A communication transmission control cabinet based on big data

Abstract

The present invention discloses a communication transmission control cabinet based on big data, which relates to the field of communication transmission technology. The present invention includes: introducing a real-time network status perception and prediction module, collecting link data through embedded sensors, combining with an LSTM time series model for prediction, estimating the future link status, introducing a multi-path transmission mechanism, comprehensively considering the bandwidth, delay and load of the path when distributing tasks, uploading in batches through the cache and compression strategy of the edge computing node, effectively reducing the occupancy of the link; when the link performance degrades, the path switching mechanism can select the alternative path with the best performance in real time, avoiding the impact of link interruption on the communication task; at the same time, the high real-time task priority strategy adopts the optimal path, reducing the delay of key tasks; non-real-time tasks are compressed and processed in batches at the edge node, do not compete for resources with real-time tasks, avoid interference, and meet the requirements of ship navigation and real-time monitoring tasks.

Description

Technical Field

[0001] The present invention relates to the field of communication transmission technology, and in particular to a communication transmission control cabinet based on big data. Background Art

[0002] In ship communications, the control cabinet communication transmission requirements are relatively strict. The network environment is highly dynamic. The link quality fluctuates with time and location. The node load is constantly adjusted according to traffic demand. The network topology also changes due to the addition and removal of equipment, which places high demands on the communication system.

[0003] However, existing communication transmission control methods mostly rely on static path planning and regularized protocols, which are unable to fully cope with the complex changes in the dynamic environment, resulting in low network resource utilization and difficulty in ensuring service quality. In ship communications, uneven sea coverage often causes link interruption or performance degradation, making it impossible to quickly adjust the route or switch to a better link.

[0004] To address the above issues, some control cabinet communication solutions use centralized routers or servers, pre-plan transmission paths based on traffic load, and regularly update routing tables to adjust network strategies. Such solutions can work effectively when the network environment is relatively stable, but when the network topology changes frequently, they cannot respond in a timely manner. At the same time, they rely on central nodes for control, which easily forms performance bottlenecks. Some control cabinet communication solutions use retransmission and timeout mechanisms. Although they can ensure data reliability to a certain extent, they will cause large delays and waste network resources. Therefore, there is an urgent need for a communication transmission control cabinet based on big data to solve such problems. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a communication transmission control cabinet based on big data to solve the problem that the static planning method adopted by the traditional control cabinet communication transmission solution cannot adapt to the dynamic network status in time, resulting in the data possibly continuing to select paths with poor performance. Secondly, the centralized control architecture increases the network burden. When the central node fails or is overloaded, it will affect the stability of the entire network.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In one aspect, the present invention provides a communication transmission control cabinet based on big data, which comprises:

[0009] Central storage system, communication system and edge computing nodes,

[0010] The central storage system is used to receive data from the communication system and store the data;

[0011] The communication system includes: network status perception module, dynamic routing control module, task offloading module, data compensation module and transmission performance analysis and optimization module;

[0012] The network status perception module is used to collect and analyze network status data in real time, providing basic information for routing selection and task offloading;

[0013] The dynamic routing control module generates a transmission path based on the data provided by the network status perception module;

[0014] In the dynamic routing control module, a dynamic routing algorithm based on reinforcement learning is used to optimize path selection in real time. A multi-path transmission mechanism is adopted to distribute task data in parallel across multiple links, and a network load balancing strategy is introduced.

[0015] Task offloading module, which performs intelligent task offloading at edge computing nodes;

[0016] Data compensation module, which adjusts transmission strategies in real time and compensates for potential data loss or link interruption during data transmission;

[0017] The transmission performance analysis and optimization module analyzes the results of the communication process and provides feedback.

[0018] Furthermore, in the task distribution module, task data is prioritized according to the transmission path generated by the dynamic routing control module:

[0019] Highly real-time tasks are preferentially transmitted to the target device via the optimal path. The target device refers to the terminal device that directly receives the transmitted data and executes the task, including the navigation system, real-time monitoring equipment and the cabin display screen on the ship;

[0020] Low real-time tasks are cached and compressed in edge computing nodes and then uploaded to the central storage system in batches.

[0021] Furthermore, the data compensation module collaborates with the edge computing node to make adjustments based on real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations. The compensation operations include packet retransmission and data fragmentation reassembly.

[0022] Furthermore, the transmission performance analysis and optimization module collects network status data of the transmission process, namely bandwidth, delay, packet loss rate and node load, analyzes the network status data, and optimizes the algorithm models of the network status perception module and the dynamic routing control module.

[0023] On the other hand, the present invention provides a communication transmission control method based on big data, comprising:

[0024] Step S1: Real-time network status perception and prediction.

[0025] In the network status perception module, embedded sensors are used to collect network status data in communication links and nodes, including bandwidth, latency, packet loss rate, and node load. This data is then combined with an LSTM time series analysis and prediction model trained on historical data to predict future network status changes.

