5g network optimization method and system based on intelligent dynamic multi-band adaptive switching system

Abstract

The present application relates to the field of vehicle-mounted communication technology, and provides a 5G network optimization method and system based on an intelligent dynamic multi-frequency band adaptive switching system, which comprises the following steps: acquiring network signal quality data, environment perception data and a communication task demand set; extracting dynamic obstacle features based on the environment perception data, and outputting signal shielding prediction through a double-branch hybrid deep learning model; screening candidate links that pass the inspection to build a candidate set, generating dynamic weights using a deep Q network, and calculating the comprehensive scores of the candidate links, determining the main communication link under the hard constraint condition, and generating a communication strategy instruction; according to the instruction type of the communication strategy instruction, performing dynamic fragmentation and multi-path distribution on the data stream, performing verification, reorganization and intelligent delivery at the receiving end, and collecting feedback data packets for network closed-loop feedback. The present application combines multi-modal perception and signal shielding prediction, and constructs an adaptive switching and collaborative transmission mechanism for multi-frequency band communication.

Description

Technical Field

[0001] This invention relates to the field of vehicle communication technology, and in particular to a 5G network optimization method and system based on an intelligent dynamic multi-band adaptive switching system. Background Technology

[0002] With the deepening application of 5G technology in intelligent transportation and autonomous driving, remote vehicle control has become a core technology for achieving high-level autonomous driving and unmanned operations. In complex and ever-changing electromagnetic environments, how to provide continuous, stable, and low-latency communication guarantees for remote control commands and high-definition video transmission, and effectively compensate for the shortcomings of traditional single-path communication solutions such as easy interruption and high latency, has become an important issue that urgently needs to be addressed in promoting the large-scale commercialization of vehicle remote control technology.

[0003] Chinese patent application CN120264315A discloses a network slicing optimization method for 5G private networks. This method includes: constructing a time-series performance model during slice operation to monitor minute changes in end-to-end latency, throughput, and packet loss rate; extracting short-cycle fluctuation patterns of the indicators and comparing gradients with historical stability thresholds; identifying indicator drift behavior with continuous jitter but no default to determine if the system is in a performance critical state; recording the non-proportional response resulting from each resource adjustment, including jitter improvement and latency reduction; performing trend fitting on the data to determine if there is a performance plateau or critical jump point; defining three intervals for resource investment in the model: an ineffective zone with unchanged investment, a highly sensitive zone with rapidly changing performance, and a saturated zone where performance is stable but resources continue to increase; identifying the current response zone of the slice through the model, and deciding that the control strategy should not focus on proportional investment but rather on the optimal response zone as the resource allocation interval.

[0004] However, current technology still faces many challenges. The propagation capability of 5G high-frequency signals is limited at the physical level, especially in densely populated urban areas or complex terrain. Their diffraction and penetration capabilities are weak, making them highly susceptible to obstruction by buildings or mountains, leading to frequent communication link interruptions. Existing systems mostly employ passive link switching mechanisms, typically triggering backup links only after the current signal quality has severely deteriorated. This lack of forward-looking prediction of future channel conditions results in switching delays or even failures. Furthermore, their switching decision logic is simplistic, failing to dynamically and comprehensively balance the real-time quality of each candidate link with differentiated service requirements, such as latency and bandwidth. This makes it difficult to achieve globally optimal scheduling, leading to lost remote control commands or video transmission stutters, posing a serious threat to driving safety. Summary of the Invention

[0005] To achieve the above objectives, this invention provides a 5G network optimization method based on an intelligent dynamic multi-band adaptive handover system, the specific technical solution of which is as follows:

[0006] A 5G network optimization method and system based on an intelligent dynamic multi-band adaptive handover system includes:

[0007] Based on an integrated vehicle-mounted multi-band device, a multi-source sensor network, a positioning system, and an application perception and recognition module, real-time network signal quality data P is acquired synchronously. sig Environmental perception data D env and the communication task requirement set R task ;

[0008] Based on environmental perception data D env Extract the set of dynamic obstacle feature vectors F obs Based on the environmental perception data D env and the set of dynamic obstacle feature vectors F obs The data is then integrated and fed into a parallel, dual-branch hybrid deep learning model M. DL Perform signal occlusion prediction and output the signal occlusion prediction results with multi-dimensional parameters. pred ;

[0009] The acquired candidate links are sequentially subjected to signal link feasibility checks and network resource availability checks to select qualified candidate links and construct a candidate set C for multi-dimensional performance profiles. set Based on the multidimensional performance profiles of each candidate link and combined with a dual-mode decision triggering mechanism, a deep Q-network is used to generate dynamic weight factors when the communication decision process is activated. The comprehensive score S of each candidate link is then calculated based on these dynamic weight factors. score Based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate links, sorted in descending order, are checked against hard constraints one by one until a candidate link that meets the hard constraints is found to be the primary communication link. Based on the overall network resource situation, a choice is made between switching policy instructions and coordination policy instructions to generate the final communication policy instruction, Start. comm ;

[0010] According to the communication strategy instruction Start comm Perform instruction type determination; if the communication strategy instruction Start... comm For a collaborative strategy, dynamic fragmentation is performed on the data stream to generate data packets P that can be transmitted concurrently. sliced At the receiving end, the data packet P... sliced Cyclic redundancy check, out-of-order reordering, and intelligent delivery are performed sequentially to restore the complete and ordered data stream D. reass ; in the communication strategy instruction Start comm After execution, a feedback data packet F containing the key performance indicators of the instructions is collected and encapsulated.back and the feedback data packet F back Iterative training of deep Q-networks is used to update decision weights and achieve adaptive closed-loop feedback.

[0011] Furthermore, the signal occlusion prediction result O pred The output methods include:

[0012] Dynamic obstacle data D sensor Multi-stage processing is performed to generate a set of dynamic obstacle feature vectors F containing morphological and motion parameters. obs ;

[0013] static environment data D map Vehicle dynamic data D pos and the set of dynamic obstacle feature vectors F obs The data is then integrated and fed into a parallel dual-branch hybrid deep learning model to extract spatial feature vectors F. cnn With time eigenvector F rnn Based on the spatial feature vector F cnn With time eigenvector F rnn Fusion is performed to construct a spatiotemporal fusion feature vector F fused The spatiotemporal fusion feature vector F is combined with the prediction decoding head. fused Decode the signal and output the predicted affected frequency band bank. k Predicted signal attenuation value A k The predicted start time of the occlusion event And the predicted duration of the occlusion event. Signal occlusion prediction results O pred .

[0014] Furthermore, the set F of dynamic obstacle feature vectors with morphological and motion parameters obs The generation includes morphological feature extraction and trajectory feature extraction;

[0015] Based on dynamic obstacle data D sensor The convex hull algorithm is used for morphological feature extraction to determine the length l, width w, and height h of the dynamic obstacle. The data points are then further divided into sub-cubes, and the data point density ρ of each sub-cube is determined accordingly. i' Perform a weighted summation to estimate the total volume V of the obstacle;

[0016] The Kalman filter algorithm is used to extract the trajectory features of the centroid of the same obstacle identified in consecutive frames. By calculating the position change of the centroid between consecutive time steps, the instantaneous velocity s and instantaneous motion direction angle θ of the obstacle are estimated. s .

[0017] Furthermore, the communication strategy instruction Start comm The generation methods include:

[0018] Obtain the candidate link list and network load rate based on real-time network signal quality data P. Sig Perform a signal link feasibility check on the candidate link list, and perform a network resource availability check on the candidate link list based on the network load rate. After completing the feasibility and availability checks, construct a candidate set C from all remaining qualified candidate links. set And based on the candidate set C set The generation of each candidate link includes signal quality index N sig Network load metric N load and distance index N dis Multidimensional performance profile;

[0019] Based on real-time network signal quality data P Sig And signal occlusion prediction results O pred The system combines a dual-mode decision triggering mechanism to determine whether to activate the communication decision process. If the communication decision process is activated, a deep Q-network is used to generate dynamic weight factors based on the state vector of the current communication environment. The comprehensive score S of each candidate link is then calculated based on the weight factors and the multi-dimensional performance profile of each candidate link. score ;

[0020] Based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate set C sorted in descending order set The candidate link ranked first is subjected to hard constraint testing. If the candidate link meets the hard constraints, it is determined as the main communication link; if the candidate link does not meet the hard constraints, the process is automatically iterated to the candidate set C after descending order. set The next candidate link is then subjected to hard constraint checks until a candidate link that meets the hard constraints is found. Based on the overall network resource situation, a choice is made between switching policy instructions and coordination policy instructions to generate the final communication policy instruction, Start. comm ;

[0021] The communication service requirement R task The hard constraint is the communication service demand R task Performance requirements of the critical data stream j of the highest priority data stream task in the middle business Including minimum protection width and maximum transmission delay

[0022] Furthermore, the dual-mode decision triggering mechanism is activated when any of the following conditions are met: if the real-time network signal quality data P Sig The display shows that the key performance indicators of the currently used link are below the preset minimum availability threshold, indicating that the service quality has deteriorated unacceptably, thus activating the passive response trigger mechanism; if the signal obstruction prediction result is O... pred The system indicates that when the current communication link is about to face a high probability and high attenuation signal blockage event in the near future, it means that the system anticipates the impending deterioration in service quality and activates the proactive prevention trigger mechanism.

[0023] Furthermore, the method for choosing between the switching strategy instruction and the cooperative strategy instruction includes:

[0024] When the system determines that the current network environment is relatively tense, that is, the quality of other candidate links is generally poor except for the selected main communication link, or when the main purpose of this decision is to avoid the deterministic failure risk of the current link, the system will start the switching strategy instruction generation process.

[0025] When the system determines that the current network environment has sufficient resources, that is, in addition to the already determined main communication link, the candidate set C set There are also several other high-quality links in the process, and the system will initiate a collaborative strategy instruction generation process.

[0026] The handover policy instruction includes the policy instruction type 'HANDOVER' and the target frequency band of the finally selected primary communication link. band The parameters and the reason parameter that triggered this link switch are recorded;

[0027] The collaborative strategy instructions include the strategy instruction type 'SYNERGY' and the control task 'control'. task Video task task and telemetry task .

