A smart scheduling method and system for port tugboat systems

Abstract

This invention discloses an intelligent scheduling method and system for port tugboat systems, belonging to the field of intelligent scheduling technology. By uniformly preprocessing and partitioning the multi-channel collected data, this invention can fully utilize the information advantages of different channels in complex port environments, avoiding scheduling decision biases caused by single data dimensions. Based on this, by combining the construction of anomaly projection vectors and bidirectional cumulative evidence matrices, it achieves timely capture and high-confidence modeling of abnormal operating states, effectively improving the accuracy and robustness of scheduling. Through joint scheduling and end-cloud collaborative feedback mechanisms, this invention forms a dynamic closed-loop intelligent scheduling system, which can significantly improve tugboat resource utilization and port operation efficiency, and reduce scheduling delays and anomaly risks.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology, and in particular to an intelligent scheduling method and system for port tugboat systems. Background Technology

[0002] With the continuous growth of global port traffic, the complexity and real-time requirements of port operations have significantly increased, especially in tugboat scheduling. Improving tugboat resource utilization efficiency while ensuring safety has become a crucial issue for intelligent port management. In recent years, with the rapid development of IoT, big data, and AI technologies, tugboat operation scheduling has gradually shifted from traditional manual experience-based decision-making to intelligent, data-driven methods. Existing research largely relies on genetic algorithms, neural networks, and reinforcement learning to optimize the scheduling process and improve scheduling rationality and operational efficiency. However, port operation scenarios involve multi-channel data collection (such as ship arrival times, weather information, tugboat status, and operational area constraints). This data is characterized by multi-source heterogeneity, high real-time requirements, and frequent dynamic changes. Existing methods still have limitations in information fusion, anomaly detection, and joint scheduling. Achieving partitioned, dynamic, and robust scheduling optimization in multi-channel data scenarios remains a core challenge that urgently needs to be overcome.

[0003] CN115471142B discloses an intelligent scheduling method for port tugboat operations based on human-machine collaboration. It generates a theoretically optimal scheduling scheme through a genetic algorithm and classifies the dispatcher's habitual characteristics using a neural network autoencoder model. Finally, it obtains a fused optimal scheduling scheme through a fusion model and reinforcement learning algorithm. This method can balance theoretical optimality and human experience to a certain extent, but it suffers from limitations due to its reliance on modeling dispatcher habits. The algorithm's real-time performance and flexibility are poor, making it difficult to handle scenarios with sudden changes in multi-source dynamic data. Furthermore, this method focuses more on optimizing individual schemes and fusing habits, lacking mechanisms for joint encoding of multi-channel heterogeneous data, partition anomaly identification, and dynamic evidence updating. Therefore, in complex port environments, the scheduling scheme is prone to becoming disconnected from real-time operational needs.

[0004] CN120525281A discloses a method, system, device, and storage medium for intelligent management of ships at the shore. It constructs a port operation knowledge graph using electronic charts, ship status, and environmental data, and optimizes operation scheduling based on reinforcement learning and simulated annealing algorithms. This method enables autonomous decision-making and plans tugboat operation demands at the predictive level. However, the scheduling process relies too heavily on static modeling of the knowledge graph, making it difficult to effectively capture and respond to dynamic anomalies in multi-source data. Furthermore, at the joint scheduling level, the method still focuses on overall optimization, lacking data processing and scheduling strategies based on partitioning and hierarchy. This leads to response delays and local bottlenecks in complex and changing operational environments, hindering the real-time scheduling and feedback loop construction of the port tugboat system. Summary of the Invention

[0005] In view of the problems existing in the current port tugboat scheduling technology, this invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to achieve dynamic, multi-dimensional scheduling instruction optimization and feedback mechanism.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides an intelligent scheduling method for a port tugboat system, comprising: collecting and uniformly preprocessing a multi-channel acquisition dataset; generating a partition set according to a preset port area division rule; and performing multi-channel joint coding on the multi-channel acquisition dataset according to each partition based on a channel priority mapping table to obtain a partitioned coded dataset; calculating a preliminary anomaly projection vector for each partition of the partitioned coded dataset, and performing differential projection with the corresponding historical partitioned coding baseline to form an anomaly projection vector set, while constructing a bidirectional cumulative evidence matrix in the partition dimension; generating a back-injection codeword and a partition-level scheduling constraint set for each partition according to a mapping rule based on the anomaly projection vector set and the bidirectional cumulative evidence matrix; performing joint scheduling with the back-injection codeword and partition-level scheduling constraint set corresponding to each partition in the partition set as input, generating an updated partition-level scheduling instruction set, and sending it to the central and edge terminals and collecting execution feedback.

[0009] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the step of performing multi-channel joint coding to obtain a partitioned coding dataset includes: mapping each channel signal to coding bits according to the multi-channel acquired data in the partition and the channel priority mapping table to form a preliminary joint coding structure; within each partition, interleaving the coding bits of different channels according to the temporal and spatial correspondence to generate a cross-coding matrix; adding redundant check bits to the cross-coding matrix to generate the final partitioned coding data and adding it to the partitioned coding dataset.

