Logistics task intelligent distribution method and system oriented to multi-source demand

Abstract

The invention discloses a logistics task intelligent distribution method and system oriented to multi-source requirements, and relates to the technical field of task distribution. The method comprises the following steps: interacting M computing power boxes, carrying out logistics equipment business data acquisition, and obtaining M logistics equipment business work log sequences and M logistics equipment business work video sequences; carrying out abnormal feature anchor point vector extraction to obtain M log abnormal feature anchor point vector sequences and M video abnormal feature anchor point vector sequences; carrying out attention interaction fusion to obtain M log abnormal feature interaction anchor vector sequences; carrying out logistics task work order generation to obtain a logistics task work order set; and obtaining logistics resource distribution information of the target area, and performing intelligent allocation in combination with the logistics task work order set to obtain a target intelligent allocation scheme. The technical problems of unbalanced logistics task allocation and low resource utilization rate in the prior art are solved, efficient matching of tasks and resources is realized, and the technical effects of improving the logistics resource utilization rate are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of task allocation, and in particular to a method and system for intelligently allocating logistics tasks oriented to multi-source demands. Background Art

[0002] In the field of logistics management, with the increasing scale and complexity of facilities such as smart campuses, intelligent buildings, hospitals, schools, enterprises, and institutions, daily logistics services encompass multiple aspects, including equipment operation and maintenance, material management, fire safety, security inspections, and environmental protection. These services demonstrate significant demand diversity, task complexity, and resource dynamism. Existing methods for allocating logistics tasks often rely on manual management or pre-set fixed rules. These methods lack the ability to uniformly analyze and intelligently process multi-source logistics demand data, making it difficult to promptly identify the priority and urgency of different tasks. Furthermore, efficient and refined scheduling and rational allocation of logistics resources are difficult to achieve. Overall response efficiency and resource utilization levels need to be improved. Summary of the Invention

[0003] The present application provides a method and system for intelligently allocating logistics tasks for multi-source demands, which solves the technical problems of unbalanced logistics task allocation and low resource utilization in the prior art.

[0004] A first aspect of the present application provides a method for intelligently allocating logistics tasks for multi-source requirements, the method comprising: The M computing power boxes corresponding to the M logistics equipment interactively distributed in the target area collect logistics equipment business data to obtain M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer; the M logistics equipment business work log sequences and the M logistics equipment business work video sequences are traversed to extract abnormal feature anchor vectors to obtain M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences; based on the M video abnormal feature anchor vector sequences, the M log abnormal feature anchor vector sequences are subjected to attention interaction fusion to obtain M log abnormal feature interaction anchor vector sequences; logistics task work orders are generated according to the M log abnormal feature interaction anchor vector sequences to obtain a logistics task work order set; the logistics resource distribution information of the target area is obtained, and intelligent allocation is performed in combination with the logistics task work order set to obtain a target intelligent allocation plan.

[0005] A second aspect of the present application provides a multi-source demand-oriented intelligent logistics task allocation system, the system comprising: Data acquisition module: interacts with M computing power boxes corresponding to M logistics equipment distributed in the target area, collects logistics equipment business data, and obtains M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer; anchor vector extraction module: traverses the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtains M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences; fusion module: performs attention interaction fusion on the M log abnormal feature anchor vector sequences based on the M video abnormal feature anchor vector sequences, and obtains M log abnormal feature interaction anchor vector sequences; work order generation module: generates logistics task work orders based on the M log abnormal feature interaction anchor vector sequences, and obtains a logistics task work order set; allocation module: obtains logistics resource distribution information of the target area, and performs intelligent allocation in combination with the logistics task work order set to obtain a target intelligent allocation plan.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, the M computing power boxes corresponding to M logistics equipment distributed in the target area interact to collect logistics equipment business data, obtaining M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer. Next, the M logistics equipment business work log sequences and M logistics equipment business work video sequences are traversed to extract abnormal feature anchor vectors, obtaining M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences. Furthermore, the M log abnormal feature anchor vector sequences are subjected to attention interaction fusion based on the M video abnormal feature anchor vector sequences, obtaining M log abnormal feature interaction anchor vector sequences. Then, logistics task work orders are generated based on the M log abnormal feature interaction anchor vector sequences, obtaining a logistics task work order set. Finally, logistics resource distribution information for the target area is obtained and intelligently allocated based on the logistics task work order set, resulting in a target intelligent allocation solution. This solves the technical problems of uneven logistics task distribution and low resource utilization in the existing technology, achieving efficient matching of tasks and resources and improving logistics resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1A flowchart of a method for intelligently allocating logistics tasks to multiple sources of demand, provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an intelligent logistics task allocation system for multi-source demand provided in an embodiment of the present application.

