of the mesh network can better adapt to dynamic environments where nodes frequently join or leave.

[0003] However, as the scale of the network expands, the number of transmission links between nodes increases, and the complexity of data transmission increases. How to optimize the data transmission path and avoid single-point overload has become an urgent problem to be solved in the current Mesh network. Summary of the invention

[0004] The present invention provides a distributed data transmission optimization method, device and equipment, which can optimize the data transmission path and avoid single point overload.

[0005] The present invention provides a distributed data transmission optimization method, comprising: Performing policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; According to the optimization strategy result, the data to be transmitted is segmented and routed to obtain segmentation information and routing paths; Performing data transmission on the data to be transmitted according to the sharding information and the routing path; During the data transmission process, the routing path is optimized based on a preset routing optimization strategy until the data transmission is completed.

[0006] As an embodiment, the optimizing the routing path based on a preset routing optimization strategy includes: Constructing a current graph structure model of the Mesh network, wherein the vertices of the current graph structure model are determined based on the nodes of the Mesh network, the edges of the current graph structure model are determined based on the links between the nodes, and the weights of the edges are determined based on the comprehensive quality of the links; Constructing a graph neural network according to the nodes in the current graph structure model and the relationship between neighboring nodes; Obtaining the current node state of any node in the Mesh network, inputting the current node state into the graph neural network, the graph neural network is used to update the current node state, and outputting the updated current node state to a preset time series prediction model to obtain a predicted node state of any node; The routing path is optimized according to the predicted node status.

[0007] As an embodiment, it also includes: Obtaining real-time link quality parameters and current transmission rate of any node in the Mesh network, wherein the real-time link quality parameters include signal strength, link packet loss rate, link delay and bandwidth utilization; Determining a comprehensive quality evaluation result of the link according to the link quality parameter, and adjusting the current transmission rate according to the comprehensive quality evaluation result; The routing path is optimized based on the adjusted transmission rate.

[0008] As an embodiment, the data transmission of the data to be transmitted according to the fragmentation information and the routing path includes: Encoding the slice information based on network coding technology to obtain a coding packet; Distributing the encoded packets to the routing path based on a multi-path transmission technology to achieve data transmission of the data to be transmitted; Correspondingly, the optimizing the routing path based on a preset routing optimization strategy includes: Obtaining real-time link quality parameters and multipath transmission results of any node in the Mesh network; According to the real-time link quality parameter and the multipath transmission result, the redundancy of the coded packet is adjusted based on the erasure coding technology to optimize the routing path.

[0009] As an embodiment, it also includes: Obtaining the real-time transmission rate and energy status of any node in the Mesh network; Training a predetermined local model according to the real-time transmission rate and the energy state to obtain update parameters corresponding to the local model; The real-time transmission rate and / or the routing path are optimized according to the update parameters corresponding to the local model.

[0010] As an embodiment, the method of performing policy optimization on each node of the Mesh network based on the proximal policy optimization algorithm to obtain an optimized policy result includes: Determine the state space of the node and the dynamic optimization target according to the energy state, communication load, signal quality and neighbor node state of each node in the Mesh network at the current time step; Initializing the state space based on a proximal strategy optimization algorithm, and inputting the initialized state space into a preset time series prediction model to obtain a predicted state space; Substituting the dynamic optimization target into the Lagrange multiplier method to optimize the energy state in the predicted state space to obtain an optimized state space; Input the optimized state space into the node policy network corresponding to the Mesh network, update the node policy network based on the policy gradient method of proximal policy optimization, and obtain the optimal action of the current time step output by the node policy network, wherein the optimal action includes node routing selection and sleep and wake-up operations; Repeat the above steps, combine the optimal actions of all time steps, and get the optimized strategy result.

[0011] As an embodiment, the data to be transmitted includes data segments with different data values. Correspondingly, the data to be transmitted is segmented and routed according to the optimization strategy result to obtain segmentation information and routing path, including: Obtaining the node storage capacity, node width, node bandwidth, number of neighbor nodes and communication load corresponding to the optimization strategy result; Substituting the data value of the data fragment, the node storage capacity and the node width into a non-uniform hash function to fragment the data to be transmitted to obtain data fragments; According to the node bandwidth and the number of neighbor nodes, the data shards are allocated to the nodes corresponding to the optimization strategy results based on a hash allocation mechanism to obtain the shard information; The routing selection optimization strategy is optimized according to the node storage capacity, the node width and the communication load to obtain the routing path.

[0012] As an embodiment, during the data transmission process, optimizing the routing path based on a preset routing optimization strategy until the data transmission is completed includes: During data transmission, obtain the real-time node storage capacity, real-time node width, real-time node bandwidth and real-time communication load; Substituting the real-time node storage capacity and the real-time node width into a non-uniform hash function to slice the remaining data to be transmitted to obtain real-time data slices; According to the real-time node bandwidth and the number of neighbor nodes, the real-time data shards are allocated and optimized based on a hash allocation mechanism to obtain optimized shard information; Performing routing optimization according to the real-time node storage capacity, the real-time node width and the real-time communication load to obtain an optimized routing path; Repeat the above steps until the data transfer is completed.

[0013] The present invention also provides a distributed data transmission optimization device, comprising: A policy optimization module, used to perform policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; A slicing and routing module, used to slice and route the data to be transmitted according to the optimization strategy results, and obtain slicing information and routing paths; A data transmission module, used for transmitting the data to be transmitted according to the sharding information and the routing path; The routing optimization module is used to optimize the routing path based on a preset routing optimization strategy during the data transmission process until the data transmission is completed.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the distributed data transmission optimization method as described above is implemented.