[0028] Furthermore, the process of generating the collaborative strategy instructions includes:

[0029] Control the highest priority control task task Assign it to the established main communication link and write it as the first key-value pair in the parameter dictionary;

[0030] Create two temporary resource pools: a task pool and an available link pool. The task pool contains communication service requirements R. task Other business flows besides the allocated critical business flows, and sorted by business priority. Sort from highest to lowest, and the available link pool contains candidate set C. setOther high-quality candidate links besides the main communication links that are already occupied are ranked according to their comprehensive score S. score Sort from highest to lowest;

[0031] Prioritize from the task pool The highest-priority business is processed, and the available link pool is traversed sequentially to find the best matching link.

[0032] Furthermore, the optimal matching link includes three scenarios:

[0033] If a single link satisfies the scenario, and the current pending business has a comprehensive score S in the available link pool. score When the performance of the highest-performing link meets the core requirements of the service, the link is allocated to the service, a corresponding key-value pair is added to the parameter dictionary, and the service and the link are removed from their respective resource pools to ensure that resources are not allocated repeatedly.

[0034] In link aggregation scenarios, if no single link in the available link pool can meet the current high-bandwidth service requirements, the system further determines whether the service requirements can be met by aggregating the bandwidth of multiple links.

[0035] In link reuse scenarios, after completing the link allocation for high-priority tasks, the system continues to process lower-priority tasks. In addition to considering unused link resources in the available link pool, it also intelligently assesses whether there are allocated links with remaining capacity that can be reused.

[0036] Furthermore, the method for implementing the adaptive closed-loop feedback includes:

[0037] According to the communication strategy instruction Start comm If the instruction type is a cooperative strategy, then the optimal fragment size S is dynamically calculated based on the real-time available bandwidth B of each link and the target transmission time window T. size And based on the optimal fragment size S size Dynamically fragment the data stream to generate data packets P that can be transmitted in parallel across multiple links. sliced If the instruction type is a switching strategy, then skip the dynamic sharding process.

[0038] For the received data packet P sliced Cyclic redundancy check (CRC) is performed. Data packets that pass the check are stored in a circular buffer based on their sequence numbers for out-of-order reassembly. A bidirectional scanning reassembly algorithm is then used to intelligently decode and deliver the data packets in the circular buffer sequentially, thereby ensuring the received data packet P is delivered. sliced Restored to a complete and ordered data stream D reass ;

[0039] Execute the communication policy instruction Start comm Then, the communication strategy instruction Start is executed. comm The execution effect is evaluated, and the key performance indicators collected in the evaluation are encapsulated into a feedback data packet F. back and the feedback data packet F back The samples are used as training data for iterative training of the deep Q-network to update the decision weights and achieve adaptive closed-loop feedback.

[0040] The 5G network optimization system based on the intelligent dynamic multi-band adaptive handover system is used to implement the above-mentioned 5G network optimization method based on the intelligent dynamic multi-band adaptive handover system, including: a multi-source heterogeneous data acquisition and sensing module, an intelligent signal blockage prediction module, a communication strategy generation module, and a closed-loop feedback optimization module.

[0041] The multi-source heterogeneous data acquisition and sensing module is used to acquire real-time network signal quality data P. sig Environmental perception data D env and the communication task requirement set R task ;

[0042] The intelligent signal obstruction prediction module is based on environmental perception data D. env Extract the set of dynamic obstacle feature vectors F obs Based on the environmental perception data D env and the set of dynamic obstacle feature vectors F obs The data is integrated and fed into a parallel dual-branch hybrid deep learning model for signal occlusion prediction, outputting a multi-dimensional parameter signal occlusion prediction result. pred ;

[0043] The communication strategy instruction generation module is used to sequentially perform signal link feasibility checks and network resource availability checks on the acquired candidate links, and select qualified candidate links to construct a candidate set C of multi-dimensional performance profiles. set Based on the multidimensional performance profiles of each candidate link, and combined with a dual-mode decision triggering mechanism, a deep Q-network is used to generate dynamic weight factors when the communication decision process is activated. The comprehensive score S of each candidate link is then calculated based on these dynamic weight factors. score Based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate links, sorted in descending order, are checked against hard constraints one by one until a candidate link that meets the hard constraints is found to be the primary communication link. Based on the overall network resource situation, a choice is made between switching policy instructions and cooperative communication policy instructions to generate the final communication policy instruction, Start. comm ;

[0044] The closed-loop feedback optimization module is used to perform optimization based on the communication strategy instruction Start. comm Perform instruction type determination; if the communication strategy instruction Start... comm When using a collaborative strategy, dynamic fragmentation is performed on the data stream to generate data packets P that can be transmitted concurrently. sliced At the receiving end, the data packet P... sliced Cyclic redundancy check, out-of-order reordering, and intelligent delivery are performed sequentially to restore the complete and ordered data stream D. reass ; in the communication strategy instruction Start comm After execution, a feedback data packet F containing the key performance indicators of the instructions is collected and encapsulated. back and the feedback data packet F back Iterative training of deep Q-networks is used to update decision weights and achieve adaptive closed-loop feedback.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] This invention improves upon the problem that traditional decision-making systems rely solely on a single network quality indicator, leading to blind decision-making and a disconnect from actual needs due to a lack of comprehensive contextual awareness, by simultaneously collecting multi-dimensional heterogeneous information such as network channels, physical environment, and upper-layer service requirements.

[0047] This application improves the problem of traditional systems relying solely on lagging network indicators and resulting in delayed handover decisions and passive responses due to the inability to anticipate physical obstructions by constructing a digital twin model to make forward-looking predictions of signal blockage.

[0048] This application uses real-time network quality, environmental awareness data, communication service requirements, and forward-looking blockage prediction as comprehensive decision-making factors to dynamically optimize the selection strategy for multi-band communication links. This avoids the risk of traditional methods relying solely on a single optimal indicator, which could lead to the selected links being mismatched with critical service requirements or about to fail.

[0049] This application dynamically segments and schedules the data stream to be transmitted in parallel based on communication strategy instructions and the real-time bandwidth of each link, thus avoiding the transmission bottleneck or resource idleness problems caused by the inability to aggregate the bandwidth of multiple links in the traditional single-link transmission mode. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the principle of the 5G network optimization method based on the intelligent dynamic multi-band adaptive switching system of the present invention.

[0052] Figure 2 This is a functional block diagram of the 5G network optimization system based on the intelligent dynamic multi-band adaptive switching system of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1

[0055] Please see Figure 1 As shown, this embodiment provides a 5G network optimization method based on an intelligent dynamic multi-band adaptive handover system, including:

[0056] The S1000, based on an integrated vehicle-mounted multi-band device, a multi-source sensor network, a positioning system, and an application perception and recognition module, synchronously acquires real-time network signal quality data P. sig Environmental perception data D env and the communication task requirement set R task .

[0057] Specifically, this step aims to provide a high-quality data input foundation for subsequent prediction and decision-making. By collecting diverse information from the vehicle itself, the surrounding environment, and real-time network channels, and standardizing it, this step addresses the issues of incomplete information from a single data source and inconsistent spatiotemporal benchmarks among multiple data sources. It serves as the starting point for the entire closed-loop optimization system.

[0058] Further, step S1000 includes:

[0059] Step S1100: Monitor the radio frequency signal of the vehicle-mounted integrated multi-band device in real time, preprocess the radio frequency signal, and output structured real-time network signal quality data P. sig .

[0060] Specifically, this step involves collecting and outputting real-time network signal quality data P through an in-vehicle integrated multi-band communication device. Sig This provides a reliable data foundation for subsequent steps.

[0061] The in-vehicle integrated multi-band communication device integrates millimeter-wave band (24.25GHz-52.6GHz), Sub-6GHz band (3.3GHz-5GHz), and 4G LTE band communication modules. Each communication module is connected to the vehicle's central control system via a high-speed PCIe data bus, ensuring that the large amount of data generated concurrently across multiple bands can be processed without bottlenecks.

[0062] In practice, the highly integrated multi-band communication device within the vehicle terminal uses a built-in signal processing unit to continuously monitor the radio frequency signals of all available communication bands—millimeter-wave, Sub-6GHz, and 4G LTE—at a high sampling frequency of millions of times per second. This high-precision signal monitoring detects the communication link status in real time and accurately measures key signal quality parameters for each band. Simultaneously, to ensure data purity and reliability, the signal processing unit utilizes bandpass filters and adaptive equalizers to preprocess the original signal, effectively filtering out noise interference and compensating for signal attenuation and distortion during transmission.

[0063] After the above measurement and processing procedures, a structured real-time network signal quality data P is finally output. Sig ={Q i |i∈B set}; where P sig Q represents the total set of network quality information collected at a certain moment, encompassing all available communication frequency bands; it is a set, namely, a real-time network signal quality data set. i ={RSSI i SNR i BER i} represents a tuple of quality parameters for frequency band i, containing several key indicators describing the current communication quality of that frequency band, namely the Received Signal Strength (RSSI). i Signal-to-noise ratio (SNR) i and Bit Error Rate (BER) i RSSI i SNR (Signal Response Rate) represents the signal power received by the vehicle-mounted terminal antenna from the base station. It is the most direct and important indicator for assessing network coverage, measured in decibels (dBm). A higher value indicates a stronger signal and better network coverage. i BER (Block Received Power) represents the ratio of the received useful signal power to the background noise power present in the same channel, reflecting the signal clarity. The unit is decibels (dB). A higher BER indicates that the signal is less susceptible to noise interference and that data transmission reliability is higher. iThis represents the proportion of erroneous bits decoded by the receiver within a certain time period to the total number of transmitted bits. It is used to measure the accuracy of the data transmission link; the lower the value, the higher the data transmission quality. 'i' represents an identifier for a specific communication frequency band; B set ={mmWave,Sub-6G,4G-LTE} represents the set of all detectable communication frequency bands in the current vehicle environment, including the millimeter wave band mmWave, the Sub-6G band, and the 4G LTE band.

[0064] Step S1200: Based on the vehicle-mounted multi-source sensor network and positioning system, collect static environmental data D. map Vehicle dynamic data D pos and dynamic obstacle data D sensor Based on the static environmental data D map Vehicle dynamic data D pos and dynamic obstacle data D sensor Constructing environmental perception data D env .