[0009] Explanation of reference numerals: data acquisition module 11 , anchor vector extraction module 12 , fusion module 13 , work order generation module 14 , allocation module 15 . DETAILED DESCRIPTION

[0010] This application solves the technical problems of unbalanced distribution of logistics tasks and low resource utilization in the prior art by providing a method and system for intelligent distribution of logistics tasks oriented to multi-source demands.

[0011] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0013] Example 1, as Figure 1 As shown, the present application provides a method for intelligently allocating logistics tasks oriented to multi-source demands, wherein the method includes: The M computing power boxes corresponding to the M logistics equipment distributed in the target area interact to collect logistics equipment business data, and obtain M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer.

[0014] In this embodiment, a computing box corresponding to each type of logistics equipment (such as elevators, air conditioners, water pumps, firefighting facilities, and inspection robots) distributed within a target area (such as a smart park, hospital, or campus) is deployed to achieve distributed deployment and local data collection and processing. The computing box is a micro-computing terminal with edge computing capabilities that can exchange data with the corresponding logistics equipment at high speed and low latency, and is used to collect multimodal data from the logistics equipment in real time during business operations.

[0015] Specifically, the computing box connects to the corresponding logistics equipment via communication interfaces (such as RS-485, Ethernet, Wi-Fi, and 5G). It periodically or on-demand collects structured business log data (including status records, fault codes, operational events, and alarm information) generated by the equipment, as well as unstructured business video data generated by connected surveillance cameras. During the collection process, the computing box performs preliminary processing on the raw data through local caching, compression, and preprocessing (such as log timestamp synchronization and video keyframe extraction) before uploading it to the central server or edge collaboration node.

[0016] Based on the correspondence between M logistics equipment and M computing power boxes, M logistics equipment business work log sequences and M logistics equipment business work video sequences can be obtained respectively, where M is a positive integer representing the total number of logistics equipment involved in data collection in the target area.

[0017] The M logistics equipment business work log sequences and the M logistics equipment business work video sequences are traversed to extract abnormal feature anchor vectors to obtain M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences.

[0018] By traversing the M collected logistics equipment business work log sequences and M logistics equipment business work video sequences, abnormal feature information related to logistics tasks is extracted and anchored in the form of structured vectors to generate M log abnormal feature anchor point vector sequences and M video abnormal feature anchor point vector sequences.

[0019] Each log sequence for a logistics device is processed using a pre-defined text anomaly feature identifier. This identifier, based on a natural language processing model (such as BERT, BiLSTM-CRF, or Transformer), combines domain knowledge dictionaries with an anomaly template library to perform semantic parsing and feature tagging on key log fields (such as device status codes, operation event types, fault identifiers, and anomaly keywords). It then extracts feature anchors reflecting anomalies and converts them into corresponding log anomaly feature anchor vectors. By traversing M log sequences, a sequence of M log anomaly feature anchor vectors is ultimately obtained.

[0020] Similarly, each video sequence of logistics equipment is processed using a pre-defined video anomaly feature identifier. This identifier uses deep learning-based visual models (such as YOLO, 3D CNN, ViT, and I3D) to extract key frames or key segments corresponding to abnormal events (such as equipment shaking, liquid leaks, abnormal personnel operation, flames, and smoke) from video clips. These key frames or segments are then converted into video anomaly feature anchor vectors that demonstrate spatial and temporal feature expression. By traversing M video sequences, a sequence of M video anomaly feature anchor vectors is ultimately obtained.

[0021] Furthermore, traversing the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtaining M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences, including: Pre-construct a text abnormality feature identifier and a video abnormality feature identifier; use the text abnormality feature identifier to extract abnormality feature anchor vectors from the M logistics equipment business work log sequences to obtain the M log abnormality feature anchor vector sequences; use the video abnormality feature identifier to extract abnormality feature anchor vectors from the M logistics equipment business work video sequences to obtain M video abnormality feature anchor vector sequences.

[0022] Preferably, a text anomaly feature identifier and a video anomaly feature identifier are pre-built, wherein the text anomaly feature identifier is used to identify abnormal events in log data, and is constructed based on natural language processing technology. Specifically, it may include a word vector embedding layer (such as Word2Vec, BERT Embedding), a sequence modeling layer (such as Bi-LSTM, Transformer Encoder) and a classification discrimination layer. It can identify abnormal patterns such as fault codes, warning states, operational anomalies, timeout events, etc. in equipment logs according to predefined abnormal type labels or weak supervision rules, and map abnormal fragments into structured vector representations to obtain abnormal feature anchor vector sequences corresponding to each log sequence; the video anomaly feature identifier is used to identify video anomaly information during the working process of logistics equipment, and is constructed based on computer vision and time series modeling methods, and specifically includes a video frame extraction module, an image / video target detection model (such as YOLOv5, Faster R-CNN), action recognition models (such as I3D, C3D, and SlowFast), etc. The video anomaly feature identifier can locate key abnormal scenes in device operation, such as smoke, liquid spills, violent device shaking, and people approaching illegally. It also maps the detected abnormal key frames or video segments into high-dimensional feature vectors to form a video anomaly feature anchor vector sequence.