[0015] The distributed data transmission optimization method, device and equipment provided by the present invention perform policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm, realize intelligent decision-making of route selection, node sleep and wake-up, and optimize the routing path based on a preset routing optimization strategy during data transmission, dynamically adjust the fragment size and transmission path, ensure the uniform distribution and efficient scheduling of data fragments in the network, avoid the phenomenon of single point overload, and improve the data storage and transmission efficiency of the entire network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is one of the flow charts of the distributed data transmission optimization method provided by the present invention.

[0018] Figure 2It is a schematic diagram of the process of performing strategy optimization based on the proximal strategy optimization algorithm provided by the present invention.

[0019] Figure 3 It is a flowchart of the training and application of the collaborative learning model provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the distributed data transmission optimization device provided by the present invention.

[0021] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] It should be noted that all actions of acquiring signals, information or data in the present invention are performed in compliance with the corresponding data protection laws and policies of the location and with the authorization given by the owner of the corresponding device.

[0024] Most existing Mesh network data transmission methods rely on static routing protocols and simple link selection strategies, such as AODV (Ad hoc On-demand Distance Vector Routing) and DSR (Dynamic Source Routing). Static routing protocols establish the shortest path by regularly broadcasting routing information to ensure that data can reach the destination with the minimum number of hops. With the dynamic changes in network topology, the limitations of traditional routing protocols gradually emerge. First, static routing strategies are prone to data loss or delay when link quality is unstable, and cannot adapt to the ever-changing network environment. Second, traditional routing algorithms mainly use the shortest path as the optimization goal, ignoring the energy status and load conditions of nodes, which can easily lead to premature energy exhaustion of some nodes and cause network splitting. Finally, the existing Mesh network routing algorithms lack real-time monitoring and feedback mechanisms for link quality, and cannot adaptively adjust transmission strategies according to real-time network status, resulting in data transmission efficiency and network stability being difficult to guarantee.

[0025] The distributed data transmission optimization method, device and equipment provided by the present invention are described in detail below with reference to the accompanying drawings.

[0026] Figure 1This is one of the flow charts of the distributed data transmission optimization method provided by the present invention, such as Figure 1 As shown, the present invention provides a distributed data transmission optimization method, comprising the following steps.

[0027] Step S100, performing policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network.

[0028] Proximal Policy Optimization (PPO) is a policy gradient algorithm for reinforcement learning. It is widely used to train intelligent agents to complete various tasks. It aims to improve the stability and efficiency of policy updates and overcome the high variance problem that may occur in traditional policy gradient methods.

[0029] Based on the proximal policy optimization algorithm, the policy of each node in the Mesh network is optimized. That is, based on the proximal policy optimization algorithm, the performance parameters of each node in the Mesh network and the performance parameters of its neighboring nodes are optimized and learned, and the routing selection, node sleep and wake-up strategies are dynamically adjusted to obtain the optimized strategy results.

[0030] Step S200, fragment and route the data to be transmitted according to the optimization strategy result to obtain fragment information and routing path. The fragment information is used to represent the data fragments that should be allocated to the next hop node or each hop node according to the optimization strategy result.

[0031] Step S300: Transmit the data to be transmitted according to the fragmentation information and the routing path.

[0032] Step S400: During the data transmission process, the routing path is optimized based on a preset routing optimization strategy until the data transmission is completed.

[0033] It can be understood that the present invention performs policy optimization on each node of the Mesh network based on the proximal policy optimization algorithm, realizes intelligent decision-making of route selection, node sleep and wake-up, and optimizes the routing path based on the preset routing optimization strategy during data transmission, dynamically adjusts the fragment size and transmission path, ensures the uniform distribution and efficient scheduling of data fragments in the network, avoids the phenomenon of single point overload, and improves the data storage and transmission efficiency of the entire network.

[0034] like Figure 2As shown, based on the above embodiment, as an optional embodiment, the proximal policy optimization algorithm is used to perform policy optimization on each node of the Mesh network to obtain an optimized policy result, including steps S110 to S160.

[0035] Step S110, determining the state space of the node and the dynamic optimization target according to the energy state, communication load, signal quality and neighbor node state of each node in the Mesh network at the current time step.

[0036] Define the energy state of the i-th node in the Mesh network at the current time step , Communication Load , signal quality and neighbor node status , the state space of the node is set to , state space Including energy consumption, communication load, number of neighbor nodes and real-time performance parameters of the link.

[0037] Define the dynamic optimization goal of the node as minimizing the nonlinear cost function : ; in, and represents the dynamically adjusted weight coefficient, k and m are the optimization times corresponding to the synchronization time t of the i-th node, and the values ​​of k and m can be adjusted dynamically until the optimal parameters are found. Indicates the total number of nodes.

[0038] Step S120, initializing the state space based on the proximal strategy optimization algorithm, inputting the initialized state space into a preset time series prediction model to obtain a predicted state space. The time series prediction model is used to predict the energy state, communication load and signal quality of each node in the next time step.

[0039] Step S130: Substitute the dynamic optimization target into the Lagrange multiplier method to optimize the energy state in the predicted state space to obtain an optimized state space.

[0040] Specifically, define the function of the Lagrange multiplier method as follows: ; in, represents the Lagrange multiplier, represents the nonlinear weight control factor, Indicates the minimum energy threshold of a node.

[0041] The Lagrange multiplier method is used to solve the constraint problem in energy and performance optimization, solve the optimized energy state, and substitute the optimized energy state into the state space to obtain the optimized state space.

[0042] Step S140, input the optimized state space into the node policy network corresponding to the Mesh network, update the node policy network based on the policy gradient method of proximal policy optimization, and obtain the optimal action of the current time step output by the node policy network (which can be set to ), the optimal action includes node routing selection and sleep and wake-up operations.

