A communication network resource management system and method based on intelligent scheduling

Abstract

This invention relates to the field of cloud-edge collaborative scheduling, and discloses a communication network resource management system and method based on intelligent scheduling. The intelligent scheduling-based communication network resource management method includes performing phase-domain splitting processing on real-time communication traces to disperse carrier modulation textures, edge channel congestion emergence sequences, and micro-burst session clusters into multiple dynamic interaction cavities; under the guidance of a time-varying perturbation skeleton diagram, performing reverse stitching and rhythmic bias folding on cloud-edge response loops to write the cloud-edge response loops into the instantaneous buffer loops of edge nodes; identifying cloud-edge response loops and constructing self-tuning sequence factors through signal folding and disordering in the frequency misalignment recursive layer; mapping each edge node to a potential allocation factor matrix based on the temporal reconstruction pattern in the transition state control domain; and performing layered fusion processing on the transition state control domain when the intelligent scheduling track enters the stable transition window. This invention has the advantage of improving network resource utilization.

Description

Technical Field

[0001] This invention relates to the field of cloud-edge collaborative scheduling, specifically to a communication network resource management system and method based on intelligent scheduling. Background Technology

[0002] With the rapid development of mobile internet and the Internet of Things (IoT), edge computing has been widely applied in scenarios such as subways, smart factories, and intelligent transportation. Terminal devices generate massive amounts of real-time data in high-density environments. This data needs to be processed quickly at local edge nodes to ensure low latency, and also uploaded to the cloud for unified analysis and optimization. However, existing communication network resource management methods typically rely on static allocation or single-layer scheduling strategies, which cannot achieve dynamic, cross-layer resource optimization based on network load, service priority, and edge node status. During peak hours or in the event of emergencies, data transmission between edge nodes and the cloud is prone to bandwidth congestion, increased link latency, and delays in processing critical business tasks, leading to low utilization of computing and network resources at edge computing nodes and impacting overall service quality assurance. For example, when a large number of terminals in a subway station simultaneously initiate high-definition video streams and real-time location requests, relying solely on fixed spectrum and traditional scheduling methods cannot guarantee low-latency service. Therefore, it is essential to design a communication network resource management system and method based on intelligent scheduling to improve network resource utilization. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a communication network resource management system and method based on intelligent scheduling, which has the advantage of improving network resource utilization and solves the problems mentioned in the background technology.

[0004] To achieve the aforementioned goal of improving network resource utilization, this invention provides the following technical solution: a communication network resource management method based on intelligent scheduling, comprising the following steps:

[0005] Phase splitting is performed on the real-time communication trace to break down the carrier demodulation texture, edge channel congestion emergence sequence and micro-burst session cluster into multiple dynamic interaction cavities, and a time-varying perturbation skeleton map is constructed in each interaction cavity;

[0006] Guided by the time-varying perturbation skeleton diagram, reverse stitching and rhythmic bias foldback are performed on the cloud edge response loop. The cloud edge response loop is written into the instantaneous buffer loop of the edge node, the micro-response surge signal of the edge node is extracted and injected back into the reference pulse pool in the cloud to form a frequency misalignment recursion layer.

[0007] By folding and disordering the signals in the frequency-misaligned recursive layer, the cloud edge response loop is identified and a self-tuning sequence factor is constructed. The cloud edge response loop is then non-uniformly reshaped using the self-tuning sequence factor to obtain the transition state control domain.

[0008] Based on the temporal reconstruction pattern in the transitional control domain, each edge node is mapped to a potential allocation factor matrix. The potential allocation factor matrix is ​​dynamically ordered in the cloud-edge response loop through a non-parallel folding transformation across nodes, forming a hierarchical intelligent scheduling track.

[0009] When the intelligent scheduling track enters the stable transition window, the transition state control domain is subjected to layered fusion processing to generate a pan-domain collaborative potential field that characterizes the global reuse trend of communication resources.

[0010] Preferably, the process of constructing a time-varying perturbation skeleton diagram within each interaction cavity is as follows:

[0011] The carrier demodulation texture fragments obtained after phase decomposition are used to perform phase slicing on the communication traces in each dynamic interactive cavity, and rhythmic jump segments are extracted from the slice sequence.

[0012] Based on the rhythmic jump segment, the local intensity envelope of the edge channel congestion emergence sequence is constructed, and the generation rate of micro-burst conversation clusters is combined to form the intracavity perturbation factor vector.

[0013] The perturbation factor vector is used to perform rhythm recalibration processing on the time scale within the interactive cavity to form a new time mapping axis;

[0014] On the time-mapping axis, a topological fragment stitching algorithm is used to connect discrete perturbation segments to generate a time-varying perturbation skeleton diagram describing the perturbation propagation mechanism of the interactive cavity.

[0015] Preferably, the process of writing the cloud-edge response loop into the transient cache loop of the edge node is as follows:

[0016] Based on the temporal directionality of the perturbation propagation path in the time-varying perturbation skeleton diagram, the response segments sent from the cloud are deduced in reverse along the perturbation skeleton to obtain the rhythmic bias in the cloud-edge link.

[0017] Perform a backoff correction on the rhythm bias, and re-stitch the response links that were originally aggregated to the cloud back to the edge nodes along the reverse path of the skeleton graph;

[0018] A transient cache loop is generated at the edge node, and the cache duration in the loop is synchronized with the rhythm bias. The cloud edge response fragment after reverse stitching is written into the transient cache loop.