[0026] Step S2, dynamic routing strategy optimization,

[0027] According to the network status estimated in step S1, a dynamic routing algorithm based on reinforcement learning is used to generate a transmission path;

[0028] A multi-path transmission mechanism is used to distribute task data in parallel across multiple links; a network load balancing strategy is introduced during path optimization;

[0029] Step S3, intelligent task diversion,

[0030] Based on the transmission path generated in step S2, the edge computing node acts as the intermediate processing layer of the communication system to offload data tasks:

[0031] Tasks with high real-time requirements are preferentially transmitted to the target device via the optimal path;

[0032] Non-real-time data is cached and compressed by edge computing nodes and then uploaded to the central storage system in batches;

[0033] Step S4: Full-link intelligent control.

[0034] During the data transmission process in step S3, adjustments are made based on real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations, including packet retransmission and data fragmentation reassembly.

[0035] Step S5, analyzing the results of the transmitted data,

[0036] After the data transmission is completed, the network status data is analyzed and the analysis results are fed back to the network status perception module and the dynamic routing control module.

[0037] Furthermore, in step S1, the network status estimation method is:

[0038] Collect network link status data to form a time series input data set Among them, x t,i ={b t，i , d t，i , l t，i， c t,i}, b t，irepresents the available bandwidth of node i at time t, d t，i represents the network delay of node i at time t, l t，i represents the packet loss rate of node i at time t, c t，i Represents the load of node i at time t, and integrates the node state sequence into a time series matrix table,

[0039] Where n is the number of nodes, m = 4 is the characteristic dimension of each node state;

[0040] Perform window processing on the input data, take the data of k historical time steps as input and the data of h future time steps as prediction targets, and generate training sample pairs Among them, X t-k：t represents the input window from time tk to t, Y t+1：t+h Represents the target prediction window from time t+1 to t+h. The training sample input shape is (k, n, m) and the output is (h, n, m).

[0041] Furthermore, in step S1, the network status estimation method further includes:

[0042] Build an LSTM time series model and define the calculation process of the LSTM unit:

[0043] The forget gate determines the proportion of past information to be retained:

[0044] f t =σ(W f ·[h t-1 , X t ]+b f ), where f t Represents the output of the forget gate, with a value range of [0, 1], W f , b f represents the weight matrix and bias vector of the forget gate, h t-1 represents the hidden state vector at time t-1, X t represents the input feature matrix at time t;

[0045] Input gate, which determines the proportion of new information retained in the current time step:

[0046] i t =σ(W i ·[h t-1 , X t ]+b i ), where i t Represents the input gate output, with a value range of [0, 1], W i , b i Represents the weight matrix and bias vector of the input gate;

[0047] Candidate state update, generate candidate memory value:

[0048] in, Represents candidate memory value, range [-1, 1], W c , b c The weight matrix and bias vector representing the candidate state;

[0049] Memory unit update, the memory unit is updated by the forget gate and the input gate:

[0050] Among them, C t Represents the memory state of the current time step, ⊙ represents the element-by-element product operation;

[0051] The output gate determines the impact of the memory state on the current hidden state:

[0052] o t =σ(W o ·[h t-1 , X t ]+b o ), where o t Represents the output of the output gate, with a value range of [0, 1], W o , b o represents the weight matrix and bias vector of the output gate,

[0053] Hidden state update, generating the hidden state vector at time t:

[0054] h t =o t ⊙tanh(C t ), where h t Represents the hidden state of the current time step, ranging from [1, 1],

[0055] Perform model training and prediction, define the loss function, and use mean square error MSE as the loss function:

[0056] Among them, Y t+j,i，k represents the true value, Represents the model prediction value, n represents the number of nodes, h represents the number of prediction steps, and m represents the feature dimension. After training is completed, the model uses the input window X t-k：t Generate future state predictions: in, Represents the predicted future network state matrix.

[0057] Furthermore, in step S2, the dynamic routing strategy optimization method is:

[0058] The dynamic routing problem is modeled as a reinforcement learning task. The router acts as an intelligent agent, whose goal is to select the appropriate path to minimize transmission delay and balance network load. The network state at time t is defined as:

[0059] Among them, B t，i represents the current remaining bandwidth of node i, D t，i represents the current network delay of node i, L t，i Indicates the packet loss rate of node i, with a value range of [0, 1], N t，i represents the current load of node i, and n represents the total number of nodes in the available path;

[0060] Define action a t is the transmission path selected from the candidate path set:

[0061] in, represents the set of all candidate paths, k represents the number of candidate paths;

[0062] Define the reward function r t Measure the performance of the selected path: in, represents the average delay of the selected path, represents the average packet loss rate of the selected path, represents the average bandwidth utilization of the selected path, α, β, and γ represent weight parameters;

[0063] Build a reinforcement learning model and use deep Q learning to reinforce learning, where the Q value function is defined as:

[0064] Among them, Q(s t , a t ) means in state s t Execute action a t The expected cumulative reward, λ represents the discount factor, r t+k represents the immediate reward at time t+k, and E[·] represents the mathematical expectation;