[0043] Step S150, repeat the above steps, combine the optimal actions of all time steps, and obtain the optimization strategy result.

[0044] It can be understood that, through the proximal strategy optimization algorithm and deep reinforcement learning, each node in the network can autonomously learn and optimize according to its energy state, communication load and the state of neighboring nodes, thereby realizing intelligent decision-making of routing selection, node sleep and wake-up, greatly improving the flexibility and accuracy of node energy consumption management in Mesh networks, extending the overall service life of the network, and avoiding the problem of network splitting caused by uneven distribution of node energy consumption.

[0045] Based on the above embodiment, as an optional embodiment, the data to be transmitted includes data segments with different data values. Correspondingly, the data to be transmitted is segmented and routed according to the optimization strategy results to obtain segmentation information and routing paths, including steps S200-S240.

[0046] Step S210, obtaining the node storage capacity, node width, node bandwidth, number of neighbor nodes and communication load corresponding to the optimization strategy result.

[0047] Step S220: Substitute the data value of the data fragment, the node storage capacity and the node width into a non-uniform hash function to slice the data to be transmitted to obtain data slices.

[0048] Distributed hash table, abbreviated as DHT, is a distributed storage method, similar to Tracker's network that returns seed information based on seed signatures. Without the need for a server, each client is responsible for routing in a small range and storing a small portion of data, thereby achieving addressing and storage of the entire DHT network. The embodiment of the present invention uses distributed hash table technology to shard data, takes the size, location and node storage capacity of the data shard as input, and constructs a hash key-value pair. ,in A hash key representing a piece of data, Indicates the storage node corresponding to the data fragment.

[0049] In the sharding process, non-uniform hash functions are used Optimize the distribution of data fragments of different sizes according to the node status, non-uniform hash function The expression is as follows: ; in, Indicates the size of the data segment, Representation Node storage capacity, Representation Node The width of represents the weight factor for balancing storage capacity and bandwidth, Indicates the total number of nodes.

[0050] Step S230, according to the node bandwidth and the number of neighboring nodes, the data shards are allocated to the nodes corresponding to the optimization strategy results based on a hash allocation mechanism to obtain the shard information. The shard information is used to characterize the shard sizes allocated to all nodes.

[0051] According to the load and bandwidth of different nodes, the shard size is dynamically adjusted. The expression is as follows: ; in, Indicates allocation to a node The fragment size, represents the node bandwidth, Indicates connection to a node The number of neighbor nodes, and represents the nonlinear adjustment parameter, Indicates the balance factor between bandwidth and the number of neighbor nodes.

[0052] The hash allocation mechanism specifically refers to the efficient search and storage of allocated data fragments, the construction of a distributed hash table, in which each node is responsible for storing a portion of hash key-value pairs, the rapid location of sharded data is achieved through hash table mapping, and the rapid distribution of data is based on hash keys.

[0053] Step S240, optimizing the routing optimization strategy according to the node storage capacity, the node width and the communication load to obtain the routing path.

[0054] The data transmission path is optimized based on the node's storage load, bandwidth, and link quality. The data transmission path is determined based on the routing optimization strategy. The optimization objective function is as follows: ; in, represents the objective function of transmission path optimization, Representation Node The communication load, Represents the weight factor that controls the influence of neighbor nodes.

[0055] In this embodiment, the optimized transmission path can be used as a routing path. In other embodiments, real-time optimization can be performed based on distributed hash table technology and hash allocation mechanism during data transmission to reduce link congestion and node overload. Specifically, during data transmission, the routing path is optimized based on a preset routing optimization strategy until data transmission is completed, including steps S411-S415.

[0056] Step S411, during the data transmission process, obtain the real-time node storage capacity, real-time node width, real-time node bandwidth and real-time communication load.

[0057] Step S412: Substitute the real-time node storage capacity and the real-time node width into a non-uniform hash function to slice the remaining data to be transmitted to obtain real-time data slices.

[0058] Step S413: optimizing the distribution of the real-time data shards based on a hash distribution mechanism according to the real-time node bandwidth and the number of neighbor nodes, to obtain optimized shard information.

[0059] Step S414: performing routing optimization according to the real-time node storage capacity, the real-time node width and the real-time communication load to obtain an optimized routing path.

[0060] Step S415, repeat the above steps until the data transmission is completed.

[0061] Step S411 to step S414 may refer to step S200 to step S240 and will not be described in detail.

[0062] It is understandable that there is a common problem in existing Mesh networks, namely how to reasonably allocate network resources to ensure load balancing and energy consumption optimization of nodes. In most cases, the energy consumption management of nodes is carried out independently, lacking the ability of coordinated optimization of the entire network. As the scale of the network increases, the energy consumption of individual nodes increases rapidly due to frequent participation in data forwarding, while the energy resources of other nodes are not fully utilized, resulting in a shortened overall network life. At the same time, the existing Mesh network transmission rate control method is relatively simple, usually through a preset rate adjustment strategy for data transmission, ignoring the influence of multi-dimensional factors such as link quality, packet loss rate, node energy, etc., resulting in low bandwidth utilization and serious network congestion problems.

[0063] By introducing distributed hash table technology, the present invention can intelligently distribute and manage data in shards, dynamically adjust the shard size and transmission path in combination with the node's bandwidth, storage capacity and load status, ensure uniform distribution and efficient scheduling of data fragments in the network, avoid single-point overload, and improve the data storage and transmission efficiency of the entire network.

[0064] Based on the above embodiment, as an optional embodiment, the optimizing the routing path based on a preset routing optimization strategy includes steps S421 to S425.