[0065] Specifically, this step involves constructing a high-fidelity digital twin environment that is synchronized with the physical world in real time through an onboard multi-source heterogeneous sensor network and a high-precision positioning system.

[0066] In the specific implementation process, key perception data is collected from three dimensions, as follows:

[0067] I. Static Environment Data D map Data collection. A pre-set high-precision map database is loaded and invoked. This database has centimeter-level spatial resolution and comprehensively covers key elements such as road geometry, lane information, and the location of traffic signs and markings. More importantly, the database records detailed three-dimensional spatial information of various static facilities along the route that may affect signal propagation, including their length, width, height, and outline features, covering typical structures such as buildings, bridges, and tunnels.

[0068] II. Vehicle's own dynamic data D pos Data Acquisition: By fusing the Global Positioning System (GPS) and the Inertial Navigation System (INS), real-time perception of the vehicle's dynamic state information is achieved. This fusion scheme ensures that even in complex environments where GPS signals are weak or interrupted (such as tunnels, urban canyons, etc.), stable and reliable position and attitude information can still be continuously calculated using the INS. The system achieves sub-meter positioning accuracy and a speed measurement error within ±0.1 m / s, demonstrating high dynamic response capability and adaptability to complex environments.

[0069] III. Dynamic Obstacle Data D sensorData Acquisition: An environmental perception sensor array is deployed on the vehicle roof or at high points along the route to continuously collect information on dynamic obstacles in the surrounding environment. The sensor array includes a 128-line mechanical LiDAR with 360° omnidirectional scanning capability, a sampling frequency of 10Hz, and a maximum detection distance of up to 200 meters; and an industrial-grade binocular camera with a baseline distance of 15cm, supporting the acquisition of high-quality images at a resolution of 1280×960 at a frame rate of 30fps.

[0070] Finally, the three types of data collected above are integrated to form a multi-layered, structured environmental perception data D. env ={D map D pos D sensor}; where D env This refers to the total collection of multimodal data collected at a specific moment, comprehensively describing the vehicle's own state and its surrounding environment; it is a set, namely, environmental perception data; D map This represents static high-precision map data used to predict signal occlusion caused by certain fixed objects (such as tall buildings and mountains), i.e., static environmental data; D pos ={p,v,θ v} represents a real-time dynamic data tuple for the vehicle, describing its position and motion trend at the current moment, including the vehicle's geographic location coordinates p, vehicle speed v, and vehicle direction angle θ. v The vehicle's geographic location coordinates p are coordinate pairs containing longitude and latitude, used to locate the vehicle to the static environmental data D. map Above; vehicle speed v represents the vehicle's current instantaneous speed, a key parameter used to predict the vehicle's future position, measured in meters per second (m / s); vehicle direction angle θ v The heading angle, expressed as the angle between the vehicle's direction of travel and true north, is used to determine the vehicle's trajectory and is measured in degrees (°). sensor Three-dimensional point cloud data representing dynamic obstacles is used to predict temporary and sudden signal occlusion caused by surrounding moving objects, i.e., dynamic obstacle data.

[0071] Step S1300: Through the vehicle application perception and recognition module, the service type of the upper-layer data stream is analyzed in real time, and a minimum guaranteed bandwidth is generated. Maximum transmission delay and business priorities The set of communication task requirements R task .

[0072] Specifically, this step aims to establish a bridge between upper-layer application requirements and lower-layer network resource allocation, dynamically mapping the abstract business requirements of upper-layer applications into a quantifiable and standardized set of communication task requirements R. taskTo ensure that we do not blindly pursue the best signal, but instead focus on the most critical and urgent business needs of the vehicle at present, we must avoid the risk of high-quality data being occupied by low-priority data, which could lead to delays in critical control commands.

[0073] In practice, the application perception and recognition module built into the vehicle's central control unit runs continuously, monitoring and parsing the data streams generated by upper-level applications (such as the remote driving control module, video streaming module, and vehicle status monitoring module) in real time. When a new data stream is detected or the attributes of an existing data stream change, the recognition module dynamically identifies the task type and tags it with a communication requirement based on a preset policy rule base. The specific recognition process includes the following three typical tasks:

[0074] I. High-bandwidth Task Identification. When the 4K high-definition video surveillance function is enabled, the identification module will determine that it is a high-bitrate data stream output by the video streaming module. According to the policy rule base, this data stream is assigned the labels of high bandwidth requirement, allowable certain latency, and medium priority.

[0075] II. Low-Latency Task Identification. When an emergency braking or obstacle avoidance steering command is detected, the identification module immediately identifies its high-risk attributes. The system assigns it the strictest latency requirements and the highest task priority, while its bandwidth requirements are relatively low.

[0076] III. Routine Task Identification. For vehicle status data (such as battery level, tire pressure, etc.) periodically reported by the vehicle status monitoring module, the system will allocate lenient latency and bandwidth requirements, as well as lower task priority.

[0077] In this way, all non-standardized business instructions are dynamically transformed into a standardized set of communication task requirements. Among them, R task This represents the set of all communication task requests that the vehicle needs to process at a given moment; j represents the data stream identifier, used to distinguish different data streams, for example, j=1 represents the control signal stream, j=2 represents the video stream; S stream This represents the set of identifiers for all currently active data streams; This represents a tuple representing the specific requirements of data stream j, including the minimum guaranteed width of data stream j. Maximum transmission delay of data stream j And the business priority of data stream j Minimum guaranteed width of data stream j This refers to the minimum data transmission rate necessary to ensure the normal operation of a service. The unit is usually Mbps, and its value varies greatly depending on the type of service. For example, the requirements for 4K high-definition video streaming... For vehicle status data or control commands, only a few hundred kbps are needed; the maximum transmission latency of data stream j is [missing information]. This represents the maximum allowed time for a data packet to travel from the sender to the receiver, typically measured in milliseconds (ms); the service priority of data stream j. The primary basis for resource scheduling is usually a non-negative integer; the smaller the value, the higher the priority. Indicates critical safety commands (such as emergency braking). This indicates real-time interactive services (such as high-definition video surveillance).

[0078] For example, consider a remotely controlled driverless truck operating in a mine. Assume that at a certain time t, the truck is being operated by a remote control operator to load ore and transport it to a designated location. At this moment, the vehicle's central control unit identifies four parallel data streams, each corresponding to different business needs, forming a set S of identifiers for all currently active data streams. stream ={1,2,3,4}. The system will precisely define the corresponding communication task requirements for each data stream based on its business characteristics, ultimately forming a complete set of communication task requirements R. task Furthermore, as shown in Table 1, Table 1 is an example table of the current communication task requirements obtained by the present invention. As can be seen from the example table, at time t, the communication service quality requirements of four parallel data streams on the remotely operated unmanned truck are as follows: For example, the first data stream (j=1) is a real-time control command from the remote operator, including critical safety operations such as steering, acceleration / deceleration, and braking. Because it is directly related to the vehicle's driving safety, the system will assign it the highest service priority. And requires the most stringent transmission latency This ensures instantaneous response to commands. Furthermore, given its small data size, only low bandwidth requirements are needed. This will satisfy the transmission requirements.

[0079] Table 1. Examples of obtaining current communication task requirements.

[0080]

[0081] S2000, based on environmental perception data D env Extract the set of dynamic obstacle feature vectors F obs Based on the environmental perception data D env and the set of dynamic obstacle feature vectors F obs The data is then integrated and fed into a parallel, dual-branch hybrid deep learning model M. DL Perform signal occlusion prediction and output the signal occlusion prediction results with multi-dimensional parameters. pred .

[0082] Specifically, this step aims to utilize the environmental perception data D collected in step S1200. env By using digital twin models, signal obstruction events on a vehicle's future path can be predicted in advance, providing forward-looking input for subsequent actions.

[0083] Further, step S2000 includes:

[0084] Step S2100, process the dynamic obstacle data D sensor Multi-stage processing is performed to generate a set of dynamic obstacle feature vectors F containing morphological and motion parameters. obs .

[0085] Specifically, this step aims to bridge the key gap between sensor data and advanced predictive models. Based on the environmental perception data D from step S1200... env Dynamic obstacle data D sensor Through a series of complex processing steps, it is transformed into a low-dimensional set of dynamic obstacle feature vectors F that is highly compatible with machine learning models. obs This greatly reduces the computational burden of subsequent steps and is a technical prerequisite for achieving real-time, high-precision signal occlusion prediction.

[0086] In the specific implementation process, based on dynamic obstacle data D sensor The process involves a series of steps, which can be divided into four core stages, as follows:

[0087] I. Spatiotemporal alignment. Due to the dynamic obstacle data D sensor Data is collected separately by a LiDAR and a binocular camera, which are physically separate heterogeneous sensors. Their coordinate systems and acquisition times differ, necessitating rigorous spatiotemporal alignment. First, a GPS timing module provides a unified time reference for both the LiDAR and the binocular camera, ensuring that the timestamp error of their respective data frames is within ±1ms, thus achieving time synchronization. Next, Zhang's calibration method is used to calculate the rotation matrix R and translation vector T between the two sensor coordinate systems to establish their spatial transformation relationship. Finally, affine transformation formulas are used to map the 3D points in the binocular camera coordinate system to the LiDAR coordinate system, completing spatial alignment. The specific process formulas are as follows:

[0088] P lidar =R·P vision +T;

[0089] Among them, P lidar R represents the final three-dimensional coordinates of a spatial point in the target coordinate system (i.e., the lidar coordinate system) after coordinate transformation; R represents the rotation matrix, used to describe the rotational attitude of the source coordinate system (i.e., the binocular camera coordinate system) relative to the target coordinate system in space; P visionIt represents the original three-dimensional coordinates of a spatial point in the source coordinate system; T represents the translation vector, which is used to describe the displacement of the origin of the source coordinate system relative to the origin of the target coordinate system in physical space.

[0090] II. Noise Reduction and Filtering. After completing spatiotemporal alignment, the acquired dynamic obstacle data D... sensor Outliers and redundant information may still exist due to factors such as sensor errors or environmental reflections. The system employs both statistical filtering and voxel filtering methods to process the dynamic obstacle data D. sensor Perform data cleaning and compression.