[0065] Get the current and future network states s from the prediction model in step S1 t , using the policy network of the reinforcement learning model, based on the current state s t Select the best path: Update the parameters of the policy network according to the reward function and optimize it using gradient descent:

[0066] Among them, θ t represents the parameters of the Q network, η represents the learning rate, and controls the step size of parameter updates;

[0067] In step S2, a proportional distribution strategy is used to distribute task data to different paths:

[0068] w i represents the weight of task data assigned to path i, B t，i represents the remaining bandwidth of path i, L t，i represents the packet loss rate of path i, and the task distribution amount is T i =w i ·T total , where T i represents the task distribution amount of path i, T total Indicates the total task data volume;

[0069] When the load on a path exceeds the threshold N max , dynamically adjust the distribution ratio:

[0070] Among them, w′ i represents the task allocation weight of path i after adjustment, N t，i Indicates the current load of path i, I(N t，i <N max ) represents the indicator function, which is 1 when the load of path i is lower than the threshold, and 0 otherwise.

[0071] Furthermore, in step S3, the intelligent task diversion method is:

[0072] According to the real-time requirements of the task, the transmission task is divided into high real-time tasks and non-real-time tasks:

[0073] If R i ≥τ, then P i =1,

[0074] If R i <τ, then P i =0, where P i Indicates the priority flag of task i, P i =1 indicates a high real-time task, P i =0 indicates a non-real-time task, R i represents the real-time requirement of task i, τ represents the real-time threshold; the priority division result, in, Represents a set of high real-time tasks, Represents a set of non-real-time tasks;

[0075] In step S3, high-real-time tasks are preferentially transmitted to the target device via the optimal path:

[0076] Path Assignment Model: Tasks Select the optimal path j : Among them, D k represents the current delay of path k, L k represents the packet loss rate of path k, λ represents the weight of the impact of the packet loss rate on the path optimization, Represents the set of all optional paths. If the remaining bandwidth of the path is B k is not enough to accommodate task i, then T i，j =min(T i , B k ), where T i represents the amount of data for task i, T i，j represents the actual transmission amount of task i assigned to path j;

[0077] Non-real-time tasks The edge computing nodes cache, compress and upload the tasks in batches to the central storage system. After the high real-time tasks are completed, their transmission delay and path performance are recorded.

[0078] Furthermore, in step S4, the full-link intelligent control step includes:

[0079] Define the performance index of link j at time t, P t，j =ω1·B t，j -ω2·D t，j -ω3·L t，j , where P t，j It represents the comprehensive performance score of link j at time t. The higher the value, the better the performance. t，j represents the remaining bandwidth of link j, D t，j represents the delay of link j, L t，j represents the packet loss rate of link j, ranging from [0, 1], ω1, ω2, ω3 represent the weights of bandwidth, delay and packet loss rate, and set the minimum link performance threshold P min , when P t，j <P min When , the path switching mechanism is triggered, and the triggering conditions are: Among them, C t Indicates the set of links that need to be switched. Indicates the set of links currently in use;

[0080] For the link that needs to be switched, select the path with the best performance among the alternative paths k : Represents a set of alternative paths, including all non-currently used paths. The path after switching is updated to The link set used after switching;

[0081] In step S4, a data compensation mechanism is introduced:

[0082] Detecting packet loss based on the confirmation mechanism: Among them, Δ i A represents the packet loss data set of task i, i Represents the set of received data packets, T i represents the total data packet set of task i,

[0083] Retransmit lost packets: Among them, R i represents the amount of retransmitted data of task i, size(p) represents the size of data packet p;

[0084] During the path switching process, large data is transmitted in fragments, and the fragment size is calculated as: Among them, S k represents the current fragment size of path k, Δt represents the fragment transmission interval, n represents the number of paths currently used, M max Indicates the maximum amount of data allowed for a single shard;

[0085] After receiving the shards, the edge computing node reassembles the complete data in sequence: Among them, T recv Indicates the complete received data set, which is formed by merging the fragmented data of each path;

[0086] After completing path switching and data compensation, the link performance and compensation results are fed back to steps S1 and S2 to optimize the prediction model and routing strategy. The feedback data includes: Among them, F t Indicates feedback results, recording link performance, retransmission volume, and fragmentation efficiency.

[0087] The beneficial effects of the present invention are:

[0088] The present invention introduces a real-time network status perception and prediction module, collects link data through embedded sensors, combines it with the LSTM time series model for prediction, estimates the future link status, and introduces a multi-path transmission mechanism. When distributing tasks, it comprehensively considers the bandwidth, delay and load of the path, adopts a proportional distribution strategy and load balancing mechanism, ensures that the task traffic is reasonably distributed to each link, and uploads it in batches through the caching and compression strategy of the edge computing node, effectively reducing the occupancy of the link and significantly improving bandwidth utilization.

