AI-based edge computing-based intelligent logistics terminal equipment collaborative management and control system

Abstract

This invention relates to the field of logistics management technology, specifically to a collaborative control system for intelligent logistics terminal equipment based on AI edge computing. The system includes an edge data fusion and status recognition module deployed at edge nodes, which collects multi-source data such as operating status, environmental perception, and communication links, and fuses this data to generate an equipment status representation vector. An intelligent prediction and task scheduling decision module uploads the status vector to the cloud, predicts task completion capabilities and failure probabilities, and generates a task scheduling decision package. A scheduling strategy distribution and edge execution coordination module distributes the scheduling package through a multi-protocol gateway. Edge nodes complete task distribution, communication switching, and resource scheduling, and cache key strategies. This invention enables real-time perception of the status of logistics terminal equipment, intelligent prediction and resource optimization for task scheduling, and rapid response and fault-tolerant control in fault scenarios, significantly improving the system's operating efficiency, intelligence level, and stability.

Description

Technical Field

[0001] This invention relates to the field of logistics management technology, and in particular to a collaborative control system for intelligent logistics terminal equipment based on AI edge computing. Background Technology

[0002] With the development of smart logistics and intelligent manufacturing, logistics terminal equipment such as delivery robots, warehousing and handling equipment, and vehicle-mounted terminals are being deployed more and more densely in logistics scenarios. These devices need to have the ability to operate continuously, be flexibly scheduled, and coordinate efficiently. At the same time, the logistics system has increasingly higher requirements for the real-time performance and accuracy of task scheduling. It is urgent to achieve dynamic perception, intelligent prediction, and resource optimization of the terminal equipment's operating status through edge computing, artificial intelligence, and cloud platform collaboration in order to cope with changing environmental conditions and task requirements.

[0003] Most existing logistics scheduling systems rely on central servers for task distribution and status assessment, which leads to problems such as delayed data updates, incomplete equipment status awareness, and fixed scheduling strategies. Equipment status information often comes from heterogeneous sources and is collected inconsistently, lacking a unified data fusion and standardization mechanism. The scheduling strategy cannot perceive the equipment's execution capabilities and failure risks, resulting in unreasonable task allocation and even equipment overload or scheduling conflicts. In addition, existing systems have limited response capabilities to communication link fluctuations and equipment anomalies, lacking sound fault tolerance mechanisms and local execution closed loops, making it difficult to meet the high stability and high intelligence requirements of large-scale intelligent logistics systems. Summary of the Invention

[0004] This invention provides a collaborative management and control system for intelligent logistics terminal equipment based on AI edge computing, which realizes intelligent, efficient and highly stable control of equipment scheduling, and solves the technical problems of existing technologies such as incomplete state perception, rigid scheduling and weak fault tolerance.

[0005] The AI-based edge computing-based intelligent logistics terminal equipment collaborative management and control system includes an edge data fusion and status recognition module, an intelligent prediction and task scheduling decision module, and a scheduling strategy distribution and edge execution collaboration module.

[0006] The edge data fusion and status recognition module collects multi-source data from logistics terminal equipment in real time through edge computing nodes deployed in the logistics scenario, including operating status data, environmental perception data and communication link status, and performs fusion processing on the multi-source data to generate equipment status representation vectors, including equipment load status, positioning information, communication quality level and equipment fault risk score.

[0007] The intelligent prediction and task scheduling decision module uploads the device status representation vector to the cloud AI scheduling engine. Based on the fused historical operation data and real-time device status representation vector, the AI ​​scheduling engine predicts the task completion capability and potential failure probability of the target device in the future scheduling cycle, and dynamically generates a task scheduling decision package based on the prediction results, including task allocation scheme, resource reservation strategy and anomaly handling priority.

[0008] The scheduling strategy distribution and edge execution coordination module distributes the task scheduling decision package to each edge computing node through a multi-protocol adaptation gateway. The edge computing node then distributes tasks, switches communication paths, and schedules resources for the connected logistics terminal devices based on the task scheduling decision package, while caching key control strategies locally.