[0065] Step S421, constructing a current graph structure model of the Mesh network, wherein the vertices of the current graph structure model are determined based on the nodes of the Mesh network, the edges of the current graph structure model are determined based on the links between the nodes, and the weights of the edges are determined based on the comprehensive quality of the links.

[0066] Graph Neural Networks (GNN) is a type of deep learning model that is specifically designed to process graph data. It extends traditional deep learning techniques to enable it to capture and utilize complex relationships in graph structures. The core idea of ​​GNN is to enable each node to aggregate its neighbor's information to update its own feature representation, which is achieved through an iterative neighbor information aggregation process.

[0067] Construct a graph structure model of the Mesh network and As vertices in the graph, links between nodes As the edge in the graph, define the weight Representation Node and neighbor nodes The overall quality of the link between: ; in, Representation Node and nodes The signal quality of the link between Representation Node The remaining energy, Representation Node The current communication load, Representation Node and nodes The signal-to-noise ratio between represents the smoothing term, and Represent the weighted index of energy and load respectively, An attenuation factor that represents the signal-to-noise ratio.

[0068] Step S422: construct a graph neural network based on the relationship between the nodes and neighbor nodes in the current graph structure model. The graph neural network is used to characterize the relationship between the nodes and neighbor nodes in the current graph structure model.

[0069] Step S423, obtain the current node state of any node in the Mesh network, input the current node state into the graph neural network, the graph neural network is used to update the current node state, and output the updated current node state to a preset time series prediction model to obtain the predicted node state of any node.

[0070] The current node status of the i-th node As shown below: ; in, Representation Node In time The remaining energy, Representation Node In time The communication load, Representation Node and nodes The link quality between Representation Node The number of neighbor nodes.

[0071] The graph neural network is used to extract the dependency between nodes and neighboring nodes through the graph convolutional neural network layer, update the node status and output it to the time series prediction model. The time series prediction model predicts the future link status and predicts the node status. As shown below: ; in, represents the weight matrix for time series prediction, Represents the bias vector.

[0072] Step S424, optimize the routing path according to the predicted node status, update the node relationship in the graph neural network according to the predicted node status, feed back the information of the future state of the node into the node status update through multi-layer graph convolution operations, and output the prediction results of link quality, packet loss rate and delay.

[0073] It is understandable that the use of graph neural networks combined with time series prediction models can effectively capture the topological structure and dependencies between nodes in the Mesh network, and accurately predict future link quality, packet loss rate and delay. Compared with traditional static routing selection methods, the present invention has higher dynamic adaptability and accuracy, and can adjust routing strategies in a timely manner to ensure that data transmission remains stable and efficient when the link status changes.

[0074] Based on the above embodiment, as an optional embodiment, the distributed data transmission optimization method provided by the present invention also includes steps S425 to S427.

[0075] Step S425, obtaining real-time link quality parameters and current transmission rate of any node in the Mesh network, wherein the real-time link quality parameters include signal strength, link packet loss rate, link delay and bandwidth utilization.

[0076] The link quality monitoring mechanism is used to obtain the link quality parameters between nodes in the Mesh network, mainly including signal strength. , Link packet loss rate , Link Delay and bandwidth utilization .

[0077] Step S426: Determine a comprehensive quality evaluation result of the link according to the link quality parameter, and adjust the current transmission rate according to the comprehensive quality evaluation result.

[0078] Function of the comprehensive quality evaluation result of the link As shown below: ; in, and represents the weight coefficient, represents the smoothing term.

[0079] Based on the link quality monitoring and evaluation results, the transmission rate is adjusted based on: ; in, Indicates time The transmission rate at Indicates time The transmission rate at represents the rate adjustment factor, Indicates the link quality threshold.

[0080] Step S427, optimize the routing path based on the adjusted transmission rate. Combined with the transmission rate adjustment mechanism, the transmission rate is adaptively adjusted, and real-time adjustment is performed according to the link state change; the transmission strategy is adjusted according to the dynamically adjusted transmission rate and the current link quality.

[0081] It can be understood that the present invention can dynamically adjust the data transmission rate according to the link quality feedback, ensuring that the bandwidth utilization is improved when the link status is good, and reducing the transmission rate when the link status is poor, so as to reduce packet loss and retransmission, significantly improve the bandwidth utilization efficiency and network stability during data transmission, and effectively avoid transmission interruption or delay problems caused by link quality fluctuations.

[0082] Based on the above embodiment, as an optional embodiment, the data transmission of the data to be transmitted according to the fragmentation information and the routing path includes steps S310 to S320.

[0083] Step S310: encoding the slice information based on network coding technology to obtain a coding packet.

[0084] Step S320: Distribute the encoded packets to the routing path based on the multi-path transmission technology to achieve data transmission of the data to be transmitted.

[0085] The present invention utilizes network coding and multi-path transmission technology to significantly enhance the fault tolerance and stability of data transmission.

[0086] Correspondingly, the optimization of the routing path based on the preset routing optimization strategy includes steps S431 and S432.

[0087] Step S431, obtaining the real-time link quality parameters and multi-path transmission results of any node in the Mesh network.

[0088] Step S432: According to the real-time link quality parameter and the multipath transmission result, the redundancy of the coded packet is adjusted based on the erasure coding technology to optimize the routing path.

[0089] Erasure coding is a data protection method widely used in distributed storage systems to improve data reliability and storage efficiency. The basic idea of ​​erasure coding is to split the original data into multiple data blocks and generate additional redundant data blocks through a coding algorithm. The redundant data blocks are stored together with the original data blocks in different physical locations, such as disks, storage nodes, or geographical locations. When some data blocks are lost or damaged, erasure coding allows the original data to be recovered from the remaining data blocks and redundant blocks, thus providing an efficient data redundancy and fault tolerance mechanism.