[0010] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the generation of the cross-coding matrix includes: dividing each partition into several grids according to predetermined rules, and mapping each collection point within the partition to the corresponding grid according to the position of the ship or tugboat; recording the coding bits of each channel on the corresponding grid within each grid; aligning the collected data of each channel within each grid according to time slices, with data at the same time point corresponding to the same time index; writing the time-aligned grid channel coding bits into the cross-coding matrix in the manner of time index as rows and grid-channel combination as columns, forming a matrix structure of: time × (grid × channel); for any missing data at any time-grid-channel position, the corresponding channel coding bits in the time neighborhood are preferentially used for filling; if there is no valid data in the time neighborhood, the corresponding channel coding bits in the spatial neighborhood are used for filling.

[0011] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the calculation of preliminary anomaly projection vectors for each partition of the partitioned coding dataset includes: for each partition, extracting the corresponding cross-coding matrix from the partitioned coding dataset; for each time-space-channel combination of the cross-coding matrix, calculating the difference between the current coded value and the corresponding historical baseline, taking the absolute value of the difference as the deviation amplitude, and taking the sign of the difference as the deviation direction. The current abnormal projection vector is formed by combining these vectors and attaching a confidence label. The current abnormal projection vectors of all time-space-channel combinations within the same partition are collected to form a preliminary abnormal projection vector set. The formation of this abnormal projection vector set includes: comparing and calculating the corresponding elements of the preliminary abnormal projection vector and historical baseline for each partition one by one to obtain multidimensional difference data for each time-space-channel combination; the multidimensional difference amplitude... The Euclidean norm for multidimensional variance data; combined with the direction of deviation. Forming difference projection vectors :

[0012]

[0013] in, As the directional weighting factor, ; All The vectors are combined into a set of anomalous projection vectors.

[0014] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the step of constructing a bidirectional cumulative evidence matrix in the partition dimension includes: initializing a forward cumulative evidence matrix and a backward cumulative evidence matrix for each partition; forward accumulation along the time series: for each time-grid-channel combination, the result of multiplying the deviation amplitude by the corresponding confidence level is added to the corresponding element of the forward cumulative evidence matrix, and the forward accumulation is completed in ascending order of time; backward accumulation along the time series: for each time-grid-channel combination, the result of multiplying the deviation amplitude by the corresponding confidence level is added to the corresponding element of the backward cumulative evidence matrix, and the backward accumulation is completed in descending order of time; and the bidirectional cumulative evidence matrix is ​​obtained by integrating the forward cumulative evidence matrix and the backward cumulative evidence matrix by an element-wise arithmetic average.

[0015] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the confidence level is obtained by multiplying the historical fluctuation reliability score and the anomaly persistence score; the historical fluctuation reliability score is calculated by taking the ratio of the current deviation to the historical maximum deviation based on the deviation between the time-space-channel combination and the historical baseline; the anomaly persistence score is calculated by statistically analyzing the time length during which the initial anomaly projection vector deviation direction is consistent in consecutive time points and then comparing it with the length of a set reference time window.

[0016] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the step of generating back-injection codewords and partition-level scheduling constraint sets according to mapping rules includes: establishing three types of bit fields for each partition: amplitude field, direction field, and confidence field; the amplitude field is discretely quantized to obtain an amplitude bit string, the direction field is a binary bit representing increase or decrease, and the confidence field is mapped through intervals to obtain a confidence bit string; at the same time point and the same grid level, the three types of bit fields of all channels at the time-grid position are concatenated bit by bit and parity is checked bit by bit to obtain the parity check bit of the time-grid unit and appended to the end of the corresponding back-injection codeword; at the partition level, bit strings are extracted from the high-priority channel set and the low-priority channel set according to the channel priority mapping table, and parity is checked bit by bit within each priority channel set, and the check results are merged to form a partition-level check bit and appended to the end of the back-injection codeword; the amplitude field corresponds to the scheduling intensity level, the direction field corresponds to the scheduling trend, and the confidence field corresponds to the constraint priority; each partition generates a corresponding partition-level scheduling constraint set.

[0017] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the joint scheduling includes: aligning the back-injection codewords generated by each partition according to the time index; extracting the amplitude domain, direction domain, and confidence domain of each partition within each time slice; calculating the confidence-weighted average of the amplitude as the initial scheduling intensity of the time slice; using confidence-weighted majority voting for direction decision: calculating the positive weighted sum and the negative weighted sum; if the positive weighted sum is greater than the negative weighted sum, the final direction is to increase; if it is less, the final direction is to decrease; if they are equal, the comprehensive intensity is split according to the proportion of the two to the total weight and allocated to the positive and negative directions respectively; combining the final direction and the comprehensive intensity to form a partition-level joint scheduling instruction, and performing hierarchical allocation between channels according to the channel priority mapping table to generate a partition-level joint scheduling scheme issued according to priority.