[0009] Optionally, the edge data fusion and state recognition module includes:

[0010] Multi-source data acquisition and standardization: In edge computing nodes, real-time data acquisition of operational status data, environmental perception data, and communication link status of logistics terminal equipment is performed, and time alignment, format conversion, and unit unification are performed on the multi-source data to form standardized data;

[0011] State feature extraction and vector construction: Based on standardized data, key features reflecting the equipment's operating status are extracted, including equipment load, location information, and communication quality level, and a structured equipment state representation vector is constructed.

[0012] Fault risk assessment and information completion: Using the prediction model built into the edge node, the state characteristics of the equipment are analyzed to generate a fault risk score, and the fault risk score is integrated into the equipment state representation vector.

[0013] Optionally, the multi-source data acquisition and standardization includes:

[0014] Multi-type data synchronous collection: Edge computing nodes periodically collect operational status data (CPU utilization, task queue length, memory usage), environmental perception data (temperature and humidity, GPS location), and communication link status data (communication signal strength, packet loss rate, average latency) of logistics terminal equipment through local interfaces, and add a local timestamp to each data item.

[0015] Time alignment and sampling window construction: applying a unified time step to multi-source data A sliding sampling window is constructed, and synchronization is performed using the nearest time point alignment principle to form an aligned data sequence;

[0016] Data format unification and unit standardization: Convert the collected multi-source data into a unified structure format, including field names, data types (such as floating-point numbers and boolean values), and unit specifications;

[0017] Standardized data structure output: Packages the processed multi-source data into structured standard data vectors. .

[0018] Optionally, the state feature extraction and vector construction include:

[0019] Device load feature extraction: CPU utilization based on standardized data Task queue length and memory usage Calculate the overall load score of the equipment. ;

[0020] Location Information Extraction and Structure: Extracting Global Positioning System (GPS) Coordinates from Standardized Data And structure it into a location information field. ,in, This is the current latitude value. This is the current longitude value;

[0021] Communication quality level assessment: based on communication signal strength Packet loss rate With average delay Calculate communication quality score And mapped to communication level labels ;

[0022] Device state representation vector construction: Encapsulate the extracted key features into a device state representation vector with a unified structure. .

[0023] Optionally, the fault risk assessment and information completion include:

[0024] State feature vector preparation: Extracting sub-feature vectors for fault prediction from the constructed device state representation vector. ;

[0025] Risk scoring model prediction: sub-feature vectors Input a risk scoring model for local deployment of edge nodes, output a device failure risk score. ;

[0026] Vector information completion and updating: This involves assigning equipment failure risk scores. These parameters are stored separately as local fault-tolerant control parameters for edge nodes. If the edge node independently performs task anomaly warning and local scheduling, an auxiliary control vector including fault scoring is constructed. .

[0027] Optionally, the intelligent prediction and task scheduling decision module includes:

[0028] Predictive Input Construction: The cloud-based AI scheduling engine receives device state representation vectors from edge nodes and fuses them with the historical operating data (task records, fault logs, communication performance) of the corresponding devices to generate an input feature set for prediction;

[0029] Capability and Risk Prediction: Based on the predicted input feature set, the output is the target device's task completion capability score and failure probability in the next scheduling cycle, which is used to evaluate its task execution efficiency and stability.

[0030] Decision package generation: Based on the results of capability and risk prediction and current task requirements, a task scheduling decision package is generated, including task allocation scheme, resource reservation strategy and exception handling priority, and is issued to edge nodes as control instructions.

[0031] Optionally, the construction of the prediction input includes:

[0032] State Vector Reception and Time Calibration: The cloud-based AI scheduling engine periodically receives device state representation vectors uploaded from edge nodes. This includes device load rating, communication level, and location information, and assigns a receiving timestamp to each. ;

[0033] Historical feature sequence extraction: Retrieving the corresponding device from the cloud database in the past. Historical data of operation within each scheduling cycle, including task completion rate Fault marking With communication delay Construct a historical sequence matrix ;

[0034] Feature vector fusion generation: This involves generating a device state representation vector. With historical sequence matrix The predicted input feature set is generated using a weighted concatenation method. .