[0089] In the present invention, when the link performance degrades, the path switching mechanism can select the alternative path with the best performance in real time to avoid the impact of link interruption on communication tasks. At the same time, the high real-time task priority strategy adopts the optimal path to reduce the delay of key tasks. Non-real-time tasks are compressed and batch processed at the edge nodes, without competing for resources with real-time tasks, avoiding interference, and meeting the requirements of ship navigation and real-time monitoring tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0091] Figure 1 This is a schematic diagram of the structure of a communication transmission control cabinet based on big data of the present invention;

[0092] Figure 2 It is a flow chart of the communication transmission control method based on big data of the present invention. DETAILED DESCRIPTION

[0093] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0094] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0095] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0096] Example 1, with reference to Figure 1 This embodiment provides a communication transmission control cabinet based on big data, including: a central storage system, a communication system and an edge computing node.

[0097] The central storage system is used to receive data from the communication system and store the data;

[0098] The communication system includes: network status perception module, dynamic routing control module, task offloading module, data compensation module and transmission performance analysis and optimization module;

[0099] The network status perception module is used to collect and analyze network status data in real time, providing basic information for routing selection and task offloading;

[0100] The dynamic routing control module generates a transmission path based on the data provided by the network status perception module;

[0101] In the dynamic routing control module, a dynamic routing algorithm based on reinforcement learning is used to optimize path selection in real time. A multi-path transmission mechanism is adopted to distribute task data in parallel across multiple links, and a network load balancing strategy is introduced.

[0102] Task offloading module, which performs intelligent task offloading at edge computing nodes;

[0103] In the task distribution module, task data is prioritized according to the transmission path generated by the dynamic routing control module:

[0104] Highly real-time tasks are preferentially transmitted to the target device via the optimal path. The target device refers to the terminal device that directly receives the transmitted data and executes the task, including the navigation system, real-time monitoring equipment and the cabin display screen on the ship;

[0105] Low real-time tasks are cached and compressed in edge computing nodes and then uploaded to the central storage system in batches;

[0106] Data compensation module, which adjusts transmission strategies in real time and compensates for potential data loss or link interruption during data transmission;

[0107] The data compensation module works with edge computing nodes to adjust according to real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations. Compensation operations include packet retransmission and data fragmentation reassembly.

[0108] The transmission performance analysis and optimization module analyzes the communication process and provides feedback;

[0109] The transmission performance analysis and optimization module collects network status data of the transmission process, namely bandwidth, delay, packet loss rate and node load, analyzes the network status data, and optimizes the algorithm model of the network status perception module and the dynamic routing control module.

[0110] Example 2, reference Figure 2 , this embodiment provides a communication transmission control method based on big data, comprising the following steps:

[0111] Step S1: Real-time network status perception and prediction.

[0112] In the network status perception module, embedded sensors are used to collect network status data in communication links and nodes, including bandwidth, latency, packet loss rate, and node load. This data is then combined with an LSTM time series analysis and prediction model trained on historical data to predict future network status changes.

[0113] In step S1, the network status estimation method is:

[0114] Collect network link status data to form a time series input data set Among them, x t，i ={b t，i , d t，i , l t，i , c t，i}, b t，i represents the available bandwidth of node i at time t, d t，i represents the network delay of node i at time t, l t，i represents the packet loss rate of node i at time t, c t，i Represents the load of node i at time t, and integrates the node state sequence into a time series matrix table,

[0115] Where n is the number of nodes, m = 4 is the characteristic dimension of each node state;

[0116] Perform window processing on the input data, take the data of k historical time steps as input and the data of h future time steps as prediction targets, and generate training sample pairs Among them, X t-k：t represents the input window from time tk to t, Y t+1：t+h Represents the target prediction window from time t+1 to t+h, the training sample input shape is (k, n, m), and the output is (h, n, m);

[0117] In step S1, the network status estimation method further includes:

[0118] Build an LSTM time series model and define the calculation process of the LSTM unit:

[0119] The forget gate determines the proportion of past information to be retained:

[0120] f t =σ(W f ·[h t-1 , X t ]+b f ), where f t Represents the output of the forget gate, with a value range of [0, 1], W f , bf represents the weight matrix and bias vector of the forget gate, h t-1 represents the hidden state vector at time t-1, X t represents the input feature matrix at time t;

[0121] Input gate, which determines the proportion of new information retained in the current time step:

[0122] i t =σ(W i ·[h t-1 , X t ]+b i ), where i t Represents the input gate output, with a value range of [0, 1], W i , b i Represents the weight matrix and bias vector of the input gate;

[0123] Candidate state update, generate candidate memory value:

[0124] in, Represents candidate memory value, range [-1, 1], W c , b c The weight matrix and bias vector representing the candidate state;

[0125] Memory unit update, the memory unit is updated by the forget gate and the input gate:

[0126] Among them, C t Represents the memory state of the current time step, ⊙ represents the element-by-element product operation;

[0127] The output gate determines the impact of the memory state on the current hidden state:

[0128] o t =σ(W o ·[h t-1 , X t ]+b o ), where o t Represents the output of the output gate, with a value range of [0, 1], W o , b o represents the weight matrix and bias vector of the output gate,

[0129] Hidden state update, generating the hidden state vector at time t:

[0130] h t =o t ⊙tanh(C t ), where h t Represents the hidden state of the current time step, ranging from [1, 1],

[0131] Perform model training and prediction, define the loss function, and use mean square error MSE as the loss function:

[0132] Among them, Y t+j,i，k represents the true value, Represents the model prediction value, n represents the number of nodes, h represents the number of prediction steps, and m represents the feature dimension. After training is completed, the model uses the input window X t-k：t Generate future state predictions: in, represents the predicted future network state matrix;

[0133] Specifically, the introduction of embedded sensors and LSTM time series analysis and prediction models significantly improves the perception accuracy of network status and the ability to predict future changes. By combining historical and real-time data, it accurately estimates dynamic changes in network bandwidth, latency, packet loss rate, etc., reducing communication delays and resource waste caused by network fluctuations. In scenarios with uneven sea coverage or complex link status, it can perceive possible bottlenecks in advance.

[0134] Step S2, dynamic routing strategy optimization,

[0135] According to the network status estimated in step S1, a dynamic routing algorithm based on reinforcement learning is used to generate a transmission path;

[0136] A multi-path transmission mechanism is used to distribute task data in parallel across multiple links; a network load balancing strategy is introduced during path optimization;

[0137] In step S2, the dynamic routing strategy optimization method is:

[0138] The dynamic routing problem is modeled as a reinforcement learning task. The router acts as an intelligent agent, whose goal is to select the appropriate path to minimize transmission delay and balance network load. The network state at time t is defined as:

[0139] Among them, B t，i represents the current remaining bandwidth of node i, D t，i represents the current network delay of node i, L t，i Indicates the packet loss rate of node i, with a value range of [0, 1], N t，i represents the current load of node i, and n represents the total number of nodes in the available path;

[0140] Define action a t is the transmission path selected from the candidate path set:

[0141] in, represents the set of all candidate paths, k represents the number of candidate paths;

[0142] Define the reward function r t Measure the performance of the selected path: in, represents the average delay of the selected path, represents the average packet loss rate of the selected path, represents the average bandwidth utilization of the selected path, α, β, and γ represent weight parameters;

[0143] Build a reinforcement learning model and use deep Q learning to reinforce learning, where the Q value function is defined as:

[0144] Among them, Q(s t , a t ) means in state s t Execute action a t The expected cumulative reward, λ represents the discount factor, r t+k represents the immediate reward at time t+k, and E[·] represents the mathematical expectation;

[0145] Get the current and future network states s from the prediction model in step S1 t , using the policy network of the reinforcement learning model, based on the current state s t Select the best path: Update the parameters of the policy network according to the reward function and optimize it using gradient descent:

[0146] Among them, θ t represents the parameters of the Q network, η represents the learning rate, and controls the step size of parameter updates;

[0147] In step S2, a proportional distribution strategy is used to distribute task data to different paths:

[0148] w i represents the weight of task data assigned to path i, B t，i represents the remaining bandwidth of path i, L t，i represents the packet loss rate of path i, and the task distribution amount is T i =w i ·T total , where T i represents the task distribution amount of path i, T total Indicates the total task data volume;

[0149] When the load on a path exceeds the threshold N max , dynamically adjust the distribution ratio:

[0150] Among them, w′ i represents the task allocation weight of path i after adjustment, N t，i Indicates the current load of path i, I(N t，i <N max ) represents the indicator function, which is 1 when the load of path i is lower than the threshold, and 0 otherwise;

[0151] Specifically, a dynamic routing strategy optimization method based on reinforcement learning DQN is adopted to make full use of network status prediction information to realize intelligent and dynamic transmission path planning in multi-path selection. The bandwidth, delay and packet loss rate are comprehensively considered through the reward function. The reinforcement learning strategy can automatically adapt to network conditions and optimize the path to reduce delay and make the load distribution more balanced.

[0152] Step S3, intelligent task diversion,

[0153] Based on the transmission path generated in step S2, the edge computing node acts as the intermediate processing layer of the communication system to offload data tasks:

[0154] Tasks with high real-time requirements are preferentially transmitted to the target device via the optimal path;

[0155] Non-real-time data is cached and compressed by edge computing nodes and then uploaded to the central storage system in batches;

[0156] In step S3, the intelligent task diversion method is:

[0157] According to the real-time requirements of the task, the transmission task is divided into high real-time tasks and non-real-time tasks:

[0158] If R i ≥τ, then P i =1,

[0159] If R i <τ, then P i =0, where P i Indicates the priority flag of task i, P i =1 indicates a high real-time task, Pi = 0 indicates a non-real-time task, R i represents the real-time requirement of task i, τ represents the real-time threshold; the priority division result, in, Represents a set of high real-time tasks, Represents a set of non-real-time tasks;

[0160] In step S3, high-real-time tasks are preferentially transmitted to the target device via the optimal path:

[0161] Path Assignment Model: Tasks Select the optimal pathj : Among them, D k represents the current delay of path k, L k represents the packet loss rate of path k, λ represents the weight of the impact of the packet loss rate on the path optimization, Represents the set of all optional paths. If the remaining bandwidth of the path is B k is not enough to accommodate task i, then T i，j =min(T i , B k ), where T i represents the amount of data for task i, T i，j represents the actual transmission amount of task i assigned to path j;

[0162] Non-real-time tasks The edge computing nodes cache, compress, and upload batches of data to the central storage system. After high-real-time tasks are completed, the transmission delay and path performance are recorded.