[0023] The text anomaly feature identifier is used to parse and extract features from M logistics equipment business work log sequences one by one, and M log anomaly feature anchor point vector sequences are obtained. The video anomaly feature identifier is used to process and detect anomalies from M logistics equipment business work video sequences section by section, and M video anomaly feature anchor point vector sequences are obtained.

[0024] Based on the M video abnormality feature anchor point vector sequences, the M log abnormality feature anchor point vector sequences are subjected to attention interaction fusion to obtain M log abnormality feature interaction anchor point vector sequences.

[0025] In an embodiment of the present application, an attention interaction fusion mechanism is used to process M video anomaly feature anchor vector sequences and M log anomaly feature anchor vector sequences, ultimately obtaining M log anomaly feature interaction anchor vector sequences. Specifically, the video anomaly feature anchor vector sequence corresponding to each logistics device is first time-synchronized and matched with the log anomaly feature anchor vector sequence. Through the timestamp alignment strategy, an association relationship is established between anchor pairs in similar time periods. Subsequently, an interactive calculation model based on the attention mechanism is constructed, using the log anchor vector as the query item and the video anchor vector as the key and value. By calculating the attention coefficient, the attention weight of each log anchor in the video feature space is obtained. Furthermore, the corresponding video anchor vector is weightedly fused using the weight coefficient, and the result is fused with the original log anchor vector to obtain a fusion vector containing multimodal anomaly feature semantics. Finally, for each logistics device, a fused log anomaly feature interaction anchor vector sequence is generated, providing an accurate multi-source anomaly expression basis for subsequent anomaly association analysis and logistics task work order generation.

[0026] Furthermore, the M log abnormality feature anchor point vector sequences are subjected to attention interaction fusion based on the M video abnormality feature anchor point vector sequences to obtain M log abnormality feature interaction anchor point vector sequences, including: The M video abnormality feature anchor point vector sequences and the M log abnormality feature anchor point vector sequences are simultaneously identified with anchor point vector mapping interaction coefficients to obtain M mapping interaction coefficient sequences; M attention interaction fusion matrices are constructed based on the M mapping interaction coefficient sequences; the M log abnormality feature interaction anchor point vector sequences are interactively fused using the M attention interaction fusion matrices to obtain M log abnormality feature interaction anchor point vector sequences.

[0027] Preferably, the similarity between the anchor vectors of each corresponding log anomaly feature anchor vector sequence and the video anomaly feature anchor vector sequence is first calculated. A similarity metric, such as cosine similarity, Euclidean distance, or a dot-product-based similarity metric is used to evaluate the similarity of anchor pairs within the same time window, obtaining M mapping interaction coefficient sequences. Next, based on each mapping interaction coefficient sequence, M corresponding attention interaction fusion matrices are constructed. During the specific construction process, each set of interaction coefficient sequences is normalized to ensure that all coefficient values ​​are within a uniform dimension, ensuring a reasonable weight distribution. The normalized interaction coefficients are then populated into the corresponding upper triangular matrix according to their chronological order in the original sequence, forming a temporally ordered fusion structure that preserves the feature evolution trend. Subsequently, the attention interaction fusion matrix is ​​used to perform a weighted fusion process on the log anomaly feature anchor vector sequence. This involves multiplying each log anchor vector by its corresponding weight factor in the video interaction matrix and performing a weighted sum. Combined with the original log anchor vector, a new fusion vector is generated through feature concatenation or residual fusion, ultimately obtaining a log anomaly feature interaction anchor vector sequence that contains video interaction semantics.

[0028] The logistics task work order is generated based on the M log abnormal feature interaction anchor point vector sequences to obtain the logistics task work order set.

[0029] Furthermore, a logistics task work order is generated based on the M log anomaly feature interaction anchor point vector sequences to obtain a logistics task work order set, including: Traversing the M log anomaly feature interaction anchor point vector sequences to perform anomaly association iterative analysis, and obtaining M target anomaly iterative memories; performing logistics task work order identification based on the M target anomaly iterative memories, and obtaining a logistics task work order set.