[0035] Optionally, the capability and risk prediction includes:

[0036] Multi-task prediction model input: the prediction input feature set The input is sent to a multi-task neural network model deployed in the cloud to simultaneously predict task completion capability scores and failure probability;

[0037] Task completion ability score calculation: In the multi-task neural network model, the first output channel performs a non-linear mapping on the input feature set to output a task completion ability score. Its value represents the device's ability to complete the current task on time within the next scheduling cycle;

[0038] Fault probability prediction output: The second output channel in the multi-task neural network model generates a device fault probability score based on the input feature set. This is used to measure the risk of the device failing in the next scheduling cycle.

[0039] Optionally, the decision package generation includes:

[0040] Matching equipment capabilities with task requirements: The cloud-based AI scheduling engine scores the task completion capability based on predictions. With the current task requirements set Complexity of each task The matching score between computing devices and tasks And select the top results from highest to lowest based on matching scores. Each task generates a task allocation scheme. ;

[0041] Resource reservation strategy formulation: scoring based on predicted equipment failure probability. Combined with task priority parameters Task allocation scheme Each task Develop a resource reservation strategy and calculate the corresponding resource reservation amount. ;

[0042] Anomaly handling priority classification: based on equipment failure probability scoring. Generate exception handling priority identifier , is represented as:

[0043] ;

[0044] in, , These are the high and low thresholds for abnormal alerts;

[0045] Scheduling decision package encapsulation and distribution: This involves packaging and distributing the task allocation scheme. Resource reservation strategy and exception handling priority identifier Encapsulated as a task scheduling decision package And it is sent to the corresponding edge computing nodes through the scheduling channel for control execution.

[0046] Optionally, the scheduling strategy distribution and edge execution coordination module includes:

[0047] Task scheduling packet protocol conversion and distribution: The cloud-based AI scheduling engine will generate task scheduling decision packets. The task scheduling decision packet is sent to the multi-protocol adaptation gateway. The multi-protocol adaptation gateway adapts and converts the communication protocol type (MQTT, CoAP, Modbus, HTTP) supported by the target edge node, and then sends the converted control command to each edge computing node.

[0048] Edge task parsing and allocation execution: Edge computing nodes that receive control commands parse the task allocation scheme in the task scheduling decision package. Based on the local available resource status table, the task distribution operation is completed to the connected logistics terminal equipment, and the communication link of the terminal equipment is switched according to the communication path identifier in the task scheduling decision package.

[0049] Resource scheduling and execution synchronization: Edge nodes allocate task resources based on the task scheduling decision package. It allocates processor time slices, memory, and communication bandwidth as needed, and implements local-level resource scheduling and queuing control for tasks in execution.

[0050] Key strategy local caching and fault tolerance preparation: Edge nodes cache copies of control strategies from the received task scheduling decision packages to the local control unit, including task priority information, exception handling level, and backup communication paths, to execute predefined fault tolerance mechanisms and emergency switching logic in the event of cloud outages or network fluctuations.

[0051] The beneficial effects of this invention are:

[0052] This invention, by deploying edge computing nodes in logistics scenarios, enables unified collection, standardization, and fusion processing of multi-source heterogeneous data such as the operating status, environmental perception, and communication link status of logistics terminal equipment. It constructs a structured equipment status representation vector and further combines it with a local lightweight model to complete the fault risk score, thereby significantly improving the real-time perception and local judgment capabilities of equipment operating status.

[0053] This invention, by inputting the fused device state vector and historical device behavior data into a multi-task prediction model, can simultaneously predict the device's task completion capability and failure probability. Combined with task complexity, task priority, and communication resource requirements, it dynamically generates a task scheduling decision package that includes task allocation, resource reservation, and anomaly handling priority. This mechanism overcomes the problems of rigid task allocation rules and weak prediction capabilities in traditional logistics systems, and realizes personalized, dynamic, and risk-aware task scheduling, significantly improving overall scheduling accuracy and system robustness.