[0163] Specifically, in the task diversion stage, edge computing nodes are used to intelligently classify and schedule real-time tasks and non-real-time tasks, ensuring that critical tasks are transmitted through the optimal path, while secondary tasks are efficiently uploaded to the central storage system in a cache-compressed manner. The delay of real-time tasks is significantly reduced, and the occupation of network bandwidth by non-real-time tasks is more reasonable, which comprehensively improves the utilization efficiency of communication resources.

[0164] Step S4: Full-link intelligent control.

[0165] During the data transmission process in step S3, adjustments are made based on real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations, including packet retransmission and data fragmentation reassembly.

[0166] In step S4, the full-link intelligent control step includes:

[0167] Define the performance index of link j at time t, P t，j =ω1·B t，j -ω2·D t，j -ω3·L t，j , where P t，j It represents the comprehensive performance score of link j at time t. The higher the value, the better the performance. t，j represents the remaining bandwidth of link j, D t，j represents the delay of link j, L t，j represents the packet loss rate of link j, ranging from [0, 1], ω1, ω2, ω3 represent the weights of bandwidth, delay and packet loss rate, and set the minimum link performance threshold P min , when P t，j<P min When , the path switching mechanism is triggered, and the triggering conditions are: Among them, C t Indicates the set of links that need to be switched. Indicates the set of links currently in use;

[0168] For the link that needs to be switched, select the path with the best performance among the alternative paths k : Represents a set of alternative paths, including all non-currently used paths. The path after switching is updated to The link set used after switching;

[0169] In step S4, a data compensation mechanism is introduced:

[0170] Detecting packet loss based on the confirmation mechanism: Among them, Δ i A represents the packet loss data set of task i, i Represents the set of received data packets, T i represents the total data packet set of task i,

[0171] Retransmit lost packets: Among them, R i represents the amount of retransmitted data of task i, size(p) represents the size of data packet p;

[0172] During the path switching process, large data is transmitted in fragments, and the fragment size is calculated as: Among them, S k represents the current fragment size of path k, Δt represents the fragment transmission interval, n represents the number of paths currently used, M max Indicates the maximum amount of data allowed for a single shard;

[0173] After receiving the shards, the edge computing node reassembles the complete data in sequence: Among them, T recv Indicates the complete received data set, which is formed by merging the fragmented data of each path;

[0174] After completing path switching and data compensation, the link performance and compensation results are fed back to steps S1 and S2 to optimize the prediction model and routing strategy. The feedback data includes: Among them, F t Indicates feedback results, recording link performance, retransmission volume, and fragmentation efficiency;

[0175] Specifically, during data transmission, the transmission strategy is dynamically adjusted according to the real-time changes in link performance, and path switching and data compensation mechanisms are introduced. When link performance degrades, path switching can be quickly triggered, and data integrity and transmission continuity are guaranteed through packet retransmission and data fragmentation reassembly, avoiding data loss caused by network interruption.

[0176] Step S5, analyzing the results of the transmitted data,

[0177] After the data transmission is completed, the network status data is analyzed and the analysis results are fed back to the network status perception module and the dynamic routing control module.