[0090] It is understandable that reliable data transmission is also a key challenge in current Mesh networks. Due to the uneven link quality between network nodes and the dynamic changes in node load, packet loss and transmission failure are prone to occur in the network. Especially in an environment with high packet loss rate, the retransmission mechanism will lead to network bandwidth waste and increased transmission delay. Most existing Mesh networks rely on simple retransmission mechanisms to deal with packet loss, without introducing more intelligent error correction mechanisms to improve the fault tolerance of data transmission.

[0091] During the data transmission process, the present invention adopts a strategy combining network coding with multi-path transmission, which can encode multiple data packets and transmit them in parallel through different paths. Even if data is lost on some paths, the receiving end can still restore the original data through the coded packets of other paths. The fault-tolerant mechanism based on network coding reduces the number of retransmissions and significantly improves the reliability and efficiency of data transmission. By introducing erasure code technology, the redundancy is dynamically adjusted according to the link quality and packet loss rate, ensuring that high fault tolerance and transmission performance can be maintained in different network environments, reducing the risk of data loss in network transmission.

[0092] On the basis of the above embodiment, as an optional embodiment, the distributed data transmission optimization method provided by the present invention further includes steps S500 to S700.

[0093] Step S500, obtaining the real-time transmission rate and energy status of any node in the Mesh network.

[0094] Step S600: training a predetermined local model according to the real-time transmission rate and the energy state to obtain update parameters corresponding to the local model.

[0095] Step S700: optimizing the real-time transmission rate and / or the routing path according to the update parameters corresponding to the local model.

[0096] Steps S500 to S700 may be performed after step S300.

[0097] like Figure 3 As shown, the present invention defines a collaborative learning model in a Mesh network, where each node Train the model locally , use the local data of the node to update the model parameters, the local model parameter vector is .

[0098] During each local training process, the node The model update is based on the local loss function The expression of the local loss function is as follows: ; in, Representation Node The number of data samples on Represents input data, represents the target label, Represents the model prediction results.

[0099] Define global model parameters After each node has trained several rounds locally, the local model parameters Upload to the collaborative server and aggregate the global model parameters. The expression is as follows: ; in, represents the global model parameters after aggregation, Indicates the total number of nodes, Representation Node The weight of the data.

[0100] The updated global model parameters After being distributed to all nodes, each node updates the local model after receiving the global model parameters, thus achieving synchronous model update across the entire network.

[0101] Each node continues to train locally based on the new local model parameters. The updated local model parameters will be used to adaptively adjust the node's transmission rate, energy consumption management or routing selection strategy, and ultimately output an optimized global transmission strategy.

[0102] It is understandable that the existing Mesh network generally lacks a collaborative learning mechanism. Data training and model optimization between nodes are mainly performed on a single node, ignoring the collaborative optimization problem of the entire network. As the network scale increases, the state differences between nodes are significant, the effect of each optimization model is greatly reduced, and it is impossible to form a globally optimal transmission strategy.

[0103] The present invention provides a global optimization solution for the transmission rate and energy management of the entire network through collaborative learning technology. By training the model locally at each node and sharing the model parameters, a globally optimized transmission model is formed. When making adaptive adjustments, each node not only considers the local network status, but also makes decisions based on the status of the entire network, thus avoiding the problem of uneven performance of the entire network caused by independent optimization of nodes. The application of collaborative learning effectively improves the transmission rate optimization effect in the network, allowing nodes to collaborate efficiently and form a globally optimal transmission strategy.

[0104] The beneficial effects of the present invention are illustrated below in conjunction with a simulation experiment.

[0105] The preferred solution of the present invention is applied to the intelligent agricultural monitoring system in a certain intelligent agricultural park. The intelligent agricultural monitoring system includes multiple IoT nodes, each of which is responsible for collecting data such as temperature, humidity, soil moisture and air quality, and transmits data through the Mesh network. Due to the large number of network nodes, wide distribution and limited node energy, traditional Mesh network transmission faces problems such as low transmission efficiency, excessive node energy consumption, and unstable link quality, and cannot meet the requirements of efficient and stable data transmission under large-scale deployment.

[0106] A total of 300 nodes are deployed in the park, and the nodes are interconnected through a wireless Mesh network, with an average transmission distance of 100 meters for each node. In order to optimize the network transmission performance, the present invention applies the distributed data transmission optimization method of the present invention, aiming to solve the problems of poor link quality, uneven node energy consumption, and lagging transmission rate.

[0107] First, the proximal policy optimization algorithm is used to optimize the policy of each node in the Mesh network. Each node monitors its energy status, communication load, and the status of neighboring nodes in real time, and dynamically adjusts its routing selection and node sleep and wake-up strategy through the proximal policy optimization algorithm. The test results show that after applying the optimization strategy, the energy consumption distribution of the nodes is obviously more balanced. Before the experiment started, some nodes in the park consumed more than 80% of their energy within 72 hours due to excessive participation in data forwarding, while other nodes consumed energy more slowly. After applying the optimization strategy, the average node energy consumption rate remained within 5% per hour, and the overall energy consumption distribution of the system became more even, effectively extending the life of the network.

[0108] Secondly, the present invention optimizes data sharding through distributed hash table technology, distributes data fragments according to the bandwidth and storage capacity of each node, ensures that data is evenly distributed in the network, and avoids individual nodes from becoming bottleneck nodes due to excessive load. Through data analysis, before the application of the present invention, the CPU utilization rate of some high-load nodes often reaches more than 85%, and the bandwidth utilization rate exceeds 90%, which easily leads to data transmission delays and packet loss. After the distributed hash table technology is applied, the data shard size is dynamically adjusted according to the storage capacity and bandwidth of the node, the average CPU utilization rate of high-load nodes drops to 65%, and the bandwidth utilization rate is also stabilized below 75%, and the overall data transmission delay is reduced by about 35%.