[0054] This invention distributes task scheduling decision packages to edge computing nodes through a multi-protocol adapter gateway. The edge nodes can then perform task distribution, resource scheduling, and communication link switching based on the package content. Simultaneously, key control strategies are cached locally, enabling automatic fault tolerance and rapid recovery in the event of cloud-based link loss or network jitter. The scheduling execution results are then synchronized back to the cloud, constructing a cloud-edge control closed loop. This achieves the system's self-awareness of its state, self-optimization of its strategies, and adaptive execution capabilities, thereby improving the logistics system's response efficiency, communication stability, and intelligent collaboration level in complex and dynamic environments. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a schematic diagram of the system functional modules according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of the intelligent prediction and task scheduling decision module in an embodiment of the present invention. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0059] like Figures 1-2 As shown, the AI ​​edge computing-based intelligent logistics terminal equipment collaborative management and control system includes an edge data fusion and status recognition module, an intelligent prediction and task scheduling decision module, and a scheduling strategy distribution and edge execution collaboration module, wherein;

[0060] The edge data fusion and status recognition module collects multi-source data from logistics terminal equipment in real time through edge computing nodes deployed in logistics scenarios, including operating status data, environmental perception data and communication link status. It then fuses the multi-source data to generate equipment status representation vectors, including equipment load status, location information, communication quality level and equipment fault risk score.

[0061] The intelligent prediction and task scheduling decision module uploads the device status representation vector to the cloud AI scheduling engine. Based on the fused historical operation data and real-time device status representation vector, the AI ​​scheduling engine predicts the task completion capability and potential failure probability of the target device in the future scheduling cycle, and dynamically generates a task scheduling decision package based on the prediction results, including task allocation scheme, resource reservation strategy and anomaly handling priority.

[0062] The scheduling strategy distribution and edge execution coordination module distributes the task scheduling decision package to each edge computing node through a multi-protocol adaptation gateway. The edge computing nodes then distribute tasks, switch communication paths, and schedule resources for the connected logistics terminal devices based on the task scheduling decision package, while caching key control strategies locally.

[0063] The edge data fusion and status recognition module includes:

[0064] Multi-source data acquisition and standardization: In edge computing nodes, real-time data acquisition of operational status data, environmental perception data, and communication link status of logistics terminal equipment is performed, and time alignment, format conversion, and unit unification are performed on the multi-source data to form standardized data;

[0065] State feature extraction and vector construction: Based on standardized data, key features reflecting the equipment's operating status are extracted, including equipment load, location information, and communication quality level, and a structured equipment state representation vector is constructed.

[0066] Fault risk assessment and information completion: Using the prediction model built into the edge node, the state characteristics of the equipment are analyzed to generate a fault risk score, and the fault risk score is integrated into the equipment state representation vector.

[0067] Multi-source data acquisition and standardization include:

[0068] Multi-type data synchronous collection: Edge computing nodes periodically collect operational status data (CPU utilization, task queue length, memory usage), environmental perception data (temperature and humidity, GPS location), and communication link status data (communication signal strength, packet loss rate, average latency) of logistics terminal equipment through local interfaces, and add a local timestamp to each data item.

[0069] Time alignment and sampling window construction: applying a unified time step to multi-source data Construct a sliding sampling window and synchronize using the nearest time point alignment principle to form an aligned data sequence, represented as:

[0070] ;

[0071] in, For the aligned multi-source data, For the first The original value of the class data, Sampling time and uniform time for each type of data Time deviation, To align with the target timeline;

[0072] Data format standardization and unit standardization: Convert collected multi-source data into a unified structure format, including field names, data types (such as floating-point numbers, Boolean values), and unit specifications. For example, convert temperature to degrees Celsius. Power is standardized to watts (W) and recorded in a standardized data dictionary;

[0073] Standardized data structure output: Packages the processed multi-source data into structured standard data vectors. , is represented as:

[0074] ;

[0075] in, The standardized first A status field.