[0030] Specifically, the system traverses M sequences of interactive anchor vectors of log anomaly features, combining temporal information, spatial correlation information, and fused multimodal feature representations to perform iterative anomaly correlation analysis. Specifically, based on recurrent neural networks (RNNs) and graph attention networks (GATs), it constructs an anomaly evolution path for each log vector sequence. Through multiple rounds of iteration, it gradually extracts feature fragments highly correlated with the target anomaly event from the original anchor vectors and stores them in a dynamically updated anomaly memory module, thus forming M iterative memories of the target anomaly. These memory sequences effectively characterize the types of problems, occurrence trends, and historical co-occurrence relationships that may exist in current logistics equipment, providing feature support for anomaly determination.

[0031] Based on the iterative memory of the M target anomalies obtained, combined with historical work order label information or preset work order classification rules, a task recognition model (such as a task matcher based on Transformer or BERT) is used to determine the task type of the anomaly features, identify the corresponding logistics task work order content, such as equipment failure, environmental anomaly, safety hazard, material shortage, etc., and output structured work order information, including task name, task type, processing priority, associated equipment ID, anomaly description summary, timestamp and geographic location information, to form a logistics task work order.

[0032] Furthermore, the M log anomaly feature interaction anchor point vector sequences are traversed to perform anomaly correlation iterative analysis to obtain M target anomaly iterative memories, including: M first log anomaly feature interaction anchor point vectors in the M log anomaly feature interaction anchor point vector sequence are used to perform anomaly association iterative extraction on M second log anomaly feature interaction anchor point vectors to obtain M first anomaly iterative memories; based on the M first anomaly iterative memories, anomaly association iterative extraction is performed on the M log anomaly feature interaction anchor point vector sequence to obtain M second anomaly iterative memories; and so on, anomaly association iterative analysis is performed on the M log anomaly feature interaction anchor point vector sequence based on the M second anomaly iterative memories to obtain the M target anomaly iterative memories.

[0033] First, from each of the M sequences of log anomaly feature interaction anchor vectors, the first log anomaly feature interaction anchor vector of the current stage in each sequence is extracted and used as the initial reference vector set. For each reference vector, a similarity-based anomaly association strategy (e.g., cosine similarity, Euclidean distance, or a custom weighted association function) is used to match the subsequent second log anomaly feature interaction anchor vectors in the sequence one by one. Anomaly feature vectors that are related to the reference vector in terms of content structure or evolutionary trajectory are identified. The matching results are then combined to construct M first anomaly iterative memories. Next, based on these M first anomaly iterative memories, the corresponding sequence of log anomaly feature interaction anchor vectors is traversed again, and the association between vectors not yet involved in the previous memory construction is determined. This allows further extraction of anomaly anchor vectors with potential extension relationships or evolutionary characteristics with the current memory content, forming M second anomaly iterative memories. Similarly, in each subsequent iteration, the anomaly memory generated in the previous iteration is used as input to continue deep-level association mining and completion of feature vectors, gradually expanding the anomaly memory coverage and strengthening the clustering effect of abnormal events until the current iteration no longer discovers new highly correlated anomaly anchors or reaches the preset iteration threshold. Ultimately, M target anomaly iterative memories with high semantic relevance and contextual coherence are obtained, providing high-quality anomaly knowledge support for subsequent logistics task work order identification and intelligent allocation.

[0034] Further, including: The M first log anomaly feature interaction anchor vectors are used to judge the degree of correlation of anchor vector elements on the M second log anomaly feature interaction anchor vectors respectively. When the result of the anchor vector element correlation judgment meets the preset correlation threshold, the anchor vector elements are combined to obtain M first correlation iterative element group sets; the M first correlation iterative element group sets and the M first log anomaly feature interaction anchor vectors are added to the M second log anomaly feature interaction anchor vectors respectively into an initially empty vector to obtain M first anomaly iterative memories.

[0035] Preferably, the M first log anomaly feature interaction anchor vectors are used to determine the anchor vector element correlation with the M second log anomaly feature interaction anchor vectors. Specifically, a dimension-by-dimensional similarity analysis is performed on the corresponding vector elements in each pair of first and second log anomaly feature interaction anchor vectors, and their element-level correlation is calculated using a predetermined similarity function (e.g., weighted cosine similarity, Manhattan distance, or a custom vector element mapping function). When the correlation determination result meets a preset correlation threshold (e.g., a correlation value greater than a certain threshold θ), the pair of anchor vectors is determined to have a valid semantic or behavioral correlation relationship. The anchor vector elements that meet the condition are then combined to generate M first correlation iteration element groups, which represent the initial anomaly feature structure to be focused on in subsequent iterations. Subsequently, these M first correlation iteration element groups, along with the corresponding M first log anomaly feature interaction anchor vectors and M second log anomaly feature interaction anchor vectors, are added to an initially empty vector container to form a first-round anomaly pattern memory set that can be used for iterative tracking, namely, the M first anomaly iteration memories.