[0109] In terms of link quality prediction, the present invention uses a graph neural network combined with a time series prediction model to predict link status. By constructing a graph structure model of the Mesh network, each node predicts future link quality, packet loss rate, and delay based on historical data. In an experiment during a certain period of time, the present invention monitored nodes with poor link quality for 72 consecutive hours and found that the data packet loss rate using traditional methods was as high as 15%. After using the graph neural network for prediction, the nodes with reduced link quality were identified in advance, and the routing strategy was adjusted, and the packet loss rate was significantly reduced to 3%. In addition, the transmission delay of the system was reduced from an average of 850ms to 600ms, and the link quality was significantly improved.

[0110] In order to improve the fault tolerance of data transmission, network coding and multi-path transmission technology were introduced in the data transmission process. Through comparative experiments, it was found that in the edge area with weak signals, the packet loss rate of the transmission path without network coding was as high as 20%. After the network coding technology was applied, multiple paths transmitted coded data packets at the same time. The receiving end successfully restored the original data after obtaining the coded packets through different paths, and the packet loss rate was significantly reduced to 5%. This improvement reduced the number of retransmissions and increased the transmission efficiency by nearly 50% in a packet loss environment.

[0111] At the same time, the present invention dynamically adjusts the transmission rate through an adaptive bit rate control algorithm. In actual monitoring of a certain period of time, when the link quality is poor, the transmission rate of the traditional fixed-rate transmission method remains at a high level, causing the packet loss rate to rise sharply to 18%. After applying adaptive bit rate control, the system adjusts the transmission rate in real time according to the feedback information of the link quality. When the link quality decreases, the rate decreases accordingly, the packet loss rate remains within 3%, and the transmission stability is significantly improved.

[0112] Finally, in order to further optimize the transmission rate and resource management of the entire network, the present invention introduces collaborative learning technology. Each node performs model training locally and achieves global optimization of the entire network by sharing model parameters. In a week-long test, the system's full network transmission rate increased by about 20%, the energy consumption distribution between nodes became more even, and the energy gap between the node with the least energy consumption and the node with the most energy consumption was reduced from the initial 30% to 15%, greatly improving the service life of the entire network.

[0113] Table 1 Performance comparison of traditional method and the method of the present invention in Mesh network

[0114] According to the comparative data in Table 1, the distributed data transmission optimization method of the present invention shows significant advantages in multiple key performance indicators.

[0115] First, the balance of node energy consumption is an important indicator for measuring the stability and life of the Mesh network. When using the traditional method, some nodes consume more than 80% of their energy within 72 hours, causing these nodes to quickly run out of energy and possibly causing the network to split. In contrast, the method of the present invention can effectively balance the energy consumption of the nodes, keeping the energy consumption of each node within 5% per hour, greatly extending the overall life of the network and preventing some nodes from failing prematurely due to excessive load.

[0116] Secondly, in terms of computing resource utilization, the average CPU utilization of some nodes under the traditional method is as high as 85% or more, and the bandwidth utilization is also over 90%, which means that some nodes may have bottlenecks under high load. However, the present invention reduces the CPU utilization to 65% and the bandwidth utilization to below 75% through intelligent sharding and transmission path selection technology, ensuring the uniform distribution of network load, reducing the phenomenon of single-point overload, and significantly improving the efficiency of resource utilization.

[0117] In addition, data transmission delay and packet loss rate are two key factors affecting network transmission performance. Without the application of the present invention, the average data transmission delay is 850ms, and the packet loss rate is as high as 15%-20%. After the application of the present invention, the link state prediction is performed with the help of graph neural network and time series prediction model, and multi-path transmission is performed in combination with network coding technology. The data transmission delay is reduced to 600ms, and the packet loss rate is also significantly reduced to 3%-5%. The present invention greatly improves the reliability and timeliness of data transmission, especially in the case of poor link quality, the number of retransmissions is reduced, and the transmission efficiency is significantly improved.

[0118] Finally, through collaborative learning technology, the present invention improves the transmission rate of the entire network by about 20%. In contrast, the transmission rate of the traditional method has not been significantly improved. In addition, the present invention also effectively reduces the difference in node energy consumption distribution, reducing the energy consumption gap between the minimum energy consumption and maximum energy consumption nodes from 30% under the traditional method to 15%, further balancing the resource allocation in the network and improving the overall performance of the network.

[0119] In summary, based on the optimization method of the present invention, the Mesh network has shown significant improvements in energy consumption management, resource allocation, data transmission efficiency and link quality stability, fully solving the bottleneck problems in traditional methods, enabling the network to maintain efficient and stable operation in a dynamic and complex environment.

[0120] The distributed data transmission optimization device provided by the present invention is described below. The distributed data transmission optimization device described below and the distributed data transmission optimization method described above can be referenced to each other.

[0121] Figure 4is a schematic diagram of the structure of the distributed data transmission optimization device provided by the present invention, such as Figure 4 As shown, the present invention also provides a distributed data transmission optimization device, including the following modules.

[0122] The policy optimization module 410 is used to perform policy optimization on each node of the Mesh network based on the proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; A slicing and routing module 420, configured to perform slicing and routing planning on the data to be transmitted according to the optimization strategy result, and obtain slicing information and routing paths; A data transmission module 430, configured to transmit the data to be transmitted according to the sharding information and the routing path; The route optimization module 440 is used to optimize the route path based on a preset route optimization strategy during the data transmission process until the data transmission is completed.