[0076] State feature extraction and vector construction include:

[0077] Device load feature extraction: CPU utilization based on standardized data Task queue length and memory usage Calculate the overall load score of the equipment. , is represented as:

[0078] ;

[0079] in, The load characteristic weighting coefficients are used to weight the load characteristics. This represents the maximum capacity of the task queue.

[0080] Location Information Extraction and Structure: Extracting Global Positioning System (GPS) Coordinates from Standardized Data And structure it into a location information field. ,in, This is the current latitude value. This is the current longitude value;

[0081] Communication quality level assessment: based on communication signal strength Packet loss rate With average delay Calculate communication quality score And mapped to communication level labels , is represented as:

[0082] ;

[0083] ;

[0084] in, For communication quality weighting coefficients, , This serves as a normalized reference value for communication signals and time delays. , Thresholds are set for communication level classification;

[0085] Communication level classification threshold , The settings include:

[0086] (1) Statistical score sample distribution: collect communication scores of different logistics terminals under various environments. To form a sample set ;

[0087] (2) Setting grade ratio requirements: High communication grade accounts for 30%, medium grade accounts for 50%, and low grade accounts for 20%;

[0088] (3) Set the initial threshold based on the quantiles:

[0089] ;

[0090] ;

[0091] in, The minimum value for the top 30% of communication scores. The highest value among the top 20% of communication scores;

[0092] Device state representation vector construction: Encapsulate the extracted key features into a device state representation vector with a unified structure. .

[0093] Fault risk assessment and information completion include:

[0094] State feature vector preparation: Extracting sub-feature vectors for fault prediction from the constructed device state representation vector. , is represented as:

[0095] ;

[0096] in, For CPU utilization, The length of the task queue. For memory usage, To score the quality of communication, The current temperature of the device;

[0097] Risk scoring model prediction: sub-feature vectors Input a risk scoring model for local deployment of edge nodes, output a device failure risk score. , is represented as:

[0098] ;

[0099] in, A fault risk score is assigned to the current equipment, representing the probability of a predicted fault. These are the standardized feature values ​​in the device state feature vector. For example, For CPU utilization, The length of the task queue. For memory usage, To score the quality of communication, For equipment temperature, This is the bias term (intercept) of the logistic regression model. These are the model weight parameters. is the base of the natural logarithm;

[0100] Vector information completion and updating: This involves assigning equipment failure risk scores. It is stored separately as a local fault-tolerant control parameter for edge nodes and is not included in the device state representation vector. The updates are not uploaded to the cloud for building the prediction input feature set. If edge nodes need to independently perform task anomaly warning and local scheduling, an auxiliary control vector containing fault scores can be built. .

[0101] The intelligent prediction and task scheduling decision-making module includes:

[0102] Predictive Input Construction: The cloud-based AI scheduling engine receives device state representation vectors from edge nodes and fuses them with the historical operating data (task records, fault logs, communication performance) of the corresponding devices to generate an input feature set for prediction;

[0103] Capability and Risk Prediction: Based on the predicted input feature set, the output is the target device's task completion capability score and failure probability in the next scheduling cycle, which is used to evaluate its task execution efficiency and stability.

[0104] Decision package generation: Based on the results of capability and risk prediction and current task requirements, a task scheduling decision package is generated, including task allocation scheme, resource reservation strategy and exception handling priority, and is issued to edge nodes as control instructions.

[0105] The prediction input construction includes:

[0106] State Vector Reception and Time Calibration: The cloud-based AI scheduling engine periodically receives device state representation vectors uploaded from edge nodes. This includes device load rating, communication level, and location information, and assigns a receiving timestamp to each. ;

[0107] Historical feature sequence extraction: Retrieving the corresponding device from the cloud database in the past. Historical data of operation within each scheduling cycle, including task completion rate Fault marking With communication delay Construct a historical sequence matrix , is represented as:

[0108] ;

[0109] in, For the first Task completion rate over a given period This is a flag value indicating whether a fault has occurred; 1 indicates a fault, and 0 indicates normal operation. For average communication delay, Set the size of the history window;

[0110] Feature vector fusion generation: This involves generating a device state representation vector. With historical sequence matrix The predicted input feature set is generated using a weighted concatenation method. , is represented as:

[0111] ;

[0112] in, To take the average of each column of historical features, Weighting coefficients are used to integrate historical features. This is a vector concatenation operation.