[0091] The statistical outlier filtering method identifies and removes outliers that deviate from the main distribution based on the statistical distribution of the distances between each point and its neighboring points. Specifically, the number of neighboring points is set to k=50, and the average Euclidean distance between each point and its 50 nearest neighbors is calculated. If the average distance between a point and its neighbors exceeds a preset threshold, the point is identified as a noise point and removed.

[0092] Voxel mesh filtering reduces data density while preserving the contour features of the point cloud. Specifically, the coordinate space of the entire point cloud is divided into a cubic voxel mesh with dimensions of 0.1m × 0.1m × 0.1m. Then, all voxels in the coordinate space are traversed. For each voxel containing one or more data points, the average of the 3D coordinates of all data points within it is calculated as a representative point, i.e., the voxel centroid, and this centroid replaces all data points within that voxel. If no data points fall within a voxel, it is considered empty after downsampling. This method significantly reduces redundant points in locally high-density regions, typically achieving a data compression of 60% to 80%, saving substantial computational resources for subsequent steps.

[0093] III. Morphological Feature Extraction. After noise reduction and filtering, based on the dynamic obstacle data D... sensor Using a convex hull algorithm (such as Graham's scan method), the minimum circumscribed cuboid enclosing a single obstacle point cloud cluster is calculated, thus accurately determining its length l, width w, and height h. Based on this, to further obtain the spatial occupancy characteristics of obstacles, the data points are divided into finer sub-cubes, and the data point density ρ of each sub-cube is determined accordingly. i' The total volume V of the obstacle is estimated by performing a weighted summation. The specific process and formula are as follows:

[0094]

[0095] Where V represents the total volume of the obstacle, in cubic meters (m³). 3 i' represents the index of each sub-cube, used to traverse all sub-cubes that make up the obstacle; n represents the total number of sub-cubes of the obstacle; ρ i'V represents the point density of the i'-th sub-cube, i.e., the number of data points within each sub-cube; i' This represents the volume of each sub-cube, a fixed value by default, in cubic meters (m). 3 .

[0096] IV. Trajectory Feature Extraction. To obtain the motion state information of obstacles, a Kalman filter algorithm is used to track and predict the trajectory of the centroid of the same obstacle identified in consecutive frames. By calculating the position change of the centroid between consecutive time steps, the instantaneous velocity s and instantaneous motion direction angle θ of the obstacle can be accurately estimated. s The specific process formula is as follows:

[0097]

[0098] Where s represents the instantaneous velocity of the obstacle, in meters per second (m / s); Δx and Δy represent the displacement components of the obstacle's center of mass in the X and Y axes respectively during two consecutive time steps, in meters (m); Δt represents the time interval between two consecutive time frames, in seconds (s); θ s The instantaneous direction of the obstacle's motion is represented by the angle between the obstacle's velocity vector and the positive X-axis of the coordinate system, expressed in degrees (°). arctan(·) represents the arctangent function, used to calculate the ratio of the increments of the Y-axis and X-axis in a rectangular coordinate system. Calculate the instantaneous angle θ of the obstacle's motion direction. s .

[0099] The centroid is a single coordinate point obtained by averaging the three-dimensional coordinates of all data points within the obstacle, and is used to represent the tracking anchor point of the entire obstacle.

[0100] Finally, all the extracted morphological and trajectory features are combined to construct a structured set of dynamic obstacle feature vectors. Among them, F obs This represents the total set of feature vectors of all identified dynamic obstacles in the scene at the current moment; j' represents the obstacle identifier, used to distinguish different obstacles, for example, j=1 represents a truck in front, j=2 represents a car on the side; This represents a single characteristic vector corresponding to the j'-th obstacle, a tuple containing obstacle shape and trajectory information, including the obstacle's length l. j' The width of the obstacle j' The height h of the obstacle j' The total volume V of the obstacle j' The instantaneous velocity s of the obstacle j' and the instantaneous direction of motion θ of the obstacle s,j' .

[0101] Step S2200, transfer static environment data D map Vehicle dynamic data D pos and the set of dynamic obstacle feature vectors F obs The data is then integrated and fed into a parallel, dual-branch hybrid deep learning model M. DL Extract spatial feature vectors F respectively cnn With time eigenvector F rnn Based on the spatial feature vector F cnn With time eigenvector F rnn Fusion is performed to construct a spatiotemporal fusion feature vector F fused The spatiotemporal fusion feature vector F is combined with the prediction decoding head. fused Decode the signal and output the predicted affected frequency band bank. k Predicted signal attenuation value A k The predicted start time of the occlusion event And the predicted duration of the occlusion event. Signal occlusion prediction results O pred .

[0102] Specifically, this step aims to generate a forward-looking and practical signal occlusion prediction result based on the multi-source environmental perception data and dynamic obstacle feature information extracted in all the preceding steps. pred This is to support subsequent proactive network scheduling and communication link optimization strategies.

[0103] In the specific implementation process, the aforementioned data information is first collected, including the environmental perception data D collected in step S1200. env Static environment data D map And vehicle's own dynamic data D pos and the set of dynamic obstacle feature vectors F constructed in step S2100 obs .

[0104] Next, the collected data is used as input and fed into a hybrid deep learning model M trained on large-scale historical data. DL The hybrid deep learning model M DL It adopts a parallel dual-branch architecture, with one branch being a convolutional neural network (CNN) for processing spatial information and the other branch being a recurrent neural network (RNN) for processing time-series information.

[0105] The spatial feature extraction branch is primarily responsible for extracting spatial morphological features that affect signal propagation from static and dynamic physical environments. First, the static environment data D... mapand the set of feature vectors of dynamic obstacles F obs ={l j' ,w j' ,h j' V j' The sequence |j'=1,2,...,m} is fused and encoded to construct a structured spatial tensor, denoted as X. spatial Subsequently, the spatial shape tensor X spatial The input is fed into a branch of a convolutional neural network (CNN), where multiple layers of convolution and pooling operations are used to progressively abstract spatial information and extract features. The final output is a compact spatial feature vector containing all key spatial information, denoted as F. cnn The CNN branch is composed of multiple convolutional layers and pooling layers stacked together. The filter parameters in each convolutional layer are optimized through supervised training, enabling it to automatically identify spatial patterns with physical meaning, such as "the edge of a tall wall", "a street canyon formed by continuous buildings", and "the metal block of a large vehicle".

[0106] The time-series feature extraction branch is primarily responsible for learning and understanding the dynamic evolution of the entire scene from the historical motion trajectories of vehicles and obstacles. First, it extracts the vehicle's own dynamic data D... pos and the set of feature vectors of dynamic obstacles F obs ={s j' ,θ s,j' The sequence |j'=1,2,...,m} is integrated temporally to construct a structured temporal state sequence tensor, denoted as X. temporal =[State t-N ,...,State t-1 State t ]; where t represents the index of the current time; N represents the length of the time window constituting the time series, i.e., the number of consecutive data frames; Srate t-N State represents the first state vector in the time series, i.e., the scene state vector N-1 time steps prior; t-1 State represents the scene state vector at the previous time step t-1; t This represents the scene state vector at the current time t. Subsequently, the time-series state sequence X... temporalThe input is fed into a Recurrent Neural Network (RNN) branch for processing, which typically employs a Long Short-Term Memory (LSTM) network capable of constructing long-term dependencies. The LSTM network, through its internal gating mechanism, selectively memorizes key historical dynamic information (such as obstacles being in an accelerated state) while filtering out irrelevant or redundant information. By progressively parsing the entire sequence, the LSTM network can capture complex dynamic interaction patterns, achieving prior perception of potential future events. Finally, it outputs a temporal feature vector containing the dynamic evolution trend of the entire scene, denoted as F. rnn .

[0107] Subsequently, the spatial feature vectors F output by the two parallel branches are... cnn and time eigenvector F rnn By concatenating the features along the feature dimension, a spatiotemporal fusion feature vector F with higher dimension and more comprehensive information is formed. fused .

[0108] Finally, the spatiotemporal fusion feature vector F is... fused The input is fed into the model's prediction decoder head. This prediction decoder head is typically a multilayer perceptron (MLP) consisting of several fully connected layers, used to fuse high-dimensional spatiotemporal feature vectors F. fused Decoding maps to understandable, structured signal occlusion prediction results. The Multilayer Perceptron (MLP) is trained using a specialized multi-task learning method and contains multiple parallel output branches, each corresponding to a different type of prediction objective, used to answer key questions about the future state of the signal. The functions of each branch are described below:

[0109] I. Communication Frequency Band Obstruction Risk Classification, i.e., a classification problem. This branch determines which communication frequency band is most likely to be affected by obstruction. Its output is activated by a Softmax function to calculate the corresponding probability value for each candidate communication frequency band (e.g., 'mmWave', 'Sub-6G', '4G-LTE'). The one with the highest probability is selected as the prediction result, i.e., the predicted affected frequency band bank. k .

[0110] II. Signal Attenuation Prediction, i.e., a regression problem. This branch predicts the signal attenuation caused by an occlusion event, outputting a continuous real value in decibels (dB), corresponding to the predicted signal attenuation value A. k .

[0111] III. Prediction of the Occlusion Event Timing, i.e., a regression problem. This branch is used to predict the proximity risk of signal terminals. Based on the relative motion between the vehicle and the obstacle, a countdown timer, measured in seconds (s), is calculated as the predicted start time of the occlusion event.

[0112] IV. Occlusion Duration Prediction, i.e., a regression problem. This branch combines the obstacle size and the vehicle's relative speed to predict the duration of the occlusion event.