[0178] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A communication transmission control cabinet based on big data, characterized by: include, Central storage system, communication system and edge computing nodes, The central storage system is used to receive data from the communication system and store the data; The communication system includes: network status perception module, dynamic routing control module, task offloading module, data compensation module and transmission performance analysis and optimization module; The network status perception module is used to collect and analyze network status data in real time, providing basic information for routing selection and task offloading; The dynamic routing control module generates a transmission path based on the data provided by the network status perception module; In the dynamic routing control module, a dynamic routing algorithm based on reinforcement learning is used to optimize path selection in real time. A multi-path transmission mechanism is adopted to distribute task data in parallel across multiple links, and a network load balancing strategy is introduced. Task offloading module, which performs intelligent task offloading at edge computing nodes; Data compensation module, which adjusts transmission strategies in real time and compensates for potential data loss or link interruption during data transmission; The transmission performance analysis and optimization module analyzes the communication process and provides feedback; The data compensation module works with edge computing nodes to adjust according to real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations. Compensation operations include packet retransmission and data fragmentation reassembly. The link performance control step includes: Define the performance index of link j at time t, P t,j =ω1·B t,j -ω2·D t,j -ω3·L t,j , where P t,j It represents the comprehensive performance score of link j at time t. The higher the value, the better the performance. t,j represents the remaining bandwidth of link j, D t,j represents the delay of link j, L t,j represents the packet loss rate of link j, ranging from [0,1], ω1, ω2, ω3 represent the weights of bandwidth, delay and packet loss rate, and set the minimum link performance threshold P min , when P t,j <P min When , the path switching mechanism is triggered, and the triggering conditions are: Among them, C t Indicates the set of links that need to be switched. Indicates the set of links currently in use; For the link that needs to be switched, select the path with the best performance among the alternative paths k : Represents a set of alternative paths, including all non-currently used paths. The path after switching is updated to The link set used after switching; In step S4, a data compensation mechanism is introduced: Detecting packet loss based on the confirmation mechanism: Among them, Δ i A represents the packet loss data set of task i, i Represents the set of received data packets, T i represents the total data packet set of task i, Retransmit lost packets: Among them, R i represents the amount of retransmitted data of task i, size(p) represents the size of data packet p; During the path switching process, large data is transmitted in fragments. The fragment size is calculated as: Among them, S k represents the current fragment size of path k, Δt represents the fragment transmission interval, n represents the number of paths currently used, M max Indicates the maximum amount of data allowed for a single shard; After receiving the shards, the edge computing node reassembles the complete data in sequence: Among them, T recv Indicates the complete received data set, which is formed by merging the fragmented data of each path; After completing path switching and data compensation, the link performance and compensation results are fed back to steps S1 and S2 to optimize the prediction model and routing strategy. The feedback data includes: Among them, F t Indicates feedback results, recording link performance, retransmission volume, and fragmentation efficiency.

2. A communication transmission control cabinet based on big data according to claim 1, characterized in that: In the task distribution module, task data is prioritized according to the transmission path generated by the dynamic routing control module: Highly real-time tasks are preferentially transmitted to the target device via the optimal path. The target device refers to the terminal device that directly receives the transmitted data and executes the task, including the navigation system, real-time monitoring equipment and the cabin display screen on the ship; Low real-time tasks are cached and compressed in edge computing nodes and then uploaded to the central storage system in batches.

3. A communication transmission control cabinet based on big data according to claim 2, characterized in that: The transmission performance analysis and optimization module collects network status data of the transmission process, namely bandwidth, delay, packet loss rate and node load, analyzes the network status data, and optimizes the algorithm model of the network status perception module and the dynamic routing control module.

4. A communication transmission control method based on big data, characterized in that: A communication transmission control cabinet based on big data according to any one of claims 1 to 3, comprising the following steps: Step S1: Real-time network status perception and prediction. In the network status perception module, embedded sensors are used to collect network status data in communication links and nodes, including bandwidth, latency, packet loss rate, and node load. The data is then combined with the LSTM time series analysis and prediction model trained with historical data to estimate future network status changes. Step S2, dynamic routing strategy optimization, based on the network status estimated in step S1, a dynamic routing algorithm based on reinforcement learning is used to generate a transmission path; A multi-path transmission mechanism is used to distribute task data in parallel across multiple links; a network load balancing strategy is introduced during path optimization; Step S3, intelligent task offloading, based on the transmission path generated in step S2, the edge computing node acts as the intermediate processing layer of the communication system to offload data tasks: Tasks with high real-time requirements are preferentially transmitted to the target device via the optimal path; Non-real-time data is cached and compressed by edge computing nodes and then uploaded to the central storage system in batches; Step S4: Full-link intelligent control. During the data transmission process of step S3, adjustments are made based on real-time changes in link performance. If the performance of a transmission link degrades, path switching is immediately triggered, and the edge computing node performs data compensation operations, including packet retransmission and data fragmentation reassembly. Step S5, analyzing the results of the transmission data, after the data transmission is completed, analyzing the network status data, and feeding back the analysis results to the network status perception module and the dynamic routing control module.

5. The communication transmission control method based on big data according to claim 4, characterized in that: In step S1, the network status estimation method is: Collect network link status data to form a time series input data set Among them, x t,i ={b t,i ,d t,i ,l t,i ,c t,i }, b t,i represents the available bandwidth of node i at time t, d t,i represents the network delay of node i at time t, l t,i represents the packet loss rate of node i at time t, c t,i Represents the load of node i at time t, and integrates the node state sequence into a time series matrix table, Where n is the number of nodes, m = 4 is the characteristic dimension of each node state; Perform window processing on the input data, take the data of k historical time steps as input and the data of h future time steps as prediction targets, and generate training sample pairs Among them, X t-h:t represents the input window from time tk to t, Y t+1:t+h Represents the target prediction window from time t+1 to t+h. The training sample input shape is (k,n,m) and the output is (h,n,m).