[0113] Capability and risk assessment includes:

[0114] Multi-task prediction model input: the prediction input feature set The input is sent to a multi-task neural network model deployed in the cloud to simultaneously predict task completion capability scores and failure probability;

[0115] Task completion ability score calculation: In the multi-task neural network model, the first output channel performs a non-linear mapping on the input feature set to output a task completion ability score. Its value represents the device's ability to complete the current task on time in the next scheduling cycle, and is expressed as:

[0116] ;

[0117] in, It is the ReLU activation function. This represents the weight vector for the task completion ability scoring channel. For bias terms, This is the Sigmoid function, used to output normalized scores;

[0118] Fault probability prediction output: The second output channel in the multi-task neural network model generates a device fault probability score based on the input feature set. This is used to measure the risk of the device failing in the next scheduling cycle, and is expressed as:

[0119] ;

[0120] in, This represents the weight vector for the fault probability prediction channel. This corresponds to the bias term.

[0121] Decision package generation includes:

[0122] Matching equipment capabilities with task requirements: The cloud-based AI scheduling engine scores the task completion capability based on predictions. With the current task requirements set Complexity of each task The matching score between computing devices and tasks And select the top results from highest to lowest based on matching scores. Each task generates a task allocation scheme. , is represented as:

[0123] ;

[0124] in, For equipment and tasks Match score, For the task complexity coefficient , This is a task complexity adjustment factor;

[0125] ;

[0126] in, The amount of computing resources required to execute the task. For task latency sensitivity, The communication bandwidth required for the task. , , The preset normalized reference upper limit value, , , These are weighting coefficients;

[0127] Resource reservation strategy formulation: scoring based on predicted equipment failure probability. Combined with task priority parameters Task allocation scheme Each task Develop a resource reservation strategy and calculate the corresponding resource reservation amount. , is represented as:

[0128] ;

[0129] in, This is the resource scheduling scaling factor;

[0130] Anomaly handling priority classification: based on equipment failure probability scoring. Generate exception handling priority identifier , is represented as:

[0131] ;

[0132] in, , These are the high and low thresholds for abnormal alerts;

[0133] High and low thresholds for abnormal alerts , The setting is expressed as:

[0134] ;

[0135] in, This indicates that the equipment is not actually faulty. For the first The probability of a sample being predicted as a fault. The allowable false alarm rate;

[0136] ;

[0137] in, This indicates that the equipment has actually malfunctioned. The allowable false negative rate;

[0138] Scheduling decision package encapsulation and distribution: This involves packaging and distributing the task allocation scheme. Resource reservation strategy and exception handling priority identifier Encapsulated as a task scheduling decision package And it is sent to the corresponding edge computing nodes through the scheduling channel for control execution.

[0139] The scheduling policy distribution and edge execution coordination module includes:

[0140] Task scheduling packet protocol conversion and distribution: The cloud-based AI scheduling engine will generate task scheduling decision packets. The task scheduling decision packet is sent to the multi-protocol adaptation gateway. The multi-protocol adaptation gateway adapts and converts the communication protocol type (MQTT, CoAP, Modbus, HTTP) supported by the target edge node, and then sends the converted control command to each edge computing node.

[0141] Edge task parsing and allocation execution: Edge computing nodes that receive control commands parse the task allocation scheme in the task scheduling decision package. Based on the local available resource status table, the task distribution operation is completed to the connected logistics terminal equipment, and the communication link of the terminal equipment is switched according to the communication path identifier in the task scheduling decision package.

[0142] Resource scheduling and execution synchronization: Edge nodes allocate task resources based on the task scheduling decision package. It allocates processor time slices, memory, and communication bandwidth as needed, and implements local-level resource scheduling and queuing control for tasks in execution.