[0113] For example, consider a scenario where an autonomous vehicle is approaching the entrance to a long tunnel beneath an urban overpass. In this scenario, the hybrid deep learning model M... DL Parallel parsing of multi-source input data is performed. The spatial feature extraction branch is responsible for interpreting spatial dimensional information by analyzing static environmental data D. map and the set of feature vectors of dynamic obstacles F obs The system quickly identifies a large concrete structure with strong electromagnetic shielding characteristics located ahead, namely the tunnel itself, which poses a significant risk of obstructing high-frequency communication signals. Simultaneously, the time-series feature extraction branch processes the time-series data based on the vehicle's own dynamic data D. pos and the set of feature vectors of dynamic obstacles F obs It captures in real-time the movement trend of the vehicle gradually approaching the tunnel entrance at a stable speed. The abstract feature vectors output by the two branches semantically represent the spatial feature vector F indicating "tunnel ahead". cnn The time feature vector F of "the vehicle is steadily approaching" rnn Subsequently, the two feature vectors are concatenated along their feature dimensions to construct a spatiotemporal fusion feature vector F. fused This constructs a complete semantic description of the event: "The vehicle is moving at a steady speed towards a large tunnel." Next, this spatiotemporal fusion feature vector F... fused The input to the model's prediction decoder head outputs a structured occlusion risk assessment result. The model determines that, in this scenario, the millimeter-wave communication frequency band is banked. k The most severe impact will be felt, with the predicted signal attenuation value A. k The obstruction will reach as high as 50 dB. Based on the vehicle's current distance from the tunnel entrance being approximately 100 meters and its speed being 20 m / s, the obstruction event is estimated to occur in approximately 5.0 seconds. This occurred subsequently. Based on the tunnel's total length of 800 meters shown on the map, it was further predicted that the obstruction would last approximately 40.0 seconds. Finally, the model outputs a signal occlusion prediction result O. predIt contains at least one key prediction tuple ('mmWave', 50, 5.0, 40.0). The model may simultaneously output secondary prediction results for other communication frequency bands in Sub-6G, such as ('Sub-6G', 25, 5.0, 40.0). Signal obstruction prediction result O pred Precise and forward-looking information, including specific frequency bands, attenuation levels, start times, and durations, is transmitted to subsequent steps to ensure that the system can take optimal measures to intervene and adjust before the signal actually deteriorates, minimizing the impact of obstruction on the continuity and reliability of communication services.

[0114] S3000 sequentially performs signal link feasibility checks and network resource availability checks on the acquired candidate links, and selects qualified candidate links to construct a candidate set C of multi-dimensional performance profiles. set Based on the multidimensional performance profiles of each candidate link, and combined with a dual-mode decision triggering mechanism, a deep Q-network is used to generate dynamic weight factors when the communication decision process is activated. The comprehensive score S of each candidate link is then calculated based on these dynamic weight factors. score Based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate links, sorted in descending order, are checked against hard constraints one by one until a candidate link that meets the hard constraints is found to be the primary communication link. Based on the overall network resource situation, a choice is made between switching policy instructions and coordination policy instructions to generate the final communication policy instruction, Start. comm .

[0115] Specifically, this step aims to integrate and analyze the real-time network signal quality data P collected in step S1100. Sig Based on the environmental perception data D collected in step S1200 env The communication service requirement R obtained in step S1300 task And the signal occlusion prediction result O generated in step S2200 pred The above-mentioned multi-source heterogeneous data information is modeled and comprehensively analyzed. Based on a multi-objective optimization algorithm, a unique and optimal communication strategy instruction, Start, is finally generated. comm It is issued to the execution layer in the form of instructions.

[0116] Further, step S3000 includes:

[0117] Step S3100: Obtain the candidate link list and network load rate based on real-time network signal quality data P. SigPerform a signal link feasibility check on the candidate link list, and perform a network resource availability check on the candidate link list based on the network load rate. After completing the above feasibility and availability checks, construct a candidate set C from all remaining qualified candidate links. set And based on the candidate set C set The generation of each candidate link includes signal quality index N sig Network load metric N load and distance index N dis A multidimensional performance profile.

[0118] In the specific implementation process, the vehicle-mounted system first sends a request to the current serving base station and obtains a NeighborRelationList. This NeighborRelationList not only contains the identification and frequency information of candidate links in neighboring cells or available backup communication frequency bands, but also includes static data such as corresponding geographical location information, providing necessary support for subsequent evaluation. After obtaining the candidate link list, a rigorous two-stage pre-qualification and pruning process is performed on each candidate link, as follows:

[0119] I. Signal Link Feasibility Verification. Based on real-time network signal quality data P Sig The signal strength of each candidate link is evaluated. Any candidate link with a received signal strength RSSI below -110dBm will be directly removed from the candidate list due to weak signal coverage and lack of basic conditions for establishing a stable physical connection.

[0120] II. Network Resource Availability Verification. If a link with a strong signal is already saturated with resources, it cannot provide quality service. Therefore, further analysis of the network load rate obtained from System Information Blocks (SIBs) broadcast by the base station or proprietary signaling is used to determine the current load status of candidate links. Any overloaded link with a network load rate exceeding 90% is pruned because it cannot guarantee service quality.

[0121] After the above feasibility and availability screening, all remaining qualified candidate links together constitute the final candidate set C. set To enable horizontal comparison of these qualified candidate links within a unified mathematical framework, a multi-dimensional performance profile is constructed for each link, and the metrics of each dimension are normalized to eliminate the influence of dimensions. The multi-dimensional performance profile covers multiple key performance indicators, each derived from different data dimensions: signal quality indicator N...sig It is based on real-time network signal quality data P Sig The signal-to-noise ratio (SNR) is used in the calculation to reflect the physical transmission capacity of the link; network load metric N... load Resource utilization of candidate links is measured by parsing System Information Blocks (SIBs) broadcast by base stations or extracting proprietary signaling; distance metric N dis Based on environmental perception data D env Vehicle dynamic data D pos Real-time Euclidean distance is calculated between the real-time GPS coordinates and the geographical coordinates of candidate base stations to estimate potential propagation delay and interference risks of the link. To achieve a unified comparison among different performance indicators, a max-min normalization paradigm is adopted, mapping all indicators to a closed interval [0,1]. Taking signal quality SNR as an example, its score N... sig The specific process formula is as follows:

[0122]

[0123] Where, N sig The signal quality index (SNR) is a dimensionless floating-point number between 0 and 1, representing the normalized signal quality. A higher SNR value indicates that the signal quality is closer to the ideal state. current The real-time signal-to-noise ratio (SNR) of the candidate link is a core metric for measuring signal clarity, measured in decibels (dB). min and SNR max These represent the system's preset acceptable lower limit for signal-to-noise ratio and the ideal upper limit for signal-to-noise ratio, respectively. These are benchmark parameters set based on engineering experience.

[0124] Step S3200, based on real-time network signal quality data P Sig And signal occlusion prediction results O pred The system combines a dual-mode decision-triggered mechanism to determine whether to activate the communication decision-making process. If the communication decision-making process is activated, a deep Q-network is used to generate dynamic weight factors based on the state vector of the current communication environment. The comprehensive score S of each candidate link is then calculated based on these dynamic weight factors and the multi-dimensional performance profiles of each candidate link. score .

[0125] In the specific implementation, to reduce the computational burden and improve response efficiency, a dual-mode decision trigger mechanism is designed to ensure that the complete communication decision process is activated only when truly necessary. This mechanism is activated when either of the following conditions is met: First, a passive response trigger mechanism. When the real-time network signal quality data P... SigThe display shows that the key performance indicators of the currently used link (such as Received Signal Strength Indicator (RSSI)) are below the preset minimum availability threshold of -110dBm, indicating that the service quality has deteriorated unacceptably; secondly, an active prevention triggering mechanism. When the signal obstruction prediction result P pred The display indicates that when the current communication link is about to face a high probability and high attenuation signal blockage event in the near future, it means that the system anticipates an impending deterioration in service quality.

[0126] Once the communication decision-making process is activated, the system enters a dynamic weighting and comprehensive scoring phase to evaluate the applicability of all candidate communication links. Based on a reinforcement learning model using a Deep Q-Network (DQN), and combined with the state vector of the current communication environment, a set of optimal dynamic weight factors (W) is dynamically generated. sig W load W dis The state vector includes the communication service requirement R. task Based on the candidate set C generated in step S3100 set and environmental perception data D env Static environment data D map Then, the candidate set C is traversed sequentially. set Each qualified candidate link is combined with a dynamic weighting factor (W). sig W load W dis The signal quality index N for each candidate link is calculated in step S3100. sig Network load metric N load and distance index N dis A comprehensive and quantitative integrated score S is calculated for each candidate link using a multi-objective optimization formula. score The specific process formula is as follows:

[0127] S score =W sig ·N sig +W load ·(1-N sig )+W dis ·(1-N dis );

[0128] Among them, S score W represents the final composite score of the candidate links. sig W load W dis These represent dynamic weighting factors for the three dimensions of signal quality, network load, and distance, respectively, with the sum of the three being 1. These factors are dynamically generated by the DQN model based on the real-time scenario; (1-N) sig ) and (1-N disThis is used to convert negative metrics, such as network load and distance metrics, where smaller values ​​are considered better, into positive metrics.

[0129] Step S3300, based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate set C sorted in descending order set The candidate link ranked first is subjected to hard constraint testing. If the candidate link meets the hard constraints, it is determined as the main communication link; if the candidate link does not meet the hard constraints, the process is automatically iterated to the candidate set C after descending order. set The next candidate link is then checked against the hard constraints until a candidate link that meets the hard constraints is found. Based on the overall network resource situation, a choice is made between the handover policy instruction and the coordination policy instruction to generate the final communication policy instruction, Start. comm .

[0130] Specifically, this step aims to act as a bridge between abstract analysis results and concrete actions, by combining the comprehensive score S calculated using S3200. score The candidate set C generated in step S3100 set Communication service requirements R obtained from S1300 task Generate a well-structured communication strategy instruction Start that can be directly parsed and executed. comm .

[0131] In the specific implementation process, firstly, the current communication service requirement R is analyzed. task Extract the data stream task with the highest business priority (i.e., priority) from it. Performance requirements of the key data stream j (the item with the smallest numerical value) And identify its core performance requirements, including minimum protection width. and maximum transmission delay As a hard constraint for link allocation, it must be met unconditionally. Subsequently, based on the comprehensive score S... score For candidate set C set The candidate links are sorted in descending order, and the top-ranked candidate link is selected as the initial optimal target. A constraint satisfaction check is then performed on this candidate link to determine whether it simultaneously satisfies the aforementioned minimum guaranteed width. and maximum transmission delay The system considers the hard constraints. If a candidate link meets all hard constraints, it is confirmed as the primary link capable of carrying critical tasks, and the system proceeds to the next step of strategy formulation. If a candidate link does not meet any hard constraints, the system discards the candidate link and automatically iterates to the descendingly sorted candidate set C.set The next candidate link (i.e., the composite score S) score The second candidate link continues to undergo the same constraint satisfaction check. This iterative process is based on the comprehensive score S. score The sorting process continues until the first candidate link that meets all key business requirements is located.