[0019] Preferably, the process of forming the frequency-shifting recursive layer is as follows:

[0020] Amplitude fluctuation contact detection is performed on response segments in the instantaneous buffer loop to identify micro-response surge points that can characterize short-term scheduling state transitions of edge nodes;

[0021] The micro-response surge points are divided into several surge clusters according to surge rate, duration, and local disturbance background, and a local recursive fragment with temporal ambiguity is constructed for each surge cluster;

[0022] Local recursive fragments are aligned with frequency shifts according to the rhythmic differences of the rising clusters, forming a stacked frequency shift structure in the time domain;

[0023] The frequency misalignment structure is injected back into the reference pulse pool in the cloud as pulse fragments, and a frequency misalignment recursive layer is generated by recursive superposition.

[0024] Preferably, the process of identifying cloud-edge response loops and constructing self-tuning sequence factors is as follows:

[0025] Based on the time back-and-forth and amplitude jump phenomena of each pulse segment in the frequency misalignment recursion layer, the back-and-forth alignment relationship between segments is structurally classified.

[0026] Based on the foldback relationship, an out-of-order indicator vector is constructed to express the reorganization partial order relationship generated by the cloud-edge response loop under the frequency mismapping.

[0027] By comparing the out-of-order indicator vector with the original response link across layers, three types of response segments in the link are identified: mismatched segments, delayed segments, and premature segments.

[0028] A self-regulating causal extraction mechanism is adopted to extract key features from the rearranged response fragments that enable the stable convergence of the response chain reconstruction, and combine them to form self-regulating sequence factors.

[0029] Preferably, the process of obtaining the transition state control domain is as follows:

[0030] By embedding self-tuned sequence factors into cloud edge response loops and performing non-uniform weight redistribution on adjacent segments of the loop, local convergence is observed in low-load segments.

[0031] Under the alternating action of expansion and convergence operations, the time structure and topological structure of the cloud edge response loop are simultaneously twisted, forming a response surface with local plasticity.

[0032] By using multi-scale folded vectors to segmentally adjust the response surface, a differentiated response density is formed between the densely distributed edge node area and the cloud scheduling core area. The adjusted response surface is then used as the transition state control domain.

[0033] Preferably, the process of mapping each edge node to a potential allocation factor matrix is ​​as follows:

[0034] In the transition state control domain, the scheduling response trajectory of the edge nodes is temporally deconstructed according to the gradient direction of the control domain and the temporal expansion and contraction relationship.

[0035] The deconstructed node trajectories are constructed into three-component descriptors based on their displacement amplitude, response sensitivity, and link dependency in the control domain.

[0036] By inputting the three-component descriptor into the multidimensional factor mapping model, the potential factor describing the nodal force field is obtained.

[0037] Arrange the relative positions of the potential factors of all nodes in the control domain into an array structure to form a potential allocation factor matrix.

[0038] Preferably, the process of forming a hierarchical intelligent scheduling track is as follows:

[0039] The interaction degree between any pair of nodes in the potential distribution factor matrix is ​​analyzed, and the transformation correlation diagram between the pairs of nodes is constructed.

[0040] Perform timing transformation on the critical path in the folded correlation graph using a non-parallel folded transformation formula;

[0041] The action amplitude is rearranged according to the node sequence after time-sequence transformation, and a hierarchical intelligent scheduling track is generated based on the rearrangement result.

[0042] Preferably, the process of generating a pan-domain cooperative potential field that characterizes the global reuse trend of communication resources is as follows:

[0043] After detecting that the track has entered the stable transition window during the time-stable segment of the hierarchical intelligent scheduling track, hierarchical sampling is performed on the transition state control domain to extract the response modes of different levels into multi-layer feature fragments.

[0044] Cross-layer connectivity processing is performed on multi-layer feature segments, incorporating time stability, link consistency, and node potential gradient into the calculation.

[0045] In the unified feature space obtained after fusion, the main direction reflecting the overall load flow, response rhythm coordination and resource reuse trend is identified, and a pan-domain collaborative potential field is generated.

[0046] A communication network resource management system based on intelligent scheduling, comprising:

[0047] Time-varying perturbation module: Disperses carrier demodulation texture, edge channel congestion sequence and micro-burst session clusters into dynamic interactive cavity, and constructs time-varying perturbation skeleton map in each cavity;

[0048] Frequency misalignment recursion module: Writes the cloud-edge response loop into the instantaneous buffer of the edge node and injects the edge micro-response signal back to the cloud to form a frequency misalignment recursion layer;

[0049] Sequence reshaping module: Constructs self-tuned sequence factors by folding out disordered signals from the frequency-misaligned recursive layer, performs non-uniform reshaping of cloud edge loops, and generates transition state control domain;

[0050] Dynamic sequencing module: Maps each edge node to a potential allocation factor matrix, and forms a hierarchical intelligent scheduling track through non-parallel folding transformation adjustment;

[0051] Layered fusion module: Performs layered fusion processing on the transition state control domain to generate a generalized cooperative potential field.

[0052] Compared with the prior art, the present invention provides a communication network resource management system and method based on intelligent scheduling, which has the following beneficial effects:

[0053] This invention achieves global perception and refined management of cloud and edge node scheduling behavior by performing phase-separated domain splitting on real-time communication traces, constructing a time-varying perturbation skeleton diagram, and forming a frequency-misaligned recursive layer in the cloud-edge collaborative loop. Through the construction and non-uniform reshaping of self-adjusting sequence factors, a transitional control domain with local plasticity and global coordination is obtained, enabling edge node responses to adaptively adjust under different loads and burst conditions. Mapping edge nodes to potential allocation factor matrices and forming hierarchical intelligent scheduling links ensures dynamic optimization of node forces and link scheduling at multiple levels and dimensions. By generating a pan-domain collaborative potential field through the layered fusion calculation of the transitional control domain, a quantitative characterization and prediction of the global reuse trend of communication resources is achieved, providing real-time reference for cloud scheduling strategies and edge resource regrouping. This significantly improves network resource utilization, reduces link conflict and congestion risks, enhances the response stability of cloud-edge collaboration, and improves the network's adaptability under high dynamic loads and micro-burst events, achieving efficient, intelligent, and sustainable communication system scheduling. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the method of the present invention;

[0055] Figure 2 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0056] 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.