[0143] Key strategy local caching and fault tolerance preparation: Edge nodes cache copies of control strategies from the received task scheduling decision packages to the local control unit, including task priority information, exception handling level, and backup communication paths, to execute predefined fault tolerance mechanisms and emergency switching logic in the event of cloud outages or network fluctuations.

[0144] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0145] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A collaborative management and control system for intelligent logistics terminal equipment based on AI edge computing, characterized in that, It includes an edge data fusion and status recognition module, an intelligent prediction and task scheduling decision module, and a scheduling strategy distribution and edge execution coordination module, among which; The edge data fusion and status recognition module collects multi-source data from logistics terminal equipment in real time through edge computing nodes deployed in the logistics scenario, including operating status data, environmental perception data and communication link status, and performs fusion processing on the multi-source data to generate equipment status representation vectors, including equipment load status, positioning information, communication quality level and equipment fault risk score. The intelligent prediction and task scheduling decision-making module specifically includes: Predictive Input Construction: The cloud-based AI scheduling engine receives device state representation vectors from edge nodes and fuses them with the historical operating data of the corresponding devices to generate an input feature set for prediction. Capability and Risk Prediction: Based on the predicted input feature set, the output is the target device's task completion capability score and failure probability in the next scheduling cycle, which is used to evaluate its task execution efficiency and stability. Decision package generation: Based on the results of capability and risk prediction and current task requirements, a task scheduling decision package is generated, including task allocation scheme, resource reservation strategy and exception handling priority, and is issued to edge nodes as control instructions; The construction of the prediction input includes: State Vector Reception and Time Calibration: The cloud-based AI scheduling engine periodically receives device state representation vectors uploaded from edge nodes. This includes device load rating, communication level, and location information, and assigns a receiving timestamp to each. ; Historical feature sequence extraction: Retrieving the corresponding device from the cloud database in the past. Historical data of operation within each scheduling cycle, including task completion rate Fault marking With communication delay Construct a historical sequence matrix ; Feature vector fusion generation: This involves generating a device state representation vector. With historical sequence matrix The predicted input feature set is generated using a weighted concatenation method. ; The capability and risk forecasts include: Multi-task prediction model input: the prediction input feature set The input is sent to a multi-task neural network model deployed in the cloud to simultaneously predict task completion capability scores and failure probability; Task completion ability score calculation: In the multi-task neural network model, the first output channel performs a non-linear mapping on the input feature set to output a task completion ability score. Its value represents the device's ability to complete the current task on time within the next scheduling cycle; Fault probability prediction output: The second output channel in the multi-task neural network model generates a device fault probability score based on the input feature set. This is used to measure the risk of the device failing in the next scheduling cycle; The decision package generation includes: Matching equipment capabilities with task requirements: The cloud-based AI scheduling engine matches the predicted task completion capability score with the current task requirement set. Complexity of each task The matching score between computing devices and tasks And select the top results from highest to lowest based on matching scores. Each task generates a task allocation scheme. ; Resource reservation strategy formulation: scoring based on predicted equipment failure probability. Combined with task priority parameters Task allocation scheme Each task Develop a resource reservation strategy and calculate the corresponding resource reservation amount. , represented as: ； in, This is the resource scheduling scaling factor; Anomaly handling priority classification: based on equipment failure probability scoring. Generate exception handling priority identifier , represented as: ； in, , These are the high and low thresholds for abnormal alerts; Scheduling decision package encapsulation and distribution: This involves packaging and distributing the task allocation scheme. Resource reservation strategy and exception handling priority identifier Encapsulated as a task scheduling decision package And it is sent to the corresponding edge computing nodes through the scheduling channel for control execution; The scheduling strategy distribution and edge execution coordination module distributes the task scheduling decision package to each edge computing node through a multi-protocol adaptation gateway. The edge computing node then distributes tasks, switches communication paths, and schedules resources for the connected logistics terminal devices based on the task scheduling decision package. At the same time, it caches key control policies locally, which are copies of the control policies in the task scheduling decision package.