[0132] Finally, after selecting a main communication link that meets the constraints, the system will, based on the overall situation of current network resources, initiate a communication strategy with two completely different structures and logics, using the Start command. comm Choose one of the following to generate:

[0133] I. Generation of Handover Policy Instructions. When the system determines that the current network environment is relatively strained (i.e., the quality of candidate links is generally poor except for the selected primary communication link), or when the main purpose of this decision is to avoid the deterministic failure risk of the current link, the system will initiate a concise handover policy instruction generation process. This process aims to construct a policy instruction tuple of the form ('HANDOVER', {params}); where 'HANDOVER' represents the policy instruction type, used by the instruction execution layer for network handover; and {params} represents the parameter dictionary, including target. band The parameters are target and reason. Among them, target... band The parameter is used to specify the frequency band of the finally selected primary communication link, and its value is determined by the final result of the candidate link constraint satisfaction test. For example, if the system ultimately selects a Sub-6GHz frequency band link as the primary communication link for the current task, then target... band The value is 'Sub-6G'. The reason parameter is used to support subsequent backtracking and model optimization, recording the triggering reason for this link switch. For example, if the decision is activated by the passive response triggering mechanism in step S3200, such as the received signal strength indicator RSSI being lower than the minimum available threshold, then reason is set to 'ReactiveThreshold Cross'; if the decision is activated by the active response triggering mechanism in step S3200, such as the signal obstruction prediction result O pred If occlusion is predicted, the reason is set to 'Proactive PredictedBlockage'.

[0134] II. Generation process of Synergy strategy instructions. When the system determines that the current network environment has sufficient resources (i.e., in addition to the already determined main communication link, the candidate set C...), setSeveral other high-quality links exist, prompting the system to initiate a more advanced collaborative policy instruction generation process. This process aims to construct a policy instruction tuple of the form ('SYNERGY', {params'}); where 'SYNERGY' represents the policy instruction type, used by the instruction execution layer for multi-path concurrent transmission; and {params'} represents the parameter dictionary of the dynamic routing table, used to describe the mapping relationship between the current multi-service flows and multiple communication links, including control... task Parameters, video task Parameters and telemetry task Parameters. Among them, control... task The parameter represents the control task, i.e., the real-time vehicle control data stream, which is the most critical and highest priority communication task; video task The parameter represents the video task, i.e., the video data stream captured and transmitted by the vehicle-mounted camera; telemetry task The parameters represent the remote sensing task, specifically the telemetry data stream of the vehicle's own operating status. The specific process is as follows:

[0135] I. Assigning Key Business Tasks. First, assign the highest priority control tasks to key business tasks. task Assigned to the determined primary communication link and written as the first key-value pair in the parameter dictionary. For example, if the system ultimately selects an mmWave band link as the primary communication link for the current task, the parameter dictionary is initialized to {'control'. task ':'mmWave'}.

[0136] II. Constructing the Resource Pool to be Allocated. Subsequently, two temporary resource pools are created: a task pool and an available link pool. The task pool contains communication service requirements R. task Other service flows besides the allocated critical services (such as video, telemetry, etc.) are prioritized according to service priority. Sort from highest to lowest; the available link pool contains candidate set C. set Other high-quality candidate links besides the main communication links that are already occupied are ranked according to their comprehensive score S. score Sort from highest to lowest.

[0137] III. Iterative Matching and Allocation. Next, the highest priority tasks in the task pool are processed, and the available link pool is traversed sequentially to find the best matching link. This matching process involves complex intelligent decision-making logic, specifically including the following three typical scenarios:

[0138] I. Single-link fulfillment scenario. For the currently pending business (e.g., video task...) task First, evaluate the overall score S in the available link pool.score The highest-performing link is used to determine whether its performance meets the core requirements of the service (e.g., minimum guaranteed bandwidth). If the conditions are met, the link will be assigned to the service, and the corresponding key-value pair will be added to the parameter dictionary, such as {'video'. task At the same time, remove the service and the link from their respective resource pools to ensure that resources are not redistributed.

[0139] II. Link Aggregation Scenario. If no single link in the available link pool can meet the current high-bandwidth service (such as 4K video) requirements, the system will further determine whether the bandwidth of multiple links can be aggregated to meet the service requirements. For example, the overall score S in the available link pool is calculated. score If the sum of the bandwidths of the two highest-bandwidth links exceeds the minimum guaranteed bandwidth... The system will decide to perform bandwidth aggregation on these two links to jointly carry the service. At this point, the link identifiers will be presented in a list format in the key-value pairs added to the parameter dictionary, such as {'video'. task ':['Sub-6G','4G-LTE']}.

[0140] III. Link Reuse Scenarios. When completing high-priority tasks (e.g., video tasks...) task After the link allocation is completed, the system continues to process lower priority tasks (such as telemetry tasks). task This not only considers unused link resources in the available link pool, but also intelligently assesses whether there are allocated links with remaining capacity available for reuse. For example, in the telemetry task... task For example, because its bandwidth requirement is extremely low, the system will prioritize routing it to a channel already used for other tasks (such as video tasks). task Instead of allocating a separate new link resource, the active link (e.g., 4G-LTE) is used. This strategy can significantly improve the overall link resource utilization and reduce system power consumption while ensuring communication quality. In this case, the parameter dictionary may contain {'telemetry_task':'4G-LTE'}, which corresponds to the video task. task They share the same link.

[0141] The iterative matching and allocation process described above continues until all services in the pending task pool have been successfully allocated, or all resources in the available link pool have been exhausted. Finally, the system outputs a parameter dictionary that constitutes an optimal communication strategy instruction, Start, tailored to the current network environment and service requirements. comm .

[0142] S4000, according to the communication strategy instruction Start comm Perform instruction type determination; if the communication strategy instruction Start... comm For a collaborative strategy, dynamic fragmentation is performed on the data stream to generate data packets P that can be transmitted concurrently. sliced At the receiving end, the data packet P... sliced Cyclic redundancy check, out-of-order reordering, and intelligent delivery are performed sequentially to restore the complete and ordered data stream D. reass ; in the communication strategy instruction Start comm After execution, a feedback data packet F containing the key performance indicators of the instructions is collected and encapsulated. back and the feedback data packet F back Iterative training of deep Q-networks is used to update decision weights and achieve adaptive closed-loop feedback.

[0143] Specifically, this step aims to implement the communication policy instruction Start generated in step S3000. comm The process is executed and feedback is given to the decision-making level, forming a closed loop.

[0144] Further, step S4000 includes:

[0145] Step S4100, according to the communication strategy instruction Start comm If the instruction type is a cooperative strategy, then the optimal fragment size S is dynamically calculated based on the real-time available bandwidth B of each link and the target transmission time window T. size And based on the optimal fragment size S size Dynamically fragment the data stream to generate data packets P that can be transmitted in parallel across multiple links. sliced If the instruction type is a switching strategy, then the dynamic sharding process is skipped.

[0146] Specifically, this step aims to start the communication policy instruction. comm This translates into executable behavior at the physical network layer, acting as a bridge and scheduler between the decision-making and execution layers. The core of this step lies in using dynamic data stream fragmentation and multi-link multiplexing to precisely translate upper-layer strategies into specific data transmission operations, ensuring the system's high efficiency and real-time performance.

[0147] In the specific implementation process, the system follows the communication strategy instruction Start comm Different instruction types execute different processing logic:

[0148] If the instruction type is 'HANDOVER', the system needs to seamlessly migrate a complete, unchanged data stream from the current link to another target link. This process mainly involves low-level signaling interaction and network path redirection, and does not require segmentation or reconstruction of the data stream itself. Therefore, the data fragmentation logic in this step is skipped directly.

[0149] However, when the instruction type is 'SYNERGY' (cooperative strategy), the core logic of this step is activated. At this time, the system needs to intelligently divide one or more large data streams (such as high-definition video streams) and schedule them in parallel across multiple communication links. Dynamic Slicing becomes a necessary and critical operation. This process aims to achieve the optimal trade-off between transmission performance and network robustness: if the fragments are too large, the loss of a single packet will result in high data retransmission overhead and be detrimental to latency control; if the fragments are too small, excessive header overhead will be introduced, reducing overall bandwidth utilization. Therefore, the system needs to dynamically calculate the optimal fragment size S based on the real-time carrying capacity of each link and the latency requirements of the application. size The specific process formula is as follows:

[0150]

[0151] Among them, S size 1 represents the optimal fragment size, i.e., the effective data volume that each data packet can carry, in bytes; B represents the current real-time available bandwidth of the target link, a value that is dynamically obtained from network measurements and is constantly changing, used to reflect the channel's instantaneous carrying capacity, in bits per second (bps); T represents the target transmission time window, used to limit the expected transmission duration of each fragment, balancing transmission delay and protocol overhead, the value of which depends on the scenario and application requirements, in seconds (s); 8 is the unit conversion factor, used to convert data streams in bits to bytes, i.e., 1 Byte = 8 bits.

[0152] Calculate the S for each link participating in the collaboration. size Then, the data fragmentation engine sequentially cuts data blocks of appropriate sizes from the original data stream. Each fragmented data packet P sliced Each packet will be appended with critical metadata, including at least a unique 64-bit sequence number to ensure the receiving end can correctly reassemble the data in the correct order, and a synchronized millisecond-level timestamp to guarantee timing consistency. Ultimately, these processed and tagged packets will be ordered according to Start... comm The defined dynamic routing table is precisely distributed to the respective designated communication links.

[0153] For example, taking the transmission of a single 4K high-definition video stream as an example, the communication strategy command Startcomm For {'SYNERGY', {'video' task If ':['Sub-6G','4G-LTE']}} is used, it indicates that the system needs to distribute the high-definition video stream simultaneously to both Sub-6G and 4G-LTE links for parallel transmission. Assume the system measures the available bandwidth B of the Sub-6G link in real time. sub6g The available bandwidth B of the 4G-LTE link is 80Mbps. 4glte The speed is 20 Mbps, and the currently set target transmission time window T is 0.2 seconds. Based on the above parameters, the system calculates the optimal fragment size for each link: Sub-6G link: 4G-LTE link: Subsequently, the segmentation engine divides the original 4K high-definition video data stream proportionally according to the above results, generating two data segments within each 0.2-second time window: one 2.0MB in size, scheduled to the Sub-6G link; and the other 0.5MB in size, allocated to the 4G-LTE link. Ultimately, the system achieves a total data throughput of 2.5MB per transmission cycle, corresponding to an equivalent total bandwidth of approximately 100Mbps (2.5MB × 8 ÷ 0.2s), maximizing the utilization of multi-link aggregation and effectively ensuring the continuity and stability of 4K high-definition video in complex network environments.