[0057] Example 1: Please refer to Figure 1 As shown in the figure, a communication network resource management method based on intelligent scheduling in an embodiment of the present invention includes the following steps:

[0058] S1: Perform phase splitting processing on the real-time communication trace, and disperse the carrier demodulation texture, edge channel congestion emergence sequence and micro-burst session cluster into multiple dynamic interaction cavities, and construct a time-varying perturbation skeleton map in each interaction cavity.

[0059] The process of constructing a time-varying perturbation skeleton diagram in each interaction cavity in S1 is as follows:

[0060] Using the carrier demodulation texture fragments obtained after phase decomposition, phase slicing is performed on the communication traces in each dynamic interactive cavity, and rhythmic jump segments are extracted from the slice sequence. The original communication traces are decomposed in the mixed time and carrier domains. Using the phase jump markers in the carrier demodulation texture fragments, the traces in the interactive cavity are sliced ​​point by point at fixed phase intervals, so that the originally continuous traces are decomposed into several sub-segments with independent phase meanings. Based on the phase continuity and energy distribution pattern, slice segments with significant rhythmic jump characteristics are identified. These segments often correspond to scenarios such as sudden increase in link load, instantaneous aggregation of scheduling commands, or dense startup of micro-sessions. The extracted rhythmic jump segments maintain a strict time index correspondence with the original traces to ensure the accuracy of disturbance analysis.

[0061] Based on the rhythmic jump segment, a local intensity envelope of the congestion emergence sequence in the edge channel is constructed, and the generation rate of micro-burst conversation clusters is combined to form an intracavity perturbation factor vector. For the rhythmic jump segment, the local energy distribution is extracted, and the intensity fluctuations formed in the time domain are used to establish an intensity envelope curve, which reflects the expansion speed and duration of the sudden congestion sequence in the edge channel, thereby providing micro-level flow fluctuation clues. The occurrence frequency, aggregation density and short-cycle refresh pattern of micro-burst conversation clusters are statistically analyzed to form a dynamic index reflecting the intensity of conversation activity transition. The intensity envelope and the conversation cluster generation rate are aligned by time label to form a set of perturbation factor vectors characterizing the overall perturbation morphology in the cavity.

[0062] The time scale within the interaction cavity is recalibrated using the perturbation factor vector to form a new time mapping axis. Regions where the perturbation factor vector shows peaks are located on the time axis and marked as high-sensitivity segments. To reveal the sequence of events within the cavity, these high-sensitivity segments are assigned higher time resolution, which is manifested as a stretching of the time axis. For segments with gentle fluctuations and sparse activity in the perturbation factor vector, they are considered low-sensitivity segments and given compressed time resolution, so that they occupy less span in the new time axis. Through this rhythm recalibration method, the resulting time mapping axis can present the actual propagation trajectory of the perturbation in a way that better conforms to the event rhythm.

[0063] On the time-mapping axis, a topological fragment stitching algorithm is used to connect discrete disturbance segments, generating a time-varying disturbance skeleton diagram describing the disturbance propagation mechanism of the interaction cavity. All rhythmic jump segments and their associated disturbance segments are repositioned in the new temporal framework, and a preliminary chain structure is constructed according to the temporal order. The topological fragment stitching algorithm is used to identify, align, and merge the boundaries between these segments, and to check whether there are shared action trigger points, resource access points, or micro-session inheritance relationships between segments. For segments that form a continuous disturbance chain, they are stitched together in the form of nodes and edges to form a coherent structure. For segments with jumps or discontinuities, local interpolation and event backtracking are used to determine their reasonable connection method in the skeleton. The resulting time-varying disturbance skeleton diagram can show the generation, diffusion, convergence, and decay process of disturbances in the interaction cavity in a dynamic topological form.

[0064] S2: Guided by the time-varying perturbation skeleton diagram, reverse stitching and rhythmic bias foldback are performed on the cloud edge response loop. The cloud edge response loop is written into the instantaneous buffer loop of the edge node, the micro-response surge signal of the edge node is extracted and injected back into the reference pulse pool in the cloud to form a frequency misalignment recursion layer.

[0065] The process of writing the cloud-edge response loop into the transient cache loop of the edge node in S2 is as follows:

[0066] Based on the temporal directionality of the perturbation propagation path in the time-varying perturbation skeleton diagram, the response segments sent from the cloud are reverse-engineered along the perturbation skeleton to obtain the rhythmic bias in the cloud-edge link. Using the time-varying perturbation skeleton diagram as an index, the original projection position of the response segments sent from the cloud in the skeleton diagram is located (this projection is determined by the triplet of the response segment's time stamp, trigger event identifier, and resource access identifier). The propagation edges of the skeleton diagram are traversed in reverse, hop by hop back to the edge node cluster. During the process, the time interval difference, trigger delay, and link state changes of each hop are recorded and serialized and stored as a reverse trajectory log. Based on the reverse trajectory log, the temporal offset features and rhythmic anomaly features describing the cloud-edge link in this segment are extracted. These features are combined to form the rhythmic bias, with time offset description, event triggering relative position, and intensity indication as the core attributes.