[0123] As an embodiment, the route optimization module 440 is further configured to: Constructing a current graph structure model of the Mesh network, wherein the vertices of the current graph structure model are determined based on the nodes of the Mesh network, the edges of the current graph structure model are determined based on the links between the nodes, and the weights of the edges are determined based on the comprehensive quality of the links; Constructing a graph neural network according to the nodes in the current graph structure model and the relationship between neighboring nodes; Obtaining the current node state of any node in the Mesh network, inputting the current node state into the graph neural network, the graph neural network is used to update the current node state, and outputting the updated current node state to a preset time series prediction model to obtain a predicted node state of any node; The routing path is optimized according to the predicted node status.

[0124] As an embodiment, the route optimization module 440 is further configured to: Obtaining real-time link quality parameters and current transmission rate of any node in the Mesh network, wherein the real-time link quality parameters include signal strength, link packet loss rate, link delay and bandwidth utilization; Determining a comprehensive quality evaluation result of the link according to the link quality parameter, and adjusting the current transmission rate according to the comprehensive quality evaluation result; The routing path is optimized based on the adjusted transmission rate.

[0125] As an embodiment, the data transmission module 430 is further used for: Encoding the slice information based on network coding technology to obtain a coding packet; Distributing the encoded packets to the routing path based on a multi-path transmission technology to achieve data transmission of the data to be transmitted; Correspondingly, the route optimization module 440 is further used for: Obtaining real-time link quality parameters and multipath transmission results of any node in the Mesh network; According to the real-time link quality parameter and the multipath transmission result, the redundancy of the coded packet is adjusted based on the erasure coding technology to optimize the routing path.

[0126] As an embodiment, a collaborative optimization module is further included, and the collaborative optimization module is used to: Obtaining the real-time transmission rate and energy status of any node in the Mesh network; Training a predetermined local model according to the real-time transmission rate and the energy state to obtain update parameters corresponding to the local model; The real-time transmission rate and / or the routing path are optimized according to the update parameters corresponding to the local model.

[0127] As an embodiment, the strategy optimization module 410 is further used for: Determine the state space of the node and the dynamic optimization target according to the energy state, communication load, signal quality and neighbor node state of each node in the Mesh network at the current time step; Initializing the state space based on a proximal strategy optimization algorithm, and inputting the initialized state space into a preset time series prediction model to obtain a predicted state space; Substituting the dynamic optimization target into the Lagrange multiplier method to optimize the energy state in the predicted state space to obtain an optimized state space; Input the optimized state space into the node policy network corresponding to the Mesh network, update the node policy network based on the policy gradient method of proximal policy optimization, and obtain the optimal action of the current time step output by the node policy network, wherein the optimal action includes node routing selection and sleep and wake-up operations; Repeat the above steps, combine the optimal actions of all time steps, and get the optimized strategy result.

[0128] As an embodiment, the data to be transmitted includes data fragments with different data values. Correspondingly, the fragment routing module 420 is further used to: Obtaining the node storage capacity, node width, node bandwidth, number of neighbor nodes and communication load corresponding to the optimization strategy result; Substituting the data value of the data fragment, the node storage capacity and the node width into a non-uniform hash function to slice the data to be transmitted to obtain data slices; According to the node bandwidth and the number of neighbor nodes, the data shards are allocated to the nodes corresponding to the optimization strategy results based on a hash allocation mechanism to obtain the shard information; The routing selection optimization strategy is optimized according to the node storage capacity, the node width and the communication load to obtain the routing path.

[0129] As an embodiment, the route optimization module 440 is further configured to: During the data transmission process, the real-time node storage capacity, real-time node width, real-time node bandwidth and real-time communication load are obtained; Substituting the real-time node storage capacity and the real-time node width into a non-uniform hash function to slice the remaining data to be transmitted to obtain real-time data slices; According to the real-time node bandwidth and the number of neighbor nodes, the real-time data shards are allocated and optimized based on a hash allocation mechanism to obtain optimized shard information; Performing routing optimization according to the real-time node storage capacity, the real-time node width and the real-time communication load to obtain an optimized routing path; Repeat the above steps until the data transfer is completed.

[0130] The distributed data transmission optimization device provided by the present invention has the technical effect corresponding to the distributed data transmission optimization method, which will not be repeated here.

[0131] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the distributed data transmission optimization method, the method comprising: performing policy optimization on each node of the Mesh network based on the proximal policy optimization algorithm to obtain an optimization policy result, the optimization policy result comprising a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; performing slicing and routing planning on the data to be transmitted according to the optimization policy result to obtain slicing information and a routing path; transmitting the data to be transmitted according to the slicing information and the routing path; in the data transmission process, optimizing the routing path based on a preset routing optimization policy until the data transmission is completed.

[0132] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0133] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the distributed data transmission optimization method provided by the above methods, and the method includes performing policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization strategy result, and the optimization strategy result includes a routing optimization strategy and a node sleep and wake-up optimization strategy of the Mesh network; according to the optimization strategy result, the data to be transmitted is fragmented and routed to obtain fragmentation information and a routing path; the data to be transmitted is transmitted according to the fragmentation information and the routing path; during the data transmission process, the routing path is optimized based on a preset routing optimization strategy until the data transmission is completed.

[0134] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the distributed data transmission optimization method provided by the above-mentioned methods, the method comprising performing policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization strategy result, wherein the optimization strategy result comprises a routing optimization strategy and a node sleep and wake-up optimization strategy of the Mesh network; slicing and routing the data to be transmitted according to the optimization strategy result to obtain slicing information and a routing path; transmitting the data to be transmitted according to the slicing information and the routing path; during the data transmission process, optimizing the routing path based on a preset routing optimization strategy until the data transmission is completed.