[0154] Step S4200, for the received data packet P sliced Cyclic redundancy check (CRC) is performed. Data packets that pass the check are stored in a circular buffer based on their sequence numbers for out-of-order reassembly. A bidirectional scanning reassembly algorithm is then used to intelligently decode and deliver the data packets in the circular buffer sequentially, thereby retrieving the received data packet P. sliced Restored to a complete and ordered data stream D reass .

[0155] Specifically, this step aims to process the data packet P from step S4100. sliced Each data stream is verified and reassembled individually to ensure that it can be accurately restored to the original, ordered, and error-free complete data stream at the millisecond level after undergoing parallel and multi-link transmission, thereby providing reliable and highly consistent data support for upper-layer applications (such as video decoding and control command parsing).

[0156] In the specific implementation process, all data packets P arriving at the receiving end sliced It will then proceed through a three-stage automated processing flow, as follows:

[0157] I. Data Integrity Check (via CRC). Each data packet P slicedUpon arrival at the receiving end, a Cyclic Redundancy Check (CRC) is required to verify whether errors occurred during transmission. To balance computational burden and reliability requirements, the system employs a hierarchical verification mechanism: For critical data packets, such as those carrying video I-frame or P-frame header information, the system prioritizes the use of CRC-32, which has stronger error detection capabilities, because the correctness of its content directly affects the decoding accuracy of subsequent image sequences. Although this adds 4 bytes of verification overhead compared to lightweight verification methods, it greatly ensures the integrity of critical structural data during transmission. For non-critical data packets, such as those carrying video B-frame content, their importance is lower than that of I-frames or P-frames. Even if some are lost, the decoder can still perform fault-tolerant reconstruction by referring to other frames. In such scenarios, the system prioritizes the use of the less computationally expensive CRC-16 verification algorithm to ensure basic integrity and reduce processing burden.

[0158] Any data packet that fails the Cyclic Redundancy Check (CRC) is considered corrupted during transmission. The system will immediately discard the data packet and initiate a retransmission request to the upstream sender to ensure the reliability and integrity of the overall data transmission.

[0159] II. Reassembly via Sequence Number. Under a cooperative strategy, data packets from multiple links often arrive out of order. To address this issue, the receiver maintains a sufficiently large circular buffer for each session. When a data packet that has passed the Cyclic Redundancy Check (CRC) enters this stage, the system reads the unique 64-bit sequence number in its header and accurately stores the data packet's payload in the corresponding slot in the buffer based on this sequence number. As data packets continue to arrive, the slots in the buffer are gradually filled, allowing the out-of-order data to be spatially reordered and restored to the correct transmission order.

[0160] III. Intelligent Decoding and Delivery. To avoid excessive initial latency caused by the waiting buffer becoming completely full, the system employs a highly efficient bidirectional scanning reassembly algorithm. This algorithm continuously scans the circular buffer, starting from the last successfully delivered position, prioritizing the identification and submission of one or more complete and independently decodeable video frames, especially the critical I-frames that serve as the starting point for decoding. Once it is confirmed that all data packets corresponding to these complete frames are in place, the bidirectional scanning reassembly algorithm immediately extracts the complete data blocks from the buffer and delivers them to upper-layer applications (such as video decoders), without waiting for subsequent data packets with more distant sequence numbers to arrive. This proactive and continuous delivery mechanism effectively reduces reassembly latency caused by network jitter and out-of-order delivery.

[0161] Finally, after a complete verification, sorting, and reassembly process, the originally scattered data packets were accurately restored into a complete and ordered data stream D. reass It is delivered to the upper-layer application.

[0162] Step S4300: Execute the communication strategy instruction Start. comm Then, the communication strategy instruction Start is executed. comm The execution effect is evaluated, and the key performance indicators collected in the evaluation are encapsulated into a feedback data packet F. back and the feedback data packet F back The samples are used as training data for iterative training of the deep Q-network to update the decision weights and achieve adaptive closed-loop feedback.

[0163] In practice, the execution process of this step mainly includes two stages: strategy execution and effect evaluation, as follows:

[0164] I. Policy Execution. In this stage, the system processes the communication policy instruction Start from step S3000, which has the instruction type 'HANDOVER'. comm Upon receiving a handover command, the system forwards it to the underlying communication module, triggering a standard Radio Resource Control (RRC) connection reconfiguration process. This process is a rigorous signaling interaction mechanism designed to achieve coordinated handover between the current base station and the target base station. By dynamically negotiating the data transmission link between the source and target paths, it minimizes interruption time during the handover process, ensuring the continuity of data transmission and the stability of communication services.

[0165] II. Effect Evaluation and Feedback Encapsulation. In each communication strategy instruction Start... commAfter the driven action is completed, whether it's the 'HANDOVER' switching strategy or the 'SYNERGY' coordination strategy, the system's monitoring module will immediately evaluate its execution effect. This evaluation process collects a series of key performance indicators and packages this collected information into a structured feedback data packet F. back The feedback data packet F back It includes four key fields: 'timestamp', which represents the timestamp of this policy execution and evaluation, used to record the specific moment the event occurred, providing a time-series reference for subsequent training; 'action', which represents the action taken. taken 'Indicates the complete communication policy instruction to be executed: Start comm Instruction tuples, such as {'SYNERGY', {'video', ... task ':['Sub-6G','4G-LTE']}}; The action result 'outcome' is used to mark the result status of the policy execution, such as 'SUCCESS' indicating success, 'FAILURE' indicating failure, 'TIMEOUT' indicating timeout, etc.; The quantitative performance metrics 'metrics' represent a dictionary of quantitative performance metrics of the network status after the policy execution, such as {'latency':85,'post action-rssi ':-89,'post action-snr ':25}. Where 'latency' represents the end-to-end communication latency after the policy is executed, in milliseconds; 'post action-rssi 'Indicates the received signal strength of the new link after the policy is executed, in dBm;'post action-snr 'Indicates the signal-to-noise ratio of the new link after the strategy is executed, in dB.

[0166] Finally, the feedback data packet F back The data is transmitted to the Deep Q-Network (DQN) learning system deployed in step S3200 and incorporated into the model update process as a new training sample. This is achieved by continuously introducing feedback data packets F from the real environment. back Through iterative training, the decision weights of the DQN model are continuously and dynamically updated. This closed-loop feedback mechanism enables the system to self-evolve and continuously optimize in complex and ever-changing communication environments, enhancing its overall intelligent decision-making capabilities and environmental adaptability.

[0167] Example 2

[0168] This embodiment, based on Embodiment 1, provides a 5G network optimization system based on an intelligent dynamic multi-band adaptive switching system, such as... Figure 2 As shown, it includes:

[0169] A multi-source heterogeneous data acquisition and sensing module is used to acquire real-time network signal quality data P. sig Environmental perception data D env and the communication task requirement set R task ;

[0170] The intelligent signal obstruction prediction module is based on environmental perception data D. env Extract the set of dynamic obstacle feature vectors F obs Based on the environmental perception data D env and the set of dynamic obstacle feature vectors F obs The data is integrated and fed into a parallel dual-branch hybrid deep learning model for signal occlusion prediction, outputting a multi-dimensional parameter signal occlusion prediction result. pred ;

[0171] The communication strategy instruction generation module is used to sequentially perform signal link feasibility checks and network resource availability checks on the acquired candidate links, and select qualified candidate links to construct a candidate set C of multi-dimensional performance profiles. set Based on the multidimensional performance profiles of each candidate link, and combined with a dual-mode decision triggering mechanism, a deep Q-network is used to generate dynamic weight factors when the communication decision process is activated. The comprehensive score S of each candidate link is then calculated based on these dynamic weight factors. score Based on communication service requirements R task The hard constraints, based on the comprehensive score S score Candidate links, sorted in descending order, are checked against hard constraints one by one until a candidate link that meets the hard constraints is found to be the primary communication link. Based on the overall network resource situation, a choice is made between switching policy instructions and cooperative communication policy instructions to generate the final communication policy instruction, Start. comm ;

[0172] The closed-loop feedback optimization module is used to optimize the communication strategy based on the Start instruction. comm Perform instruction type determination; if the communication strategy instruction Start... comm When using a collaborative strategy, dynamic fragmentation is performed on the data stream to generate data packets P that can be transmitted concurrently. sliced At the receiving end, the data packet P... sliced Cyclic redundancy check, out-of-order reordering, and intelligent delivery are performed sequentially to restore the complete and ordered data stream D. reass ; in the communication strategy instruction Start comm After execution, a feedback data packet F containing the key performance indicators of the instructions is collected and encapsulated. back and the feedback data packet F back Iterative training of deep Q-networks is used to update decision weights and achieve adaptive closed-loop feedback.