[0067] A backtracking correction is performed on the rhythm bias, re-stitching the response chain that was originally aggregated to the cloud back to the edge node along the reverse path of the skeleton graph. At each target edge node, a list of response segments requiring compensation or rearrangement is calculated, with entries including the original sequence identifier, expected write point, and desired time window constraints. The correction process is implemented through a three-step mechanism: time relocation, fine-tuning the internal time tags of the response segments according to the relative displacement indicated in the rhythm bias to match the local timeline baseline of the edge node; event chain reconnection, re-establishing the sequential constraints between segments based on trigger dependencies and resource access relationships to ensure logical consistency of the reassembled chain; and consistency verification, using lightweight sequence consistency checks (e.g., forward checks and backward confirmations based on event signatures) to determine whether the reconnected chain segments meet the executability conditions. For segments that conflict or do not meet the time window during the correction process, delayed writing, segmented downsampling, or rollback to the cloud for readjustment are selected based on priority marking, thus ensuring that the stitching operation respects the real-time capabilities of the edge nodes while maintaining the integrity of the overall response chain.

[0068] A transient cache loop is generated at the edge node, and the cache duration in the loop is synchronized with the rhythm bias. The reverse-stitched cloud-edge response fragment is written into the transient cache loop. After receiving the stitching instruction, the edge node dynamically creates or adjusts the transient cache loop to carry the response fragment locally. The cache loop is based on a circular buffer structure and combined with time window index and version number management to ensure the traceability of concurrent writes. The cache duration is determined by two parts: the target storage window indicated by the rhythm bias (the minimum length to ensure that the fragment can be triggered or traced locally); and adaptive pruning based on the current load and storage availability of the edge node (automatically reducing the residence time of non-critical fragments under high load). During the write phase, the reverse-stitched response fragment is written to the corresponding position in the loop according to the corrected timestamp, and a lightweight metadata entry is generated at the same time as the write.

[0069] The process of forming the frequency-misaligned recursive layer in S2 is as follows:

[0070] Amplitude fluctuation touchpoint detection is performed on response segments in the transient buffer loop to identify micro-response surge points that characterize short-term scheduling state transitions of edge nodes. The cloud-edge response segments and their metadata (including original cloud identifier, local time stamp, confidence level, etc.) written to the transient buffer loop of the edge node are read in time sequence. Noise baseline estimation is first performed on the amplitude sequence of each segment to eliminate the observation bias caused by differences in receiver sensitivity and local interference. A multi-criteria touchpoint detector is used to distinguish amplitude fluctuations. The detector can apply three strategies in parallel: envelope peak detection, short-term energy jump judgment, and spectrum transient identification. The final touchpoint candidate is output by majority voting or weighted decision. For each touchpoint candidate, its trigger time, amplitude surge increment, duration before and after, and trigger background (such as the number of concurrent sessions and channel quality indicators) are recorded and labeled with a confidence score as the micro-response surge point output.

[0071] Micro-response surge points are divided into several surge clusters based on surge rate, duration, and local disturbance background. For each surge cluster, a local recursive fragment with temporal ambiguity is constructed. Surge points are standardized in the time domain and attribute space (surge rate, duration, local session density, and channel background). Surge clusters are identified using density-based or hierarchical clustering methods (using time-constrained spectral clustering to capture clusters with similar temporal and attribute characteristics). For each cluster, the temporal distribution range and attribute distribution characteristics of surge points within the cluster are statistically analyzed, and a prototype description of the local recursive fragment is generated based on this. Considering the temporal uncertainty and synchronization deviation in edge observations, the local recursive fragment is not a precise time period but a temporally ambiguous interval description (e.g., described by a combination of start and end intervals and confidence intervals), including fuzzy start and end boundaries, typical surge spectrum, representative amplitude curves within the cluster, and confidence profiles.

[0072] Local recursive fragments are aligned by frequency misalignment according to the rhythm differences of the surge clusters, forming a stacked frequency misalignment structure in the time domain. The rhythm characteristics of each fragment (such as surge interval distribution, relative position of main frequency components, and periodicity measurement) are compared and sorted to identify fragment groups with large rhythm differences. Frequency misalignment strategies are adopted for different rhythm groups: for faster rhythm fragments, their high-resolution temporal representation is preserved and adjacent fragments are allowed to be misaligned during the alignment process, while slower rhythm fragments are expressed in a time-stretched manner so that they can be presented in parallel within the same time window. The alignment process focuses on preserving the fuzzy boundary information of each local recursive fragment. The actual alignment position is determined by a priority criterion based on event similarity and sequence relationship to avoid information loss due to mechanical overlap. After frequency misalignment, the fragments are arranged in a stacked manner in the time domain to form a multi-level frequency misalignment structure. Each level corresponds to a different rhythm cluster and carries its own confidence weight and temporal fuzzy interval, which facilitates the identification, weighting, or priority processing of these levels in the cloud.

[0073] The frequency misalignment structure is injected back into the reference pulse pool in the cloud as pulse fragments, and a frequency misalignment recursive layer is generated through recursive superposition. The generated layered frequency misalignment structure is transformed into a lightweight set of pulse fragments. Each pulse fragment includes the original fragment identifier, local representative sequence, temporal ambiguity boundary, confidence label, and source edge node identifier. To save bandwidth and accelerate cloud processing, the pulse fragments are compressed before injection (e.g., representative sampling points are selected, principal components are extracted, or low-confidence parts are pruned according to importance). A summary metadata is also attached for fast indexing in the cloud. The pulse fragments are transmitted back to the reference pulse pool in the cloud through a secure and reliable uplink channel. After receiving them, the cloud recursively superimposes these pulse fragments according to temporal order and source grouping: fragments from the same edge domain or with similar rhythms are locally synthesized to improve the signal-to-noise ratio. Then, at a higher level, different rhythm layers are fused and sorted to form a frequency misalignment recursive layer. This retains the independent rhythm information of each layer and reveals the trigger chain and potential amplification path across layers through recursive superposition. This information is used by the global scheduler in the cloud to determine resource priorities, issue correction strategies, or trigger cross-domain coordination actions.