[0018] As a preferred embodiment of the intelligent scheduling method for port tugboat systems described in this invention, the step of sending instructions to the central and edge terminals and collecting execution feedback includes: packaging the partition-level joint scheduling instruction set into instruction frames according to channel priority; the central node sends the instruction frames to the corresponding edge terminals, and the edge terminals directly execute tugboat scheduling operations according to the instruction frames; each edge terminal generates feedback information after execution, the feedback information including the actual number of tugboats executed, the execution direction, and the execution timestamp; the central node receives the feedback information from each edge terminal and compares it with the partition-level joint scheduling instruction set one by one, calculating the absolute value of the difference between the actual execution value and the corresponding scheduling instruction value, as well as the consistency between the scheduling direction and the execution feedback direction; if the absolute value of the difference exceeds a preset difference threshold or the direction is inconsistent, the corresponding partition is marked as a scheduling anomaly, and the historical baseline data is updated to provide a basis for correction for subsequent scheduling optimization.

[0019] Secondly, the present invention provides an intelligent scheduling system for port tugboat systems, comprising: a partition coding module, used to collect and uniformly preprocess multi-channel collected datasets, generate a partition set according to a preset port area division rule, and perform multi-channel joint coding on the multi-channel collected datasets according to each partition based on a channel priority mapping table to obtain a partitioned coded dataset;

[0020] An anomaly projection module is used to calculate a preliminary anomaly projection vector for each partition of the partitioned coding dataset, and perform differential projection with the corresponding historical partition coding baseline to form an anomaly projection vector set. At the same time, a bidirectional cumulative evidence matrix is ​​constructed in the partition dimension.

[0021] The back-injection generation module is used to generate back-injection codewords and partition-level scheduling constraint sets in each partition according to mapping rules based on the set of anomaly projection vectors and the bidirectional cumulative evidence matrix.

[0022] The scheduling instruction module is used to perform joint scheduling with the back-injection codeword corresponding to each partition in the partition set and the partition-level scheduling constraint set as input, generate an updated partition-level scheduling instruction set, send it to the center and edge terminals, and collect execution feedback.

[0023] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the intelligent scheduling method for a port tugboat system as described in the first aspect of the present invention.

[0024] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the intelligent scheduling method for a port tugboat system as described in the first aspect of the present invention.

[0025] The beneficial effects of this invention are as follows: By performing unified preprocessing and partition coding on multi-channel acquired data, this invention can make full use of the information advantages of different channels in a complex port environment, avoiding scheduling decision bias caused by a single data dimension; on this basis, by combining the construction of anomaly projection vectors and bidirectional cumulative evidence matrices, timely capture and high-confidence modeling of abnormal operating states are realized, effectively improving the accuracy and robustness of scheduling.

[0026] Meanwhile, this invention establishes a collaborative constraint mechanism between the inside and outside of the partition by generating back-injection codewords and partition-level scheduling constraint sets, so that the scheduling scheme not only has global optimization capabilities, but also partition adaptive adjustment capabilities; through joint scheduling and end-cloud collaborative feedback mechanism, a dynamic closed-loop intelligent scheduling system is formed, which can significantly improve the utilization rate of tugboat resources and port operation efficiency, and reduce scheduling delays and anomaly risks. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0028] Figure 1 This is a flowchart of an intelligent scheduling method for port tugboat systems.

[0029] Figure 2 This is a structural diagram of an intelligent scheduling system for port tugboat systems. Detailed Implementation

[0030] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0031] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0032] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0033] Figure 1 This is a flowchart of an intelligent scheduling method for a port tugboat system according to an embodiment of the present invention. Figure 1 As shown, the intelligent scheduling method for port tugboat systems includes:

[0034] S1: Collect and uniformly preprocess the multi-channel acquisition dataset, generate a set of partitions according to the preset port area division rules, and perform multi-channel joint encoding on the multi-channel acquisition dataset according to each partition based on the channel priority mapping table to obtain the partitioned encoded dataset.

[0035] S1.1: Based on the multi-channel acquisition data in the partition, and according to the channel priority mapping table, map each channel signal to the encoding bits, with higher priority channels allocated more redundant check bits to form a preliminary joint encoding structure.

[0036] For example, for high-priority channels, the allocated bit width is determined by the product of the mapping scaling factor and the priority. A larger bit width results in more redundancy check bits, indicating stronger fault tolerance for the channel during joint encoding. This ensures that high-priority channels can still provide sufficient recovery information in the event of packet loss, data distortion, or interference in subsequent communication, while low-priority channels maintain a minimum information set with less bit width, avoiding excessive bandwidth consumption. The rules for determining whether a channel is high-priority or low-priority are pre-defined.