[0173] The parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0174] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A 5G network optimization method based on an intelligent dynamic multi-band adaptive handover system, characterized in that, include: Acquire real-time network signal quality data, environmental perception data, and a set of communication task requirements; Based on environmental perception data, a set of dynamic obstacle feature vectors is extracted. The environmental perception data and the set of dynamic obstacle feature vectors are integrated and input into a parallel dual-branch hybrid deep learning model to predict signal occlusion. The signal occlusion prediction result with multi-dimensional parameters is output. The obtained candidate links are sequentially subjected to signal link feasibility test and network resource availability test, and qualified candidate links are selected to construct a candidate set of multi-dimensional performance profiles; Based on the multidimensional performance profiles of each candidate link; Based on real-time network signal quality data and signal blockage prediction results, a dual-mode decision triggering mechanism is used to determine whether to activate the communication decision process. If the communication decision process is activated, a deep Q-network is used to generate dynamic weight factors based on the state vector of the current communication environment, and the comprehensive score of each candidate link is calculated based on the dynamic weight factors and the multi-dimensional performance profile of each candidate link. The dual-mode decision triggering mechanism is activated when any of the following conditions are met: if real-time network signal quality data shows that the key performance indicators of the currently used link are lower than the preset minimum availability threshold, indicating that the service quality has deteriorated unacceptably, then the passive response triggering mechanism is activated; if the signal blockage prediction results show that the current communication link will face a high probability, high attenuation signal blockage event in the near future, representing that the system foresees the impending service quality deterioration, then the active prevention triggering mechanism is activated. Based on the hard constraints of communication service requirements, the candidate links sorted in descending order of the comprehensive score are checked one by one for hard constraints until a candidate link that meets the hard constraints is found to be the main communication link. Based on the overall situation of the current network resources, a choice is made between switching policy instructions and coordination policy instructions to generate the final communication policy instructions. The handover strategy instruction includes the strategy instruction type 'HANDOVER', the frequency band parameters of the finally selected primary communication link, and the triggering reason for this link handover; the coordination strategy instruction includes the strategy instruction type 'SYNERGY', control task, video task, and remote sensing task; Based on the communication policy instruction, the instruction type is determined. If the communication policy instruction is a cooperative policy, dynamic fragmentation is performed on the data stream to generate data packets that can be transmitted concurrently. At the receiving end, cyclic redundancy check, out-of-order reordering, and intelligent delivery are sequentially performed on the data packets to restore them into a complete and ordered data stream. After the communication policy instruction is executed, feedback data packets containing the key performance indicators of the instruction are collected and encapsulated. The feedback data packets are then used for iterative training of the deep Q-network to update the decision weights, thereby achieving adaptive closed-loop feedback.

2. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 1, characterized in that, The method for outputting the signal occlusion prediction result includes: The dynamic obstacle data is processed in multiple stages to generate a set of dynamic obstacle feature vectors containing morphological and motion parameters; The static environment data, vehicle dynamic data, and dynamic obstacle feature vector set are integrated and input into a parallel dual-branch hybrid deep learning model. Spatial feature vectors and temporal feature vectors are extracted respectively. The spatial feature vectors and temporal feature vectors are fused to construct a spatiotemporal fusion feature vector. The spatiotemporal fusion feature vector is decoded by a prediction decoding head, and the signal occlusion prediction result is output, which includes the predicted affected frequency band, the predicted signal attenuation value, the predicted start time of the occlusion event, and the predicted duration of the occlusion event.

3. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 2, characterized in that, The generation of the dynamic obstacle feature vector set of morphology and motion parameters includes morphological feature extraction and trajectory feature extraction; Based on dynamic obstacle data, the convex hull algorithm is used to extract morphological features, determine the length, width and height of the dynamic obstacle, further divide the data points into sub-cubes, and perform weighted summation based on the data point density of each sub-cube to estimate the total volume of the obstacle; The Kalman filter algorithm is used to extract the trajectory features of the centroid of the same obstacle identified in consecutive frames. By calculating the position change of the centroid between consecutive time steps, the instantaneous velocity and instantaneous motion direction angle of the obstacle are estimated.

4. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 1, characterized in that, The method for generating the communication policy instructions includes: Obtain a candidate link list and network load rate. Perform a signal link feasibility test on the candidate link list based on real-time network signal quality data. Perform a network resource availability test on the candidate link list based on the network load rate. After completing the feasibility and availability tests, construct a candidate set by combining all remaining qualified candidate links. Generate a multi-dimensional performance profile containing signal quality indicators, network load indicators, and distance indicators based on each candidate link in the candidate set. Based on real-time network signal quality data and signal blockage prediction results, a dual-mode decision triggering mechanism is used to determine whether to activate the communication decision process. If the communication decision process is activated, a deep Q-network is used to generate dynamic weight factors based on the state vector of the current communication environment, and the comprehensive score of each candidate link is calculated based on the dynamic weight factors and the multi-dimensional performance profile of each candidate link. Based on the hard constraints of communication service requirements, the candidate link ranked first in the candidate set sorted in descending order of comprehensive score is tested for hard constraints. If the candidate link meets the hard constraints, it is determined as the main communication link. If the candidate link does not meet the hard constraints, the process automatically iterates to the next candidate link in the descending order of candidate set and continues to test for hard constraints until a candidate link that meets the hard constraints is found. Based on the overall situation of current network resources, a choice is made between switching policy instructions and coordination policy instructions to generate the final communication policy instructions. The hard constraints of the communication service requirements are the performance requirements of the key data streams of the highest priority data stream tasks in the communication service requirements, including minimum guaranteed bandwidth and maximum transmission delay.

5. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 4, characterized in that, The method for choosing between the switching strategy command and the cooperative strategy command includes: When the system determines that the current network environment is relatively tense, that is, the quality of other candidate links is generally poor except for the selected main communication link, or when the main purpose of this decision is to avoid the deterministic failure risk of the current link, the system will start the switching strategy instruction generation process. When the system determines that the current network environment has sufficient resources, that is, in addition to the determined main communication link, there are multiple other high-quality links in the candidate set, the system will start the collaborative strategy instruction generation process.

6. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 5, characterized in that, The process of generating the cooperative strategy instructions includes: Assign the highest priority control task to the established main communication link and write it as the first key-value pair in the parameter dictionary; Two temporary resource pools are created: a task pool and an available link pool. The task pool contains other service flows in the communication service requirements besides the allocated key services, and they are sorted from high to low according to service priority. The available link pool contains other high-quality candidate links in the candidate set besides the main communication links that are already occupied, and they are sorted from high to low according to comprehensive score. Process the highest priority business from the task pool, and then iterate through the available link pool to find the best matching link.

7. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 6, characterized in that, The optimal matching link includes three scenarios: If a single link satisfies a scenario, and the performance of the link with the highest overall score in the available link pool meets the core requirements of the current pending business, then the link is allocated to the business, and a corresponding key-value pair is added to the parameter dictionary. At the same time, the business and the link are removed from their respective resource pools to ensure that resources are not allocated repeatedly. In link aggregation scenarios, if no single link in the available link pool can meet the current high-bandwidth service requirements, the system further determines whether the service requirements can be met by aggregating the bandwidth of multiple links. In link reuse scenarios, after completing the link allocation for high-priority tasks, the system continues to process lower-priority tasks. In addition to considering unused link resources in the available link pool, it also intelligently assesses whether there are allocated links with remaining capacity that can be reused.

8. The 5G network optimization method based on an intelligent dynamic multi-band adaptive switching system according to claim 1, characterized in that, The method for implementing the adaptive closed-loop feedback includes: Based on the instruction type of the communication strategy instruction, if the instruction type is a cooperative strategy, the optimal fragment size is dynamically calculated based on the real-time available bandwidth of each link and the target transmission time window, and dynamic fragmentation is performed on the data stream according to the optimal fragment size to generate data packets that can be transmitted in parallel across multiple links; if the instruction type is a switching strategy, the dynamic fragmentation process is skipped. Cyclic redundancy check is performed on the received data packets. Data packets that pass the check are stored in a circular buffer according to their sequence numbers for out-of-order reassembly. A bidirectional scanning reassembly algorithm is used to intelligently decode and deliver the data packets in the circular buffer in sequence, so as to restore the received data packets into a complete and ordered data stream. After executing the communication strategy instruction, the execution effect of the communication strategy instruction is evaluated. The key performance indicators collected in the evaluation are encapsulated into feedback data packets, and the feedback data packets are used as training samples for iterative training of the deep Q network to update the decision weights, thereby realizing adaptive closed-loop feedback.

9. A 5G network optimization system based on an intelligent dynamic multi-band adaptive handover system, used to implement the 5G network optimization method based on an intelligent dynamic multi-band adaptive handover system as described in any one of claims 1-8, characterized in that, include: Multi-source heterogeneous data acquisition and sensing module, intelligent signal occlusion prediction module, communication strategy instruction generation module, and closed-loop feedback optimization module; The multi-source heterogeneous data acquisition and sensing module is used to acquire real-time network signal quality data, environmental sensing data, and communication task requirements. The intelligent signal occlusion prediction module extracts a set of dynamic obstacle feature vectors based on environmental perception data, integrates the environmental perception data and the set of dynamic obstacle feature vectors, and inputs them into a parallel dual-branch hybrid deep learning model to predict signal occlusion, and outputs a signal occlusion prediction result with multi-dimensional parameters. The communication strategy instruction generation module is used to sequentially perform signal link feasibility checks and network resource availability checks on the acquired candidate links, and select qualified candidate links to construct a candidate set of multi-dimensional performance profiles. Based on the multidimensional performance profiles of each candidate link; Based on real-time network signal quality data and signal blockage prediction results, a dual-mode decision triggering mechanism is used to determine whether to activate the communication decision process. If the communication decision process is activated, a deep Q-network is used to generate dynamic weight factors based on the state vector of the current communication environment, and the comprehensive score of each candidate link is calculated based on the dynamic weight factors and the multi-dimensional performance profile of each candidate link. The dual-mode decision triggering mechanism is activated when any of the following conditions are met: If real-time network signal quality data shows that the key performance indicators of the currently used link are lower than the preset minimum availability threshold, indicating that the service quality has deteriorated unacceptably, then the passive response triggering mechanism is activated; if the signal blockage prediction results show that the current communication link will face a high probability, high attenuation signal blockage event in the near future, indicating that the system anticipates the impending service quality deterioration, then the active prevention triggering mechanism is activated; based on the hard constraints of communication service requirements, the candidate links sorted in descending order of the comprehensive score are checked one by one for hard constraints until a candidate link that meets the hard constraints is found to be the main communication link; based on the overall situation of the current network resources, a choice is made between the switching strategy instruction and the coordination strategy instruction to generate the final communication strategy instruction; The closed-loop feedback optimization module is used to determine the instruction type based on the communication strategy instruction. If the communication strategy instruction is a cooperative strategy, the data stream is dynamically fragmented to generate data packets that can be transmitted concurrently. At the receiving end, the data packets are sequentially subjected to cyclic redundancy check, out-of-order reordering, and intelligent delivery to restore them into a complete and ordered data stream. After the communication strategy instruction is executed, a feedback data packet containing the key performance indicators of the instruction is collected and encapsulated. The feedback data packet is then used for iterative training of the deep Q-network to update the decision weights, thereby achieving adaptive closed-loop feedback.

Images