[0074] S3: By folding and disordering the signals in the frequency-misaligned recursive layer, the cloud edge response loop is identified and a self-tuning sequence factor is constructed. The cloud edge response loop is then non-uniformly reshaped using the self-tuning sequence factor to obtain the transition state control domain.

[0075] The process of identifying cloud-edge response loops and constructing self-tuning sequence factors in S3 is as follows:

[0076] Based on the time reversal and amplitude jump phenomena of each pulse segment in the frequency recursion layer, the reversal alignment relationship between segments is structurally classified; all pulse segments and their time boundaries, amplitude characteristics and source identifiers are read from the generated frequency recursion layer, and the reversal events of each segment on the time axis are identified, including amplitude segments that regress forward, extend backward or increase locally. Based on these reversal events and amplitude jump characteristics, the segments are grouped according to their reversal patterns, such as being classified into the categories of sequence preservation, delayed response or early triggering, and the time span, amplitude magnitude and rhythm characteristics of each type of segment are recorded.

[0077] Based on the back-and-forth relationship, an out-of-order indicator vector is constructed to express the reorganization partial order relationship generated by the cloud-edge response loop under frequency mismapping. The time sequence and back-and-forth mode of each segment are encoded, and its deviation from the original cloud-edge link under frequency mismapping is marked as advanced, delayed, or order-preserved. The partial order state of each segment is summarized into a vectorized form to form the out-of-order indicator vector. Each vector element corresponds to the position, deviation type, and amplitude change characteristics of the segment in the link. The out-of-order indicator vector is grouped by time period or level to represent the response reorganization of different rhythms or different edge domains in the frequency mismapping recursion layer. It can reveal local misalignment and rhythm mismatch in the link and provide the cloud scheduler with a quantitative basis for analyzing link stability, identifying potential conflicts, or optimizing synchronization strategies.

[0078] By comparing the out-of-order indicator vector with the original response link across layers, three types of response segments in the link are identified: mismatched segments, delayed segments, and premature segments. The generated out-of-order indicator vector is then mapped and compared layer by layer with the original link sequence of the cloud-edge response loop. By comparing the start and end positions of each segment, amplitude changes, and rhythmic characteristics, the type of offset relative to the original link is determined. Segments with significant deviations and sudden increases in amplitude are marked as mismatched segments, indicating misalignment of information or control triggers in the link. Segments with time lags are marked as delayed segments, representing lag in cloud commands or edge responses. Segments that are prematurely triggered and whose amplitudes are within the expected range are marked as premature segments, representing the advanced behavior of edge nodes in response to cloud responses. Each segment is also accompanied by its hierarchical information, source node identifier, and confidence score, providing accurate data support for the construction of self-tuning sequences and allowing for fault tolerance processing of abnormal or incomplete segments.

[0079] A self-regulating causal extraction mechanism is employed to extract key features from the rearranged response fragments that enable stable convergence of the response chain reconstruction, combining them to form self-regulating sequence factors. Identified mismatched, delayed, and premature segments are analyzed, and the self-regulating causal extraction mechanism is used to evaluate the impact of each segment on the overall cloud-edge response chain, including its triggering stability, amplitude fluctuation characteristics, rhythm consistency, and dependency on adjacent segments. By selecting key segments that can improve link convergence, reduce conflicts, and optimize rhythm alignment, the features of these segments (such as time interval, amplitude range, rhythm identifier, and node origin) are combined to form self-regulating sequence factors. Each self-regulating sequence factor acts as a stabilizing primitive during link reconstruction, preserving local dynamic information and being used for global scheduling strategy optimization, achieving dynamic adaptive coordination between cloud and edge responses. In implementation, dynamic updates and priority adjustments of self-regulating sequence factors are supported, enabling them to cope with continuous changes in link status and sudden load disturbances.

[0080] The process of obtaining the transition state control domain in S3 is as follows:

[0081] By embedding self-tuning sequence factors into the cloud-edge response loop, non-uniform weight redistribution is performed on adjacent segments of the loop, resulting in local convergence in low-load sections. Key segment features are extracted from the generated self-tuning sequence factors and embedded into the corresponding positions in the cloud-edge response loop, so that the trigger weight of each segment in the loop is coupled with the original link state. For adjacent response segments, non-uniform weight adjustment is performed according to their load conditions, node response rates, and amplitude differences. That is, a certain dynamic fluctuation is maintained in high-load sections, and the weight convergence effect is increased in low-load sections, so that these low-load segments form a local steady state. Through this embedding and weight redistribution, the time delay and rhythm deviation that originally existed in the cloud-edge loop are buffered, ensuring that the edge nodes respond stably under low-load conditions and avoiding local overshoot or resource waste.

[0082] Through the alternating action of expansion and convergence operations, the time structure and topology of the cloud-edge response loop are simultaneously twisted, forming a response surface with local plasticity. In high-load or dynamically fluctuating sections, the time span and topology connectivity are extended through expansion operations, while in low-load sections, the time span is compressed and the node connection weights are strengthened through convergence operations. During this process, the response paths, triggering sequences, and link dependencies between the cloud and edge nodes are gradually adjusted. The interaction between the time structure and the topology generates twisting, causing the originally linear or single-path response links to form a surface structure similar to wave undulations. The surface structure is moderately adaptively adjusted by the edge nodes and the cloud scheduling core according to the local load and link status without destroying the continuity of the global link, thus providing an operable basic structure for multi-scale segmented adjustment.