[0037] S1.2: Within each partition, the encoded bits of different channels are cross-coded according to the temporal and spatial correspondence to generate a cross-coding matrix, so as to enhance the interdependence between multiple channels and avoid overall failure when a single channel data is missing.

[0038] The generation of the cross-coding matrix includes the following steps:

[0039] Each zone is divided into several grids according to predetermined rules, and each collection point in the zone is mapped to the corresponding grid according to the location of the ship or tugboat.

[0040] Within each grid, the encoded bits of each channel on the corresponding grid are recorded.

[0041] The acquired data from each channel within each grid is aligned by time slice, with data from the same time point corresponding to the same time index. In this way, the data at each moment is aligned in the matrix rows, and each column is uniquely identified by a "grid-channel" combination.

[0042] The time-aligned grid channel encoding bits are written into the cross-coding matrix using the time index as the row and the grid-channel combination as the column. The resulting matrix structure is: time × (grid × channel).

[0043] Furthermore, for any missing data at any time-grid-channel location, the missing data is first filled using the corresponding channel encoding bits of the time neighborhood (adjacent time points); if no valid data exists in the time neighborhood, the missing data is filled using the corresponding channel encoding bits of the spatial neighborhood (adjacent grids). This filling method ensures that the matrix has a dense structure at any time, thus supporting subsequent projection calculations and redundancy checks.

[0044] S1.3: Add redundant check bits to the cross-coding matrix to generate the final partitioned coding data, and add it to the partitioned coding dataset.

[0045] The calculation rules for redundancy check bits are as follows: Perform a bitwise XOR operation on all coded bits in each row (i.e., the same time slice) of the matrix to generate row parity bits; perform a bitwise XOR operation on all time slice coded bits in each column (i.e., the same grid-channel combination) to generate column parity bits; then perform a global XOR operation on all row and column parity bits to generate global parity bits. The calculated row, column, and global parity bits are then appended sequentially to the end of the cross-coding matrix to form the final partitioned coded data.

[0046] Redundancy check bits provide a self-checking mechanism within a partition, which can maintain data integrity in the event of packet loss or local anomalies, improving data robustness and reliability compared to existing simple encoding methods.

[0047] It should be noted that the multi-channel joint encoding of the present invention adopts a hierarchical encoding structure of cross-index and redundant check bits. The mapping table injects channel reliability information during encoding, thereby retaining the minimum information set that can be used for subsequent correction in the case of single-channel distortion or communication packet loss.

[0048] S2: Calculate the preliminary anomaly projection vector for each partition of the partitioned coding dataset, and perform differential projection with the corresponding historical partition coding baseline to form an anomaly projection vector set. At the same time, construct a bidirectional cumulative evidence matrix in the partition dimension.

[0049] S2.1: Calculate the preliminary anomaly projection vector.

[0050] First, for each partition, extract the corresponding cross-coding matrix from the partition coding dataset.

[0051] For each time-space-channel combination of the cross-coding matrix, the difference between the current encoded value and the corresponding historical baseline is calculated. The absolute value of the difference is taken as the deviation magnitude, and the sign of the difference is taken as the deviation direction. The deviation direction is +1 or -1, which respectively represent the trend of increasing or decreasing. By combining the deviation magnitude and deviation direction, the current abnormal projection vector is formed and a confidence label is attached.

[0052] The confidence level is obtained by multiplying the historical volatility reliability score and the anomaly persistence score.

[0053] Specifically, the historical fluctuation reliability score is calculated by taking the ratio of the current deviation to the historical maximum deviation based on the deviation of the time-space-channel combination from the historical baseline; the anomaly persistence score is calculated by statistically analyzing the time length during which the initial anomaly projection vector deviation direction is consistent across consecutive time points and then comparing it with the set reference time window length.

[0054] Collect the current abnormal projection vectors of all time-space-channel combinations in the same partition to form a preliminary abnormal projection vector set.

[0055] S2.2: Form a set of anomalous projection vectors.

[0056] The initial anomaly projection vector of each partition is compared and calculated one by one with the corresponding elements of the historical baseline (to obtain the deviation magnitude respectively), resulting in multidimensional difference data for each time-space-channel combination.

[0057] Among them, multidimensional difference amplitude The Euclidean norm for multidimensional variance data, combined with the direction of bias. Forming difference projection vectors :

[0058]

[0059] in, As the directional weighting factor, ; All The vectors are combined into a set of anomalous projection vectors.

[0060] S2.3: Construct a two-way cumulative evidence matrix in the partition dimension.

[0061] First, initialize the forward cumulative evidence matrix and the backward cumulative evidence matrix for each partition.

[0062] Forward accumulation along the time series: For each time-grid-channel combination, the result of multiplying the deviation magnitude by the corresponding confidence level is added to the corresponding element of the forward accumulation evidence matrix, and the forward accumulation is completed in ascending order of time.