[0036] Obtain the logistics resource distribution information of the target area, combine it with the logistics task work order set to perform intelligent allocation, and obtain the target intelligent allocation plan.

[0037] By connecting data with logistics resource perception terminals deployed in the target area (such as personnel positioning terminals, equipment status monitoring nodes, material storage and allocation systems, etc.), the real-time distribution information of logistics resources in the area is collected and summarized to form a resource distribution data structure covering various resource types (including manpower, tools, equipment, materials, etc.); at the same time, the task type, service content, estimated time, urgency and location information of each work order are extracted from the generated logistics task work order set to form a logistics task work order location information set. On this basis, combining the location information set of logistics task work orders with the current resource distribution data, by constructing a multi-objective adaptive scheduling model, taking into account factors such as minimizing task response time, maximizing resource load balancing, and optimizing task priority matching, all work order tasks to be executed are resource bound and location adapted to obtain an initial intelligent allocation plan; then, a resource utilization analysis is performed on the initial allocation results to judge the rationality and overall efficiency of resource allocation under the current plan. If the analysis results do not meet the preset performance indicators (for example, the utilization rate of a certain type of resource is lower than the threshold or the task waiting time is too long), the policy-driven intelligent adjustment mechanism is further enabled to optimize and iterate the allocation results, and finally generate a target intelligent allocation plan that meets the requirements of scheduling efficiency and resource utilization balance.

[0038] Furthermore, the logistics resource distribution information of the target area is obtained, and intelligent allocation is performed based on the logistics task work order set to obtain a target intelligent allocation plan, including: Extract the work order location information of the logistics task work order set to obtain the logistics task work order location information set; combine the logistics resource distribution information and the logistics task work order location information set to adaptively allocate the logistics task work order set to obtain an initial intelligent allocation plan; perform resource utilization analysis on the initial intelligent allocation plan, and if the analysis result does not meet the requirements, adjust the initial intelligent allocation plan to obtain a target intelligent allocation plan.

[0039] First, the logistics resource management system collects the real-time distribution status of various logistics resources within the target area. These resources include, but are not limited to, the location and availability of logistics personnel, the storage location and usage status of various equipment and tools, and the inventory and distribution of materials, forming a logistics resource distribution information database. Subsequently, information such as the specific execution location, task type, and task urgency of each work order is extracted from the collection of logistics task work orders to construct a collection of logistics task work order location information. Based on these two key pieces of information, an adaptive allocation mechanism is used to perform preliminary resource matching on the collection of logistics task work orders. This mechanism comprehensively considers multiple factors, such as task urgency, resource allocation cost, path distance, and execution capability matching, and generates an initial intelligent allocation plan using a multi-objective optimization algorithm. Specifically, the real-time location, available status, and capacity parameters of various logistics resources in the target area are matched with the task location and demand information in the logistics task work order location information set to construct a resource-task association matrix. Based on this association matrix, an adaptive scheduling algorithm is used to dynamically adjust the matching relationship between tasks and resources by comprehensively considering multiple factors such as the urgency of the task, resource availability, task execution distance, and resource load balancing. Through iterative optimization, the allocation plan is continuously adjusted to reduce task response time and resource idle rate. Finally, an initial intelligent allocation plan is generated that can effectively cover all task requirements and has high resource utilization.

[0040] The system analyzes resource utilization based on this initial intelligent allocation plan, calculating workload balance for logistics personnel and equipment, task completion time estimates, and overall resource utilization efficiency. If the analysis indicates issues such as idle resources or delayed task responses, a rule-based and machine learning-based adjustment mechanism is activated to dynamically adjust the resource allocation ratio, task sequence, and task-resource matching within the allocation plan, continuously optimizing the allocation results. Ultimately, the system outputs a targeted intelligent allocation plan that meets performance indicators, maximizes resource utilization, and satisfies logistics mission requirements, providing scientific and efficient decision-making support for logistics management.

[0041] Further, including: According to the preset adjustment strategy, the initial intelligent allocation plan is randomly adjusted to obtain an adjusted intelligent allocation plan set; the adjusted intelligent allocation plan set is traversed to perform resource utilization analysis to obtain an adjusted intelligent allocation plan resource utilization set; it is determined whether there is an adjusted intelligent allocation plan resource utilization rate greater than or equal to the resource utilization rate of the initial intelligent allocation plan in the adjusted intelligent allocation plan resource utilization rate set; if so, the adjusted intelligent allocation plan corresponding to the maximum value in the adjusted intelligent allocation plan resource utilization rate set is used as the target intelligent allocation plan.