[0083] By using multi-scale folded vectors to segmentally adjust the response surface, differentiated response densities are formed between densely distributed edge nodes and the core cloud scheduling region. The adjusted response surface serves as the transitional control domain. Folded vectors covering different spatial and temporal scales are generated to characterize the load density, node distribution, and link rhythm characteristics of each segment in the response surface. The surface is segmentally adjusted according to the folded vectors, that is, the distribution density of local response segments is enhanced in densely distributed edge node regions to improve node processing capacity and scheduling accuracy, while the response density is moderately compressed in the core cloud scheduling region to form a gradient distribution, reflecting the difference between the core scheduling capacity and the edge node carrying capacity. Through this multi-scale adjusted response surface, the dynamic load change characteristics and link adaptive capabilities under cloud-edge collaboration can be captured and used as the transitional control domain.

[0084] S4: Based on the temporal reconstruction pattern in the transitional control domain, each edge node is mapped to a potential allocation factor matrix. The potential allocation factor matrix is ​​dynamically ordered in the cloud-edge response loop through a non-parallel folding transformation across nodes, forming a hierarchical intelligent scheduling track.

[0085] The process of mapping each edge node to a potential allocation factor matrix in S4 is as follows:

[0086] In the transitional control domain, based on the gradient direction of the control domain and the temporal scaling relationship, the scheduling response trajectory of the edge nodes is temporally defolded. The response surface of each node in the transitional control domain is analyzed, including local load density, link rhythm, and time scaling characteristics. Along the gradient direction of the control domain, the scheduling response trajectory of the node is unfolded to eliminate the nonlinear overlap caused by time scaling and topology twisting, so that the node response behavior presents a continuous and comparable sequence on the time axis. Through this temporal defolding operation, the dynamic scheduling state of each edge node in the control domain can be accurately mapped.

[0087] The deconstructed node trajectories are constructed into three-component descriptors based on their displacement amplitude, response sensitivity, and link dependency in the control domain. Three key features are extracted from each deconstructed node trajectory: displacement amplitude, which quantifies the degree of movement of the node along the time and topology directions in the control domain; response sensitivity, which characterizes the node's response strength to cloud scheduling commands and changes in the state of neighboring nodes; and link dependency, which reflects the interaction between the node and its neighboring nodes and cloud paths. These three features are integrated into a unified three-component descriptor, enabling the scheduling behavior of each node to be fully represented in multiple dimensions, providing rich input information for potential factor calculation.

[0088] By inputting the three-component descriptor into the multidimensional factor mapping model, potential factors describing the force field of the nodes are obtained. The constructed three-component descriptor is then input into the multidimensional factor mapping model. Through nonlinear mapping and feature weighting operations, the displacement amplitude, response sensitivity, and link dependence of each node are converted into potential factors of a uniform scale. The potential factors are indicators of the force intensity or influence of the node in the control domain, reflecting the node's contribution to the overall scheduling link and its adjustability. Through this mapping, each node is transformed from the original trajectory feature space to the force field representation space, while taking into account the dynamic influence of the node under different load and topology conditions.

[0089] The potential factors of all nodes are arranged in an array structure according to their relative positions in the control domain, forming a potential allocation factor matrix. Based on the spatial topology and temporal distribution order in the control domain, the potential factors of each node are arranged according to their position in the transition state control domain, generating a two-dimensional or multi-dimensional array structure. In the array structure, the relative positions of adjacent elements reflect the topological proximity and response timing relationship between nodes, while the value of each element represents the magnitude of the force exerted by that node. The resulting potential allocation factor matrix can not only intuitively demonstrate the scheduling capability of each edge node in the control domain, but also provide a complete quantitative basis for the generation of non-parallel folding transformation and hierarchical intelligent scheduling links, while ensuring the operability of global link adjustment behavior.

[0090] The process of forming a hierarchical intelligent scheduling link in S4 is as follows:

[0091] The interaction degree between any pair of nodes in the potential allocation factor matrix is ​​analyzed, and a folded correlation graph between the node pairs is constructed. Pairwise analysis is performed on each pair of nodes in the potential allocation factor matrix to evaluate their mutual influence in the cloud-edge response loop, including the complementarity of the forces between nodes, the relativity of the response timing, and the topological proximity in the control domain. The interaction relationship of each pair of nodes is graphically represented in the form of nodes as edges and the intensity of the interaction as weights, forming a folded correlation graph. In the folded correlation graph, the weight of the edge reflects the mutual interference or cooperation effect that the node pairs may produce in the link scheduling, while the topological connectivity between nodes reflects the potential critical path of the scheduling link.

[0092] The critical path in the folded correlation graph is subjected to time-series transformation using a non-parallel folding transformation method. The time sequence of nodes on the critical path is reordered and adjusted to ensure that the action order between nodes is adapted to the nonlinear characteristics of the actual response rhythm. During the transformation process, the original response amplitude of the nodes, the topological interval between nodes, and the link dependency are all taken into account in the adjustment strategy. This ensures that the critical path can maintain the overall link stability in the adjusted link and reflect the non-parallel time sequence distribution characteristics, thereby achieving efficient dynamic arrangement of scheduling links.

[0093] The impact amplitudes of nodes are rearranged according to the time-series transformation, and a hierarchical intelligent scheduling link is generated based on the rearrangement result. According to the node order after the critical path transformation, the impact amplitudes of all nodes are rearranged along the rearranged sequence. Through this arrangement, the position of each node in the link reflects its priority, impact force, and degree of influence on adjacent nodes in the cloud-edge scheduling process. The resulting hierarchical intelligent scheduling link not only presents the hierarchical structure characteristics of the link, but also reflects the rhythmic misalignment and nonlinear interaction mode between nodes. This enables the entire cloud-edge scheduling link to achieve hierarchical resource allocation and dynamic optimization while ensuring timely response and stability, providing an operable scheduling basis for improving network resource utilization.