[0063] Backward accumulation along the time series: For each time-grid-channel combination, the result of multiplying the deviation magnitude by the corresponding confidence level is accumulated in decreasing order over time to the corresponding element of the backward accumulation evidence matrix, completing the backward accumulation. This process is similar to forward accumulation, only in the opposite direction, i.e., the deviation values ​​are accumulated in the direction of decreasing time. The purpose of forward and backward accumulation is to distinguish between short-term abrupt changes and long-term drift. Forward accumulation mainly captures short-term abrupt changes, while backward accumulation is more inclined to identify long-term deviations.

[0064] The forward cumulative evidence matrix and the backward cumulative evidence matrix are integrated by element-wise arithmetic average to obtain the bidirectional cumulative evidence matrix. That is, the final evidence value of each time-grid-channel position is equal to (forward cumulative value + backward cumulative value) / 2.

[0065] It should be noted that through this process, a partition-level bidirectional cumulative evidence matrix is ​​constructed, which will effectively improve the stability and traceability of anomaly detection, ensure that small and continuous deviations can trigger corrections, and smooth out transient noise.

[0066] S3: Based on the set of abnormal projection vectors and the bidirectional cumulative evidence matrix, generate back-injection codewords and partition-level scheduling constraint sets in each partition according to the mapping rules.

[0067] S3.1: Establish three types of bit fields for each partition: amplitude field, direction field, and confidence field. Among them, the amplitude field is used to store the deviation amplitude, the direction field is used to store the deviation direction, and the confidence field is used to store the confidence value.

[0068] S3.2: The amplitude domain is obtained by discrete quantization to obtain the amplitude bit string, the direction domain is a binary bit representing increase / decrease, and the confidence domain is obtained by interval mapping to obtain the confidence bit string.

[0069] It should be noted that after establishing the three types of bit fields, the amplitude field and confidence field of continuous values ​​need to be quantized and mapped to discrete bit strings.

[0070] Specifically, the quantization operation in the amplitude domain is defined as follows: the amplitude interval is evenly divided into multiple amplitude levels to obtain an amplitude bit string, where each level corresponds to a fixed-length bit code; the direction domain, as a binary variable, is directly mapped to a single bit, with +1 corresponding to logic "1" and -1 corresponding to logic "0", thus obtaining the direction bit; the mapping method for the confidence domain is as follows: the interval is divided into multiple sub-intervals according to the segmented interval, with each sub-interval corresponding to a confidence level, ultimately obtaining a confidence bit string. For example, when the confidence value is within the interval [0.75, 1], it corresponds to the highest level bit sequence. In this way, all time-grid-channel combinations are converted into a bit set consisting of amplitude bit strings, direction bits, and confidence bit strings, forming a standardized basic codeword unit.

[0071] S3.3: At the same time point and the same grid level, the three types of bit fields of all channels at the time-grid position are concatenated bit by bit and parity check is performed bit by bit (by bit XOR) to obtain the parity check bit of the time-grid unit and append it to the end of the corresponding back-injection codeword.

[0072] The specific method is as follows: For each bit of the composite bit string, perform a bitwise XOR operation on the corresponding bit value of all channels under the time-grid unit to obtain the parity bit at the corresponding position. In this way, the parity bit obtained is appended to the end of the back-injection codeword of the corresponding time-grid unit.

[0073] The introduction of parity checking provides the ability to detect local errors. Even when some channel signals are missing or distorted, the parity bit can still be used to locate and correct errors, thereby improving the robustness and reliability of scheduled data transmission.

[0074] S3.4: At the partition level, extract bit strings from the high-priority channel set and the low-priority channel set according to the channel priority mapping table, and perform parity checks bit by bit within each priority channel set. Combine the check results to form a partition-level check bit and append it to the end of the back-injection codeword.

[0075] For example, firstly, based on the channel priority mapping table, all channels within a partition are divided into a high-priority channel set and a low-priority channel set; then, a bit-by-bit parity check operation is performed on the channel bit strings within each set, that is, a bit-by-bit XOR operation is performed on the amplitude bit strings, direction bit strings, and confidence bit strings of all channels within the high-priority channel set to obtain the high-priority parity bit; similarly, the same operation is performed on the low-priority channel set to obtain the low-priority parity bit; finally, the high-priority and low-priority parity bits are merged and appended to the end of the partition back-injection codeword to form a complete partition-level back-injection codeword structure.

[0076] The above process ensures that channels of different priorities are treated differently in data recovery and scheduling decisions, so that high-priority channels can still obtain higher protection and recovery capabilities in communication-restricted or abnormal situations.

[0077] S3.5: The amplitude domain corresponds to the scheduling intensity level, the direction domain corresponds to the scheduling trend (increase / decrease), and the confidence domain corresponds to the constraint priority; each partition generates a corresponding partition-level scheduling constraint set.

[0078] After generating the partition-level back-injection codewords, it is necessary to further map the three types of bit fields to specific scheduling constraint indicators and construct a partition-level scheduling constraint set.