[0135] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A distributed data transmission optimization method, characterized in that: include: Performing policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; According to the optimization strategy result, the data to be transmitted is segmented and routed to obtain segmentation information and routing paths; Performing data transmission on the data to be transmitted according to the sharding information and the routing path; During the data transmission process, the routing path is optimized based on a preset routing optimization strategy until the data transmission is completed.

2. The distributed data transmission optimization method according to claim 1, characterized in that: The optimizing the routing path based on a preset routing optimization strategy includes: Constructing a current graph structure model of the Mesh network, wherein the vertices of the current graph structure model are determined based on the nodes of the Mesh network, the edges of the current graph structure model are determined based on the links between the nodes, and the weights of the edges are determined based on the comprehensive quality of the links; Constructing a graph neural network according to the nodes in the current graph structure model and the relationship between neighboring nodes; Obtaining the current node state of any node in the Mesh network, inputting the current node state into the graph neural network, the graph neural network is used to update the current node state, and outputting the updated current node state to a preset time series prediction model to obtain a predicted node state of any node; The routing path is optimized according to the predicted node status.

3. The distributed data transmission optimization method according to claim 2, characterized in that: Also includes: Obtaining real-time link quality parameters and current transmission rate of any node in the Mesh network, wherein the real-time link quality parameters include signal strength, link packet loss rate, link delay and bandwidth utilization; Determining a comprehensive quality evaluation result of the link according to the link quality parameter, and adjusting the current transmission rate according to the comprehensive quality evaluation result; The routing path is optimized based on the adjusted transmission rate.

4. The distributed data transmission optimization method according to claim 1, characterized in that: The data transmission of the data to be transmitted according to the fragmentation information and the routing path includes: Encoding the slice information based on network coding technology to obtain a coding packet; Distributing the encoded packets to the routing path based on a multi-path transmission technology to achieve data transmission of the data to be transmitted; Correspondingly, the optimizing the routing path based on a preset routing optimization strategy includes: Obtaining real-time link quality parameters and multipath transmission results of any node in the Mesh network; According to the real-time link quality parameter and the multipath transmission result, the redundancy of the coded packet is adjusted based on the erasure coding technology to optimize the routing path.

5. The distributed data transmission optimization method according to any one of claims 2 to 4, characterized in that: Also includes: Obtaining the real-time transmission rate and energy status of any node in the Mesh network; Training a predetermined local model according to the real-time transmission rate and the energy state to obtain update parameters corresponding to the local model; The real-time transmission rate and / or the routing path are optimized according to the update parameters corresponding to the local model.

6. The distributed data transmission optimization method according to any one of claims 1 to 4, characterized in that: The proximal policy optimization algorithm is used to optimize the policy of each node in the Mesh network to obtain the optimized policy result, including: Determine the state space of the node and the dynamic optimization target according to the energy state, communication load, signal quality and neighbor node state of each node in the Mesh network at the current time step; Initializing the state space based on a proximal strategy optimization algorithm, and inputting the initialized state space into a preset time series prediction model to obtain a predicted state space; Substituting the dynamic optimization target into the Lagrange multiplier method to optimize the energy state in the predicted state space to obtain an optimized state space; Input the optimized state space into the node policy network corresponding to the Mesh network, update the node policy network based on the policy gradient method of proximal policy optimization, and obtain the optimal action of the current time step output by the node policy network, wherein the optimal action includes node routing selection and sleep and wake-up operations; Repeat the above steps, combine the optimal actions of all time steps, and get the optimized strategy result.

7. The distributed data transmission optimization method according to any one of claims 1 to 4, characterized in that: The data to be transmitted includes data segments with different data values. Correspondingly, the data to be transmitted is segmented and routed according to the optimization strategy result to obtain segmentation information and routing path, including: Obtaining the node storage capacity, node width, node bandwidth, number of neighbor nodes and communication load corresponding to the optimization strategy result; Substituting the data value of the data fragment, the node storage capacity and the node width into a non-uniform hash function to fragment the data to be transmitted to obtain data fragments; According to the node bandwidth and the number of neighbor nodes, the data shards are allocated to the nodes corresponding to the optimization strategy results based on a hash allocation mechanism to obtain the shard information; The routing selection optimization strategy is optimized according to the node storage capacity, the node width and the communication load to obtain the routing path.

8. The distributed data transmission optimization method according to claim 7, characterized in that: During the data transmission process, optimizing the routing path based on a preset routing optimization strategy until the data transmission is completed includes: During data transmission, obtain the real-time node storage capacity, real-time node width, real-time node bandwidth and real-time communication load; Substituting the real-time node storage capacity and the real-time node width into a non-uniform hash function to slice the remaining data to be transmitted to obtain real-time data slices; According to the real-time node bandwidth and the number of neighbor nodes, the real-time data shards are allocated and optimized based on a hash allocation mechanism to obtain optimized shard information; Performing routing optimization according to the real-time node storage capacity, the real-time node width and the real-time communication load to obtain an optimized routing path; Repeat the above steps until the data transfer is completed.

9. A distributed data transmission optimization device, characterized in that: include: A policy optimization module, used to perform policy optimization on each node of the Mesh network based on a proximal policy optimization algorithm to obtain an optimization policy result, wherein the optimization policy result includes a routing optimization policy and a node sleep and wake-up optimization policy of the Mesh network; A slicing and routing module, used to slice and route the data to be transmitted according to the optimization strategy results, and obtain slicing information and routing paths; A data transmission module, used for transmitting the data to be transmitted according to the sharding information and the routing path; The routing optimization module is used to optimize the routing path based on a preset routing optimization strategy during the data transmission process until the data transmission is completed.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the distributed data transmission optimization method according to any one of claims 1 to 8 is implemented.

Images