[0042] Specifically, according to a preset adjustment strategy, the number of work orders assigned to each handler in the initial intelligent allocation plan is moderately increased or decreased. The matching order and specific allocation relationships between tasks and resources may also be adjusted, generating multiple different adjustment plans. These adjustment plan sets are automatically and randomly generated by the system, ensuring that a variety of possible resource allocations are explored within a reasonable range. Subsequently, the system analyzes resource utilization for each plan in the set, primarily by calculating metrics such as the workload balance of logistics personnel and equipment, task response time, resource idleness, and overall task completion efficiency. This generates corresponding resource utilization values ​​and constructs a set of adjusted intelligent allocation plan resource utilizations. The system further compares the resource utilization metrics of all adjustment plans in this set to determine whether any adjustment plan has a resource utilization greater than or equal to that of the initial intelligent allocation plan. If so, the adjustment plan with the highest resource utilization is selected as the final target intelligent allocation plan, achieving optimal allocation and maximum utilization of logistics resources, ensuring the efficient completion of logistics tasks and the rational scheduling of resources.

[0043] Furthermore, the preset adjustment strategy is to increase or decrease the number of work orders handled by each processing personnel in the initial intelligent allocation plan.

[0044] The preset adjustment strategy specifically involves adjusting the number of work orders assigned to each handler in the initial intelligent allocation plan to a moderate degree. Specifically, the system randomly increases or decreases the number of work orders assigned to a handler based on a specific adjustment range and rules, thereby fine-tuning the overall task allocation plan.

[0045] In summary, the embodiments of the present application have at least the following technical effects: First, the M computing power boxes corresponding to M logistics equipment distributed in the target area interact to collect logistics equipment business data, obtaining M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer. Next, the M logistics equipment business work log sequences and M logistics equipment business work video sequences are traversed to extract abnormal feature anchor vectors, obtaining M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences. Furthermore, the M log abnormal feature anchor vector sequences are subjected to attention interaction fusion based on the M video abnormal feature anchor vector sequences, obtaining M log abnormal feature interaction anchor vector sequences. Then, logistics task work orders are generated based on the M log abnormal feature interaction anchor vector sequences, obtaining a logistics task work order set. Finally, logistics resource distribution information for the target area is obtained and intelligently allocated based on the logistics task work order set, resulting in a target intelligent allocation solution. This solves the technical problems of uneven logistics task distribution and low resource utilization in the existing technology, achieving efficient matching of tasks and resources and improving logistics resource utilization.

[0046] The second embodiment is based on the same inventive concept as the method for intelligently allocating logistics tasks facing multi-source demands in the previous embodiment. Figure 2 As shown, the present application provides a multi-source demand-oriented logistics task intelligent allocation system, wherein the system includes: Data acquisition module 11: interacts with M computing power boxes corresponding to M logistics equipment distributed in the target area, collects logistics equipment business data, and obtains M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer; anchor vector extraction module 12: traverses the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtains M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences; fusion module 13: performs attention interaction fusion on the M log abnormal feature anchor vector sequences based on the M video abnormal feature anchor vector sequences, and obtains M log abnormal feature interaction anchor vector sequences; work order generation module 14: generates logistics task work orders based on the M log abnormal feature interaction anchor vector sequences, and obtains a logistics task work order set; allocation module 15: obtains logistics resource distribution information of the target area, and performs intelligent allocation in combination with the logistics task work order set to obtain a target intelligent allocation plan.

[0047] Furthermore, the anchor vector extraction module 12 is configured to perform the following method: Pre-construct a text abnormality feature identifier and a video abnormality feature identifier; use the text abnormality feature identifier to extract abnormality feature anchor vectors from the M logistics equipment business work log sequences to obtain the M log abnormality feature anchor vector sequences; use the video abnormality feature identifier to extract abnormality feature anchor vectors from the M logistics equipment business work video sequences to obtain M video abnormality feature anchor vector sequences.

[0048] Furthermore, the fusion module 13 is configured to perform the following method: The M video abnormality feature anchor point vector sequences and the M log abnormality feature anchor point vector sequences are simultaneously identified with anchor point vector mapping interaction coefficients to obtain M mapping interaction coefficient sequences; M attention interaction fusion matrices are constructed based on the M mapping interaction coefficient sequences; the M log abnormality feature interaction anchor point vector sequences are interactively fused using the M attention interaction fusion matrices to obtain M log abnormality feature interaction anchor point vector sequences.