[0079] The specific mapping rules are as follows: the amplitude bit string corresponding to the amplitude field is mapped to the scheduling intensity level, which represents the execution intensity requirement of the corresponding partition at the current time point; the direction bit corresponding to the direction field is mapped to the scheduling trend, which indicates whether the scheduling instruction of the corresponding partition increases or decreases tug resources; the confidence bit string corresponding to the confidence field is mapped to the constraint priority, which determines the priority order of the corresponding partition constraints in cross-partition integrated scheduling. Through the above mapping rules, the final partition-level scheduling constraint set includes three dimensions: intensity level, scheduling trend, and constraint priority, which can provide complete constraints for cross-partition consistent scheduling.

[0080] As can be seen, this invention transforms complex anomaly projection vectors and bidirectional cumulative evidence matrices into structured back-injection codewords and scheduling constraint sets, thereby achieving the encoding, verification, and constraint generation of anomaly information at the data level. Specifically, the time-grid cell-level parity check mechanism ensures rapid detection and repair of local errors, while the partition-level priority check bit design guarantees the weighted advantage of critical channels in scheduling decisions. The final mapped partition-level scheduling constraint set not only describes the intensity requirements and trend direction of partitions but also quantifies the priority of constraints, enabling cross-partition scheduling to achieve efficient, consistent, and reliable resource allocation when facing complex anomalies and conflicts.

[0081] S4: Perform joint scheduling with the back-injection codeword and partition-level scheduling constraint set corresponding to each partition in the partition set as input, generate an updated partition-level scheduling instruction set, send it to the center and edge terminals and collect execution feedback.

[0082] A. Align the back-injection codewords generated by each partition according to the time index, and extract the amplitude domain, direction domain and confidence domain of each partition within each time slice.

[0083] B. Calculate the confidence-weighted average of the amplitude as the initial scheduling intensity of the time slice.

[0084] C. Directional decision-making adopts confidence-weighted majority voting: calculate the positive weighted sum and the negative weighted sum. If the positive weighted sum is greater than the negative weighted sum, the final direction is to increase; if it is less, the final direction is to decrease; if they are equal, the overall strength is split according to the proportion of the two to the total weight and distributed to the positive and negative directions respectively.

[0085] Specifically, the positive weighted sum is calculated as follows: multiply the deviation magnitude of all projection vectors with positive deviation directions by their corresponding confidence values ​​to obtain the weighted result. Summing all weighted results yields the positive weighted sum. The positive weighted sum represents the cumulative influence intensity of all increasing directions, obtained after confidence weighting. The negative weighted sum is calculated similarly.

[0086] D. Combine the final direction and overall strength to form a partition-level joint scheduling instruction, and allocate it hierarchically among channels according to the channel priority mapping table to generate a partition-level joint scheduling scheme issued according to priority.

[0087] E. The partition-level joint scheduling instruction set is packaged into instruction frames according to channel priority. Each instruction frame contains a partition index, scheduling strength, scheduling direction, and confidence weight. The central node sends the instruction frames to the corresponding edge terminals, which then directly execute the tugboat scheduling operation based on the instruction frames.

[0088] F. Each edge terminal generates feedback information after execution, including the actual number of tugboats executed, the execution direction, and the execution timestamp.

[0089] G. The central node receives feedback information from each edge terminal and compares it with the partition-level joint scheduling instruction set one by one. It calculates the absolute value of the difference between the actual execution value and the corresponding scheduling instruction value, as well as the consistency between the scheduling direction and the execution feedback direction. If the absolute value of the difference exceeds the preset difference threshold or the direction is inconsistent, the corresponding partition is marked as a scheduling anomaly, and the historical baseline data is updated to provide a basis for correction for subsequent scheduling optimization.

[0090] Furthermore, such as Figure 2 As shown, this embodiment also provides an intelligent scheduling system for a port tugboat system, including,

[0091] The partition coding module is used to collect and uniformly preprocess multi-channel acquisition datasets, generate a set of partitions according to the preset port area division rules, and perform multi-channel joint coding on the multi-channel acquisition datasets according to each partition based on the channel priority mapping table to obtain the partitioned coding dataset.

[0092] An anomaly projection module is used to calculate a preliminary anomaly projection vector for each partition of the partitioned coding dataset, and perform differential projection with the corresponding historical partition coding baseline to form an anomaly projection vector set. At the same time, a bidirectional cumulative evidence matrix is ​​constructed in the partition dimension.

[0093] The back-injection generation module is used to generate back-injection codewords and partition-level scheduling constraint sets in each partition according to mapping rules based on the set of anomaly projection vectors and the bidirectional cumulative evidence matrix.

[0094] The scheduling instruction module is used to perform joint scheduling with the back-injection codeword corresponding to each partition in the partition set and the partition-level scheduling constraint set as input, generate an updated partition-level scheduling instruction set, send it to the center and edge terminals, and collect execution feedback.