[0094] S5: When the intelligent scheduling track enters the stable transition window, the transition state control domain is subjected to layered fusion processing to generate a pan-domain collaborative potential field that characterizes the global reuse trend of communication resources.

[0095] The process of generating the pan-domain cooperative potential field representing the global reuse trend of communication resources in S5 is as follows:

[0096] After detecting that the track has entered the stable transition window during the time-stable segment of the hierarchical intelligent scheduling track, hierarchical sampling is performed on the transition state control domain to extract the response modes of different levels into multi-level feature segments. The time series of the hierarchical intelligent scheduling track is monitored. When the response fluctuation of the track enters the controllable stable segment, that is, when the track reaches the stable transition window state, the response signal in the transition state control domain is divided into layers. Sampling operations are performed on the node response, link amplitude and time rhythm of different levels to divide the key response modes of each level into several feature segments. These feature segments can reflect the instantaneous behavior, local resource occupation and response interaction of nodes and links at different time scales.

[0097] Cross-layer interconnection processing is performed on multi-layer feature segments, incorporating time stability, link consistency, and node potential gradient into the calculation. The correlation between feature segments of each layer is analyzed, and response modes of different layers are integrated through cross-layer interconnection processing. During the processing, the time stability index of each feature segment, the synchronization consistency between links, and the potential gradient change of nodes in the control domain are considered simultaneously. By comprehensively analyzing this information, the degree of response coordination, link scheduling matching degree, and node role strength between different layers can be quantified, realizing multi-layer and multi-dimensional feature integration.

[0098] In the unified feature space obtained after fusion, the main direction reflecting the overall load flow, response rhythm coordination, and resource reuse trend is identified, and a pan-domain collaborative potential field is generated. The processed multi-layer feature fragments are mapped to the unified feature space. In the unified feature space, the load flow characteristics, response rhythm coordination relationship, and resource occupation and release trend of the overall cloud-edge network are comprehensively analyzed. By identifying the dominant direction and pattern in this space, the key evolution path reflecting the global communication resource reuse is determined and represented as a pan-domain collaborative potential field. This not only indicates the current network resource usage status, but also guides the cloud scheduler and edge nodes to dynamically optimize resource allocation in future scheduling cycles, thereby improving the efficiency and stability of the entire network resource reuse.

[0099] Example 2: As Figure 2 As shown, a communication network resource management system based on intelligent scheduling includes:

[0100] Time-varying perturbation module: Disperses carrier demodulation texture, edge channel congestion sequence and micro-burst session clusters into dynamic interactive cavity, and constructs time-varying perturbation skeleton map in each cavity;

[0101] Frequency misalignment recursion module: Writes the cloud-edge response loop into the instantaneous buffer of the edge node and injects the edge micro-response signal back to the cloud to form a frequency misalignment recursion layer;

[0102] Sequence reshaping module: Constructs self-tuned sequence factors by folding out disordered signals from the frequency-misaligned recursive layer, performs non-uniform reshaping of cloud edge loops, and generates transition state control domain;

[0103] Dynamic sequencing module: Maps each edge node to a potential allocation factor matrix, and forms a hierarchical intelligent scheduling track through non-parallel folding transformation adjustment;

[0104] Layered fusion module: Performs layered fusion processing on the transition state control domain to generate a generalized cooperative potential field.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A communication network resource management method based on intelligent scheduling, characterized in that, Includes the following steps: Phase splitting is performed on the real-time communication trace to break down the carrier demodulation texture, edge channel congestion emergence sequence and micro-burst session cluster into multiple dynamic interaction cavities, and a time-varying perturbation skeleton map is constructed in each interaction cavity; Guided by the time-varying perturbation skeleton diagram, reverse stitching and rhythmic bias foldback are performed on the cloud edge response loop. The cloud edge response loop is written into the instantaneous buffer loop of the edge node, the micro-response surge signal of the edge node is extracted and injected back into the reference pulse pool in the cloud to form a frequency misalignment recursion layer. By folding and disordering the signals in the frequency-misaligned recursive layer, the cloud edge response loop is identified and a self-tuning sequence factor is constructed. The cloud edge response loop is then non-uniformly reshaped using the self-tuning sequence factor to obtain the transition state control domain. Based on the temporal reconstruction pattern in the transitional control domain, each edge node is mapped to a potential allocation factor matrix. The potential allocation factor matrix is ​​dynamically ordered in the cloud-edge response loop through a non-parallel folding transformation across nodes, forming a hierarchical intelligent scheduling track. When the intelligent scheduling track enters the stable transition window, the transition state control domain is subjected to layered fusion processing to generate a pan-domain collaborative potential field that characterizes the global reuse trend of communication resources.

2. The communication network resource management method based on intelligent scheduling according to claim 1, characterized in that, The process of constructing a time-varying perturbation skeleton diagram within each interaction cavity is as follows: The carrier demodulation texture fragments obtained after phase decomposition are used to perform phase slicing on the communication traces in each dynamic interactive cavity, and rhythmic jump segments are extracted from the slice sequence. Based on the rhythmic jump segment, the local intensity envelope of the edge channel congestion emergence sequence is constructed, and the generation rate of micro-burst conversation clusters is combined to form the intracavity perturbation factor vector. The perturbation factor vector is used to perform rhythm recalibration processing on the time scale within the interactive cavity to form a new time mapping axis; On the time-mapping axis, a topological fragment stitching algorithm is used to connect discrete perturbation segments to generate a time-varying perturbation skeleton diagram describing the perturbation propagation mechanism of the interactive cavity.