[0049] Furthermore, the work order generation module 14 is used to execute the following method: Traversing the M log anomaly feature interaction anchor point vector sequences to perform anomaly association iterative analysis, and obtaining M target anomaly iterative memories; performing logistics task work order identification based on the M target anomaly iterative memories, and obtaining a logistics task work order set.

[0050] Furthermore, the work order generation module 14 is used to execute the following method: M first log anomaly feature interaction anchor point vectors in the M log anomaly feature interaction anchor point vector sequence are used to perform anomaly association iterative extraction on M second log anomaly feature interaction anchor point vectors to obtain M first anomaly iterative memories; based on the M first anomaly iterative memories, anomaly association iterative extraction is performed on the M log anomaly feature interaction anchor point vector sequence to obtain M second anomaly iterative memories; and so on, anomaly association iterative analysis is performed on the M log anomaly feature interaction anchor point vector sequence based on the M second anomaly iterative memories to obtain the M target anomaly iterative memories.

[0051] Furthermore, the work order generation module 14 is used to execute the following method: The M first log anomaly feature interaction anchor vectors are used to judge the degree of correlation of anchor vector elements on the M second log anomaly feature interaction anchor vectors respectively. When the result of the anchor vector element correlation judgment meets the preset correlation threshold, the anchor vector elements are combined to obtain M first correlation iterative element group sets; the M first correlation iterative element group sets and the M first log anomaly feature interaction anchor vectors are added to the M second log anomaly feature interaction anchor vectors respectively into an initially empty vector to obtain M first anomaly iterative memories.

[0052] Furthermore, the allocation module 15 is configured to execute the following method: Extract the work order location information of the logistics task work order set to obtain the logistics task work order location information set; combine the logistics resource distribution information and the logistics task work order location information set to adaptively allocate the logistics task work order set to obtain an initial intelligent allocation plan; perform resource utilization analysis on the initial intelligent allocation plan, and if the analysis result does not meet the requirements, adjust the initial intelligent allocation plan to obtain a target intelligent allocation plan.

[0053] Furthermore, the allocation module 15 is configured to execute the following method: According to the preset adjustment strategy, the initial intelligent allocation plan is randomly adjusted to obtain an adjusted intelligent allocation plan set; the adjusted intelligent allocation plan set is traversed to perform resource utilization analysis to obtain an adjusted intelligent allocation plan resource utilization set; it is determined whether there is an adjusted intelligent allocation plan resource utilization rate greater than or equal to the resource utilization rate of the initial intelligent allocation plan in the adjusted intelligent allocation plan resource utilization rate set; if so, the adjusted intelligent allocation plan corresponding to the maximum value in the adjusted intelligent allocation plan resource utilization rate set is used as the target intelligent allocation plan.

[0054] Furthermore, the allocation module 15 is configured to execute the following method: The preset adjustment strategy is to increase or decrease the number of work orders handled by each processing personnel in the initial intelligent allocation plan.

[0055] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0056] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0057] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A method for intelligently allocating logistics tasks based on multi-source demands, characterized in that: The method comprises: Interact with the M computing power boxes corresponding to the M logistics equipment distributed in the target area to collect logistics equipment business data, and obtain M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer; Traversing the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtaining M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences; Performing attention interaction fusion on the M log abnormality feature anchor point vector sequences based on the M video abnormality feature anchor point vector sequences to obtain M log abnormality feature interaction anchor point vector sequences; Generate logistics task work orders based on the M log abnormal feature interaction anchor point vector sequences to obtain a set of logistics task work orders; Obtain the logistics resource distribution information of the target area, combine it with the logistics task work order set to perform intelligent allocation, and obtain the target intelligent allocation plan.

2. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 1, characterized in that: Traversing the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtaining M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences, including: Pre-built text anomaly feature identifier and video anomaly feature identifier; Using a text abnormality feature identifier to extract abnormality feature anchor point vectors from the M logistics equipment business work log sequences, respectively, to obtain the M log abnormality feature anchor point vector sequences; The abnormal feature anchor vectors of the M logistics equipment business work video sequences are extracted using a video abnormal feature identifier to obtain M video abnormal feature anchor vector sequences.

3. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 1, characterized in that: Performing attention interaction fusion on the M log abnormality feature anchor point vector sequences based on the M video abnormality feature anchor point vector sequences to obtain M log abnormality feature interaction anchor point vector sequences, including: Performing simultaneous anchor vector mapping interaction coefficient identification on the M video anomaly feature anchor vector sequences and the M log anomaly feature anchor vector sequences to obtain M mapping interaction coefficient sequences; Constructing M attention interaction fusion matrices based on the M mapping interaction coefficient sequences; The M attention interaction fusion matrices are used to interactively fuse the M log anomaly feature interaction anchor point vector sequences to obtain M log anomaly feature interaction anchor point vector sequences.

4. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 1, characterized in that: Generate logistics task work orders based on the M log anomaly feature interaction anchor point vector sequences to obtain a set of logistics task work orders, including: Traversing the M log anomaly feature interaction anchor point vector sequences to perform anomaly correlation iterative analysis and obtain M target anomaly iterative memories; Logistics task work orders are identified based on the M target abnormal iterative memories to obtain a logistics task work order set.

5. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 4, characterized in that: Traverse the M log anomaly feature interaction anchor point vector sequences to perform anomaly correlation iterative analysis to obtain M target anomaly iterative memories, including: Using the M first log anomaly feature interaction anchor point vectors in the M log anomaly feature interaction anchor point vector sequence to perform anomaly association iterative extraction on the M second log anomaly feature interaction anchor point vectors, respectively, to obtain M first anomaly iterative memories; Performing abnormal correlation iterative extraction on the M log abnormal feature interaction anchor point vector sequences based on the M first abnormal iterative memories to obtain M second abnormal iterative memories; Similarly, based on the M second abnormality iterative memories, abnormality association iterative analysis is performed on the M log abnormality feature interaction anchor point vector sequences to obtain the M target abnormality iterative memories.

6. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 5, characterized in that: include: Using the M first log anomaly feature interaction anchor point vectors to judge the anchor point vector element correlation of the M second log anomaly feature interaction anchor point vectors, when the anchor point vector element correlation judgment result meets the preset correlation threshold, the anchor point vector elements are combined to obtain M first correlation iterative element group sets; The M first associated iterative element group sets, the M first log anomaly feature interaction anchor point vectors, and the M second log anomaly feature interaction anchor point vectors are respectively added to an initially empty vector to obtain M first anomaly iterative memories.

7. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 1, characterized in that: Obtain the logistics resource distribution information of the target area, combine it with the logistics task work order set to perform intelligent allocation, and obtain the target intelligent allocation plan, including: Extracting the work order location information of the logistics task work order set to obtain a logistics task work order location information set; Combining the logistics resource distribution information and the logistics task work order location information set, adaptively allocating the logistics task work order set to obtain an initial intelligent allocation plan; A resource utilization analysis is performed on the initial intelligent allocation plan. If the analysis result does not meet the requirements, the initial intelligent allocation plan is adjusted to obtain a target intelligent allocation plan.

8. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 7, characterized in that: include: According to a preset adjustment strategy, the initial intelligent allocation scheme is randomly adjusted to obtain an adjusted intelligent allocation scheme set; Traversing the set of adjusted intelligent allocation solutions to perform resource utilization analysis and obtain a set of adjusted intelligent allocation solution resource utilization rates; Determine whether there is an adjusted intelligent allocation scheme resource utilization rate greater than or equal to the resource utilization rate of the initial intelligent allocation scheme in the adjusted intelligent allocation scheme resource utilization rate set. If so, the adjusted intelligent allocation scheme corresponding to the maximum value in the adjusted intelligent allocation scheme resource utilization rate set is used as the target intelligent allocation scheme.

9. The method for intelligently allocating logistics tasks for multi-source demand as claimed in claim 8, characterized in that: The preset adjustment strategy is to increase or decrease the number of work orders handled by each processing personnel in the initial intelligent allocation plan.

10. An intelligent logistics task allocation system for multi-source demand, characterized by: A system for implementing a multi-source demand-oriented logistics task intelligent allocation method according to any one of claims 1 to 9, comprising: Data collection module: interacts with the M computing power boxes corresponding to the M logistics equipment distributed in the target area to collect logistics equipment business data, obtaining M logistics equipment business work log sequences and M logistics equipment business work video sequences, where M is a positive integer; Anchor vector extraction module: traverses the M logistics equipment business work log sequences and the M logistics equipment business work video sequences to extract abnormal feature anchor vectors, and obtains M log abnormal feature anchor vector sequences and M video abnormal feature anchor vector sequences; Fusion module: performing attention interaction fusion on the M log abnormality feature anchor point vector sequences based on the M video abnormality feature anchor point vector sequences to obtain M log abnormality feature interaction anchor point vector sequences; Work order generation module: Generates logistics task work orders based on the M log anomaly feature interaction anchor point vector sequences to obtain a set of logistics task work orders; Allocation module: obtains the logistics resource distribution information of the target area, combines it with the logistics task work order set to perform intelligent allocation, and obtains the target intelligent allocation plan.

Images