[0095] This embodiment also provides a computer device applicable to the intelligent scheduling method for a port tugboat system, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent scheduling method for a port tugboat system as proposed in the above embodiment.

[0096] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0097] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the intelligent scheduling method for a port tugboat system as proposed in the above embodiments.

[0098] In summary, by performing unified preprocessing and partitioned coding on multi-channel acquired data, this invention can fully utilize the information advantages of different channels in complex port environments, avoiding scheduling decision biases caused by a single data dimension. On this basis, by combining the construction of anomaly projection vectors and bidirectional cumulative evidence matrices, timely capture and high-confidence modeling of abnormal operating states are achieved, effectively improving the accuracy and robustness of scheduling.

[0099] Meanwhile, this invention establishes a collaborative constraint mechanism between the inside and outside of the partition by generating back-injection codewords and partition-level scheduling constraint sets, so that the scheduling scheme not only has global optimization capabilities, but also partition adaptive adjustment capabilities; through joint scheduling and end-cloud collaborative feedback mechanism, a dynamic closed-loop intelligent scheduling system is formed, which can significantly improve the utilization rate of tugboat resources and port operation efficiency, and reduce scheduling delays and anomaly risks.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent scheduling method for a port tugboat system, characterized in that: include: Collect and preprocess multi-channel acquisition datasets in a unified manner, generate a set of partitions according to the preset port area division rules, and perform multi-channel joint encoding on the multi-channel acquisition datasets according to each partition based on the channel priority mapping table to obtain the partitioned encoded dataset; For the partitioned coding dataset, a preliminary anomaly projection vector is calculated in each partition, and a difference projection is performed with the corresponding historical partition coding baseline to form an anomaly projection vector set. At the same time, a two-way cumulative evidence matrix is ​​constructed in the partition dimension. Based on the set of abnormal projection vectors and the bidirectional cumulative evidence matrix, back-injection codewords and partition-level scheduling constraint sets are generated in each partition according to the mapping rules. The joint scheduling is performed using the back-injection codeword and partition-level scheduling constraint set corresponding to each partition in the partition set as input, generating an updated partition-level scheduling instruction set, which is then sent to the central and edge terminals and execution feedback is collected. Calculating the preliminary anomaly projection vector for each partition of the partitioned coding dataset includes: for each partition, extracting the corresponding cross-coding matrix from the partitioned coding dataset; for each time-space-channel combination of the cross-coding matrix, calculating the difference between the current encoded value and the corresponding historical baseline, taking the absolute value of the difference as the deviation magnitude, and taking the sign of the difference as the deviation direction. The current abnormal projection vector is formed by combining these vectors and attaching a confidence label. The current abnormal projection vectors of all time-space-channel combinations within the same partition are collected to form a preliminary abnormal projection vector set. The formation of this abnormal projection vector set includes: comparing and calculating the corresponding elements of the preliminary abnormal projection vector and historical baseline for each partition one by one to obtain multidimensional difference data for each time-space-channel combination; the multidimensional difference amplitude... The Euclidean norm for multidimensional variance data; combined with the direction of deviation. Forming difference projection vectors : ； in, As the directional weighting factor, ; All The combination forms a set of anomalous projection vectors; The construction of the bidirectional cumulative evidence matrix at the partition dimension includes: initializing a forward cumulative evidence matrix and a backward cumulative evidence matrix for each partition; forward accumulation along the time series: for each time-grid-channel combination, the result of multiplying the deviation magnitude by the corresponding confidence level is added to the corresponding element of the forward cumulative evidence matrix, and the forward accumulation is completed in ascending order over time; backward accumulation along the time series: for each time-grid-channel combination, the result of multiplying the deviation magnitude by the corresponding confidence level is added to the corresponding element of the backward cumulative evidence matrix, and the backward accumulation is completed in descending order over time; and integrating the forward cumulative evidence matrix and the backward cumulative evidence matrix by element-wise arithmetic average to obtain the bidirectional cumulative evidence matrix. The confidence level is obtained by multiplying the historical fluctuation reliability score and the anomaly persistence score. The historical fluctuation reliability score is calculated by taking the ratio of the current deviation to the historical maximum deviation based on the deviation of the time-space-channel combination from the historical baseline. The anomaly persistence score is calculated by statistically analyzing the time length during which the initial anomaly projection vector deviation direction is consistent in consecutive time points and then comparing it with the set reference time window length.

2. The intelligent scheduling method for port tugboat systems as described in claim 1, characterized in that: The multi-channel joint encoding process yields a partitioned encoded dataset including: Based on the multi-channel acquired data in the partition, and according to the channel priority mapping table, the signals of each channel are mapped to the coding bits to form a preliminary joint coding structure; Within each partition, the encoded bits of different channels are interleaved according to temporal and spatial correspondences to generate an interleaved encoding matrix; Redundant check bits are added to the cross-coding matrix to generate the final partitioned coding data, which is then added to the partitioned coding dataset.