6. The communication transmission control method based on big data according to claim 5, characterized in that: In step S1, the network status estimation method further includes: Build an LSTM time series model and define the calculation process of the LSTM unit: The forget gate determines the proportion of past information to be retained: f t =σ(W f ·[h t-1 ,X t ]+b f ), where f t Represents the output of the forget gate, with a value range of [0,1], W f ,b f represents the weight matrix and bias vector of the forget gate, h t-1 represents the hidden state vector at time t-1, X t represents the input feature matrix at time t; Input gate, which determines the proportion of new information retained in the current time step: i t =σ(W i ·[h t-1 ,X t ]+b i ), where i t Represents the input gate output, with a value range of [0,1], W i ,b i Represents the weight matrix and bias vector of the input gate; Candidate state update, generate candidate memory value: in, Represents candidate memory value, range [-1,1], W c ,b c The weight matrix and bias vector representing the candidate state; Memory unit update, the memory unit is updated by the forget gate and the input gate: Among them, C t Represents the memory state of the current time step, ⊙ represents the element-by-element product operation; The output gate determines the impact of the memory state on the current hidden state: o t =σ(W o ·[h t-1 ,X t ]+b o ), where o t Represents the output of the output gate, with a value range of [0,1], W o ,b o represents the weight matrix and bias vector of the output gate, Hidden state update, generating the hidden state vector at time t: h t =o t ⊙tanh(C t ), where h t Represents the hidden state of the current time step, range [1,1], Perform model training and prediction, define the loss function, and use mean square error MSE as the loss function: Among them, Y t+j,i,k represents the true value, Represents the model prediction value, n represents the number of nodes, h represents the number of prediction steps, and m represents the feature dimension. After training is completed, the model uses the input window X t-k:t Generate future state predictions: in, Represents the predicted future network state matrix.

7. The communication transmission control method based on big data according to claim 6, characterized in that: In step S2, the dynamic routing strategy optimization method is: The dynamic routing problem is modeled as a reinforcement learning task. The router acts as an intelligent agent, whose goal is to select the appropriate path to minimize transmission delay and balance network load. The network state at time t is defined as: Among them, B t,i represents the current remaining bandwidth of node i, D t,i represents the current network delay of node i, L t,i Indicates the packet loss rate of node i, with a value range of [0,1], N t,i represents the current load of node i, and n represents the total number of nodes in the available path; Define action a t is the transmission path selected from the candidate path set: in, represents the set of all candidate paths, k represents the number of candidate paths; Define the reward function r t Measure the performance of the selected path: in, represents the average delay of the selected path, represents the average packet loss rate of the selected path, represents the average bandwidth utilization of the selected path, α, β, and γ represent weight parameters; Build a reinforcement learning model and use deep Q learning to reinforce learning, where the Q value function is defined as: Among them, Q(s t ,a t ) means in state s t Execute action a t The expected cumulative reward, λ represents the discount factor, r t+k represents the immediate reward at time t+k, and E[·] represents the mathematical expectation; Get the current and future network states s from the prediction model in step S1 t , using the policy network of the reinforcement learning model, based on the current state s t Select the best path: Update the parameters of the policy network according to the reward function and optimize it using gradient descent: Among them, θ t represents the parameters of the Q network, η represents the learning rate, and controls the step size of parameter updates; In step S2, a proportional distribution strategy is used to distribute task data to different paths: w i represents the weight of task data assigned to path i, B t,i represents the remaining bandwidth of path i, L t,i represents the packet loss rate of path i, and the task distribution amount is T i =w i ·T total , where T i represents the task distribution amount of path i, T total Indicates the total task data volume; When the load on a path exceeds the threshold N max , dynamically adjust the distribution ratio: Among them, w ' i represents the task allocation weight of path i after adjustment, N t,i Indicates the current load of path i, I(N t,i <N max ) represents the indicator function, which is 1 when the load of path i is lower than the threshold, and 0 otherwise.

8. The communication transmission control method based on big data according to claim 7, characterized in that: In step S3, the intelligent task diversion method is: According to the real-time requirements of the task, the transmission task is divided into high real-time tasks and non-real-time tasks: If R i ≥τ, then P i =1, If R i <τ, then P i =0, where P i Indicates the priority flag of task i, P i =1 indicates a high real-time task, P i =0 indicates a non-real-time task, R i represents the real-time requirement of task i, τ represents the real-time threshold; the priority division result, in, Represents a set of high real-time tasks, Represents a set of non-real-time tasks; In step S3, high-real-time tasks are preferentially transmitted to the target device via the optimal path: Path Assignment Model: Tasks Select the optimal path j : Among them, D k represents the current delay of path k, L k represents the packet loss rate of path k, λ represents the weight of the impact of the packet loss rate on the path optimization, Represents the set of all optional paths. If the remaining bandwidth of the path is B k is not enough to accommodate task i, then T i,j =min(T i ,B k ), where T i represents the amount of data for task i, T i,j represents the actual transmission amount of task i assigned to path j; Non-real-time tasks The edge computing nodes cache, compress and upload the tasks in batches to the central storage system. After the high real-time tasks are completed, their transmission delay and path performance are recorded.

Images