2. The AI ​​edge computing-based intelligent logistics terminal equipment collaborative management and control system according to claim 1, characterized in that, The edge data fusion and state recognition module includes: Multi-source data acquisition and standardization: In edge computing nodes, real-time data acquisition of operational status data, environmental perception data, and communication link status of logistics terminal equipment is performed, and time alignment, format conversion, and unit unification are performed on the multi-source data to form standardized data; State feature extraction and vector construction: Based on standardized data, key features reflecting the equipment's operating status are extracted, including equipment load, location information, and communication quality level, and a structured equipment state representation vector is constructed. Fault risk assessment and information completion: Using the prediction model built into the edge node, the state characteristics of the equipment are analyzed to generate a fault risk score, and the fault risk score is integrated into the equipment state representation vector.

3. The AI ​​edge computing-based intelligent logistics terminal equipment collaborative management and control system according to claim 2, characterized in that, The multi-source data acquisition and standardization includes: Synchronous collection of multiple types of data: Edge computing nodes periodically collect operational status data, environmental perception data, and communication link status data of logistics terminal equipment through local interfaces, and add a local timestamp to each data item; Time alignment and sampling window construction: applying a unified time step to multi-source data A sliding sampling window is constructed, and synchronization is performed using the nearest time point alignment principle to form an aligned data sequence; Data format unification and unit standardization: Convert the collected multi-source data into a unified structure format, including field names, data types and unit specifications; Standardized data structure output: Packages the processed multi-source data into structured standard data vectors. .

4. The AI ​​edge computing-based intelligent logistics terminal equipment collaborative management and control system according to claim 3, characterized in that, The state feature extraction and vector construction include: Device load feature extraction: CPU utilization based on standardized data Task queue length and memory usage Calculate the overall load score of the equipment. ; Location Information Extraction and Structuring: Extracting Global Positioning System Coordinates from Standardized Data And structure it into a location information field. ,in, This is the current latitude value. This is the current longitude value; Communication quality level assessment: based on communication signal strength Packet loss rate With average delay Calculate communication quality score And mapped to communication level labels ; Device state representation vector construction: Encapsulate the extracted key features into a device state representation vector with a unified structure. .

5. The AI ​​edge computing-based intelligent logistics terminal equipment collaborative management and control system according to claim 4, characterized in that, The fault risk assessment and information completion include: State feature vector preparation: Extracting sub-feature vectors for fault prediction from the constructed device state representation vector. ; Risk scoring model prediction: sub-feature vectors Input a risk scoring model for local deployment of edge nodes, output a device failure risk score. ; Vector information completion and updating: This involves assigning equipment failure risk scores. These parameters are stored separately as local fault-tolerant control parameters for edge nodes. If the edge node independently performs task anomaly warning and local scheduling, an auxiliary control vector including fault scoring is constructed. .

6. The AI-based edge computing-based intelligent logistics terminal equipment collaborative management and control system according to claim 1, characterized in that, The scheduling strategy distribution and edge execution coordination module includes: Task scheduling packet protocol conversion and distribution: This involves converting the generated task scheduling decision packet into a single packet. The task scheduling decision packet is sent to the multi-protocol adaptation gateway. The multi-protocol adaptation gateway adapts and converts the format of the task scheduling decision packet according to the communication protocol type supported by the target edge node, and then sends the converted control command to each edge computing node. Edge task parsing and allocation execution: Edge computing nodes that receive control commands parse the task allocation scheme in the task scheduling decision package. Based on the local available resource status table, the task distribution operation is completed to the connected logistics terminal equipment, and the communication link of the terminal equipment is switched according to the communication path identifier in the task scheduling decision package. Resource scheduling and execution synchronization: Edge nodes allocate task resources based on the task scheduling decision package. It allocates processor time slices, memory, and communication bandwidth as needed, and implements local-level resource scheduling and queuing control for tasks in execution. Local caching and fault tolerance preparation of key control strategies: Edge nodes cache copies of control strategies from the received task scheduling decision packages to the local control unit, including task priority information, exception handling level, and backup communication paths, to execute predefined fault tolerance mechanisms and emergency switching logic in the event of cloud outages or network fluctuations.

Images