3. The intelligent scheduling method for port tugboat systems as described in claim 2, characterized in that: The generation of the cross-coding matrix includes: Each zone is divided into several grids according to predetermined rules, and each collection point in the zone is mapped to the corresponding grid according to the location of the ship or tugboat. Within each grid, the encoded bits of each channel in the corresponding grid are recorded by channel; The collected data from each channel within each grid are aligned according to time slices, and data from the same time point corresponds to the same time index; The time-aligned grid channel encoding bits are written into the cross-coding matrix using the time index as the row and the grid-channel combination as the column. The resulting matrix structure is: time × (grid × channel). For any missing data at any time-grid-channel location, the corresponding channel code bit in the time neighborhood is used for filling first; if there is no valid data in the time neighborhood, the corresponding channel code bit in the spatial neighborhood is used for filling.

4. The intelligent scheduling method for port tugboat systems as described in claim 3, characterized in that: The generation of back-injection codewords and partition-level scheduling constraint sets according to mapping rules includes: Three types of bit fields are established for each partition: amplitude field, direction field, and confidence field; The amplitude domain is discretely quantized to obtain the amplitude bit string, the direction domain is a binary bit representing increase or decrease, and the confidence domain is mapped to obtain the confidence bit string. At the same time point and the same grid level, the three types of bit fields of all channels at the time-grid position are concatenated bit by bit and parity is checked bit by bit to obtain the parity check bit of the time-grid unit and appended to the end of the corresponding back-injection codeword. At the partition level, the bit strings of the high-priority channel set and the low-priority channel set are extracted according to the channel priority mapping table, and parity checks are performed bit by bit within each priority channel set. The check results are merged to form a partition-level check bit and appended to the end of the back-injection codeword. The amplitude domain corresponds to the scheduling intensity level, the direction domain corresponds to the scheduling trend, and the confidence domain corresponds to the constraint priority; each partition generates a corresponding partition-level scheduling constraint set.

5. The intelligent scheduling method for port tugboat systems as described in claim 4, characterized in that: The joint scheduling includes: Align the back-injection codewords generated by each partition according to the time index, and extract the amplitude domain, direction domain and confidence domain of each partition within each time slice; The confidence-weighted average of the amplitude is used as the initial scheduling intensity of the time slice; Directional decisions are made using a confidence-weighted majority vote: the positive weighted sum and the negative weighted sum are calculated. If the positive weighted sum is greater than the negative weighted sum, the final direction is to increase; if it is less, the final direction is to decrease; if they are equal, the overall strength is split according to the proportion of the two to the total weight and distributed to the positive and negative directions respectively. The final direction and overall strength are combined to form a partition-level joint scheduling instruction, and the instruction is distributed hierarchically among channels according to the channel priority mapping table to generate a partition-level joint scheduling scheme issued according to priority.

6. The intelligent scheduling method for port tugboat systems as described in claim 5, characterized in that: The process of sending data to central and edge terminals and collecting execution feedback includes: The partition-level joint scheduling instruction set is packaged into instruction frames according to channel priority; The central node sends the instruction frame to the corresponding edge terminal, and the edge terminal directly executes the tugboat scheduling operation according to the instruction frame; Each edge terminal generates feedback information after execution, including the actual number of tugboats executed, the execution direction, and the execution timestamp. The central node receives feedback information from each edge terminal and compares it with the partition-level joint scheduling instruction set one by one, calculating the absolute value of the difference between the actual execution value and the corresponding scheduling instruction value, as well as the consistency between the scheduling direction and the execution feedback direction. If the absolute value of the difference exceeds the preset difference threshold or the direction is inconsistent, the corresponding partition will be marked as a scheduling anomaly, and the historical baseline data will be updated to provide a basis for correction in subsequent scheduling optimization.

7. An intelligent scheduling system for a port tugboat system, based on the intelligent scheduling method for a port tugboat system according to any one of claims 1 to 6, characterized in that: Also includes: The partition coding module is used to collect and uniformly preprocess multi-channel acquisition datasets, generate a set of partitions according to the preset port area division rules, and perform multi-channel joint coding on the multi-channel acquisition datasets according to each partition based on the channel priority mapping table to obtain the partitioned coding dataset. An anomaly projection module is used to calculate a preliminary anomaly projection vector for each partition of the partitioned coding dataset, and perform differential projection with the corresponding historical partition coding baseline to form an anomaly projection vector set. At the same time, a bidirectional cumulative evidence matrix is ​​constructed in the partition dimension. The back-injection generation module is used to generate back-injection codewords and partition-level scheduling constraint sets in each partition according to mapping rules based on the set of anomaly projection vectors and the bidirectional cumulative evidence matrix. The scheduling instruction module is used to perform joint scheduling with the back-injection codeword corresponding to each partition in the partition set and the partition-level scheduling constraint set as input, generate an updated partition-level scheduling instruction set, send it to the center and edge terminals, and collect execution feedback.

Images