3. The communication network resource management method based on intelligent scheduling according to claim 2, characterized in that, The process of writing the cloud-edge response loop into the transient cache loop of the edge node is as follows: Based on the temporal directionality of the perturbation propagation path in the time-varying perturbation skeleton diagram, the response segments sent from the cloud are deduced in reverse along the perturbation skeleton to obtain the rhythmic bias in the cloud-edge link. Perform a backoff correction on the rhythm bias, and re-stitch the response links that were originally aggregated to the cloud back to the edge nodes along the reverse path of the skeleton graph; A transient cache loop is generated at the edge node, and the cache duration in the loop is synchronized with the rhythm bias. The cloud edge response fragment after reverse stitching is written into the transient cache loop.

4. The communication network resource management method based on intelligent scheduling according to claim 3, characterized in that, The process of forming a frequency misrecursive layer is as follows: Amplitude fluctuation contact detection is performed on response segments in the instantaneous buffer loop to identify micro-response surge points that can characterize short-term scheduling state transitions of edge nodes; The micro-response surge points are divided into several surge clusters according to surge rate, duration, and local disturbance background, and a local recursive fragment with temporal ambiguity is constructed for each surge cluster; Local recursive fragments are aligned with frequency shifts according to the rhythmic differences of the rising clusters, forming a stacked frequency shift structure in the time domain; The frequency misalignment structure is injected back into the reference pulse pool in the cloud as pulse fragments, and a frequency misalignment recursive layer is generated by recursive superposition.

5. The communication network resource management method based on intelligent scheduling according to claim 4, characterized in that, The process of identifying cloud-edge response loops and constructing self-tuning sequence factors is as follows: Based on the time back-and-forth and amplitude jump phenomena of each pulse segment in the frequency misalignment recursion layer, the back-and-forth alignment relationship between segments is structurally classified. Based on the foldback relationship, an out-of-order indicator vector is constructed to express the reorganization partial order relationship generated by the cloud-edge response loop under the frequency mismapping. By comparing the out-of-order indicator vector with the original response link across layers, three types of response segments in the link are identified: mismatched segments, delayed segments, and premature segments. A self-regulating causal extraction mechanism is adopted to extract key features from the rearranged response fragments that enable the stable convergence of the response chain reconstruction, and combine them to form self-regulating sequence factors.

6. The communication network resource management method based on intelligent scheduling according to claim 5, characterized in that, The process of obtaining the transition state control domain is as follows: By embedding self-tuned sequence factors into cloud edge response loops and performing non-uniform weight redistribution on adjacent segments of the loop, local convergence is observed in low-load segments. Under the alternating action of expansion and convergence operations, the time structure and topological structure of the cloud edge response loop are simultaneously twisted, forming a response surface with local plasticity. By using multi-scale folded vectors to segmentally adjust the response surface, a differentiated response density is formed between the densely distributed edge node area and the cloud scheduling core area. The adjusted response surface is then used as the transition state control domain.

7. A communication network resource management method based on intelligent scheduling according to claim 6, characterized in that, The process of mapping each edge node to a potential allocation factor matrix is ​​as follows: In the transition state control domain, the scheduling response trajectory of the edge nodes is temporally deconstructed according to the gradient direction of the control domain and the temporal expansion and contraction relationship. The deconstructed node trajectories are constructed into three-component descriptors based on their displacement amplitude, response sensitivity, and link dependency in the control domain. By inputting the three-component descriptor into the multidimensional factor mapping model, the potential factor describing the nodal force field is obtained. Arrange the relative positions of the potential factors of all nodes in the control domain into an array structure to form a potential allocation factor matrix.

8. A communication network resource management method based on intelligent scheduling according to claim 7, characterized in that, The process of forming a hierarchical intelligent scheduling track is as follows: The interaction degree between any pair of nodes in the potential distribution factor matrix is ​​analyzed, and the transformation correlation diagram between the pairs of nodes is constructed. Perform timing transformation on the critical path in the folded correlation graph using a non-parallel folded transformation formula; The action amplitude is rearranged according to the node sequence after time-sequence transformation, and a hierarchical intelligent scheduling track is generated based on the rearrangement result.

9. A communication network resource management method based on intelligent scheduling according to claim 8, characterized in that, The process of generating a pan-domain cooperative potential field that characterizes the global reuse trend of communication resources is as follows: After detecting that the track has entered the stable transition window during the time-stable segment of the hierarchical intelligent scheduling track, hierarchical sampling is performed on the transition state control domain to extract the response modes of different levels into multi-layer feature fragments. Cross-layer connectivity processing is performed on multi-layer feature segments, incorporating time stability, link consistency, and node potential gradient into the calculation. In the unified feature space obtained after fusion, the main direction reflecting the overall load flow, response rhythm coordination and resource reuse trend is identified, and a pan-domain collaborative potential field is generated.

10. A communication network resource management system based on intelligent scheduling, applied to the method described in any one of claims 1-9, characterized in that, include: Time-varying perturbation module: Disperses carrier demodulation texture, edge channel congestion sequence and micro-burst session clusters into dynamic interactive cavity, and constructs time-varying perturbation skeleton map in each cavity; Frequency misalignment recursion module: Writes the cloud-edge response loop into the instantaneous buffer of the edge node and injects the edge micro-response signal back to the cloud to form a frequency misalignment recursion layer; Sequence reshaping module: Constructs self-tuned sequence factors by folding out disordered signals from the frequency-misaligned recursive layer, performs non-uniform reshaping of cloud edge loops, and generates transition state control domain; Dynamic sequencing module: Maps each edge node to a potential allocation factor matrix, and forms a hierarchical intelligent scheduling track through non-parallel folding transformation adjustment; Layered fusion module: Performs layered fusion processing on the transition state control domain to generate a generalized cooperative potential field.

Images