Logistics storage monitoring system based on artificial intelligence

Abstract

The invention relates to the technical field of monitoring systems, and discloses a logistics storage monitoring system based on artificial intelligence, and the system comprises a data collection module which is composed of an environment sensor array, a medicine batch identification unit and a space positioning device which are disposed on a storage three-dimensional shelf, the environment sensor array collects temperature, humidity and vibration parameters in real time, and the medicine batch identification unit stores unique codes and batch attribute data of medicines. According to the method, the dynamic association model of the medicine batch-environment sensor is established, and the spatial interpolation algorithm and the three-dimensional coordinate mapping technology are combined, so that the accurate association of the abnormal event and the medicine batch is realized. When an environment abnormity occurs in a certain area, the system can quickly position the batch of the affected medicine and generate the tracing map, and compared with a scheme in which environment data is disjointed from the position of the medicine in the prior art, the problems that the abnormity tracing efficiency is low and the high-value medicine is easy to deteriorate are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring systems, and in particular to a logistics warehousing monitoring system based on artificial intelligence. Background Art

[0002] Warehousing logistics is the use of self-built or leased warehouses and sites to store, load, unload, transport and distribute goods. The traditional definition of warehousing logistics is given from the perspective of material storage, but modern "warehousing logistics" is not a "warehouse" or "warehouse management" in the traditional sense, but warehousing logistics in the context of economic globalization and supply chain integration. In addition, in the actual use of warehousing logistics warehouses, monitoring systems are required to monitor and ensure the safety of goods.

[0003] For example, the Chinese invention application with publication number CN119539501A discloses an artificial intelligence-based logistics and warehousing monitoring system, including a warehouse environment collection module, an entry vehicle collection module, a departure vehicle collection module, a fire equipment collection module, a shelf collection module, a data processing module and an information sending module. The warehouse environment collection module is used to collect warehouse environment information, the entry vehicle collection module is used to collect entry vehicle information, the departure vehicle collection module is used to collect departure vehicle information, the shelf collection module is used to collect shelf information, and the fire equipment collection module is used to collect fire equipment information. The present invention can perform more comprehensive logistics and warehousing monitoring to ensure the safety of goods.

[0004] For example, the Chinese invention application with publication number CN114548876A discloses an artificial intelligence-based logistics and warehousing monitoring system, which includes an item monitoring end and a warehouse monitoring end. The item monitoring end is provided with a market analysis unit, a supply chain analysis unit and a characteristic analysis unit, and the warehouse monitoring end is provided with a location division unit, a storage matching unit and an inventory monitoring unit; real-time analysis of goods is performed to prevent risks in the storage of goods, which leads to increased storage costs and easily reduces logistics efficiency; real-time monitoring of the space in the warehouse is performed to prevent uneven space distribution in the warehouse, which leads to reduced warehouse goods storage efficiency and unnecessary cost increases.

[0005] The shortcomings of the above patents are:

[0006] Traditional systems can collect environmental data, but they lack a dynamic model linking drug storage locations with sensor nodes. This makes it difficult to quickly identify the specific drug batches affected when an environmental anomaly occurs in a specific area. This is especially true in high-rise warehouses storing multiple batches of vaccines and insulin. This significantly prolongs the time it takes to trace an anomaly and increases the risk of spoilage for high-value drugs.

[0007] Traditional threshold alarms trigger when environmental parameters exceed set thresholds. They fail to intelligently analyze the temporal characteristics of temperature change rates and fluctuation duration, making it difficult to reserve buffer time for measures such as refrigeration system adjustments and emergency drug transfers. This can easily lead to irreversible losses in biological storage scenarios where a constant temperature environment is required.

[0008] To this end, the present invention proposes a logistics warehousing monitoring system based on artificial intelligence to solve the above-mentioned problems. Summary of the Invention

[0009] In view of the shortcomings of the existing technology, the present invention provides a logistics warehousing monitoring system based on artificial intelligence to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a logistics warehousing monitoring system based on artificial intelligence, the logistics warehousing monitoring system comprising:

[0011] The data acquisition module consists of an environmental sensor array, a drug batch identification device, and a spatial positioning device deployed on the warehouse's three-dimensional shelves;

[0012] The spatial mapping module receives the three-dimensional coordinates of the drug storage location and environmental parameters from the data acquisition module, establishes a drug-sensor association model based on a spatial interpolation algorithm, and outputs environmental mapping data with distance weights and signal compensation coefficients;

[0013] The time series processing module receives the environmental mapping data from the spatial mapping module, processes the time series data through the gated recurrent unit network, and outputs the predicted values ​​of the environmental parameters in the future time window;

[0014] The risk assessment module receives the environmental parameter prediction value from the time sequence processing module and the preset drug thermal sensitivity coefficient, and calculates the comprehensive risk value including the temperature deviation and change rate;

[0015] The decision execution module matches the preset strategy library according to the comprehensive risk value of the risk assessment module, sends control instructions to the refrigeration unit, automatic guided vehicle, and alarm device, and generates a traceability data package containing drug codes, location coordinates, and environmental curves;

[0016] The feedback calibration module receives the control instruction execution status data and traceability data packet from the decision execution module, calculates the deviation between the actual effect of the thermal sensitivity coefficient and the expected value, and generates the drug thermal sensitivity coefficient correction parameter.

[0017] Preferably, the multi-parameter environmental sensor array is deployed in a cellular topology with a node spacing of less than 1.5 meters, includes a temperature sensor, a humidity sensor, and a vibration sensor, and has an operating temperature range of -40°C to 60°C.

[0018] Preferably, the environmental sensor array collects temperature, humidity, and vibration parameters in real time, the drug batch identification unit stores the drug's unique code and batch attribute data, and the spatial positioning device obtains the three-dimensional coordinates of the drug storage location through ultra-wideband technology.

[0019] Preferably, the batch identification unit adopts dual identification of ultra-high frequency radio frequency identification tag and low-temperature resistant QR code, the lower limit of the label working temperature is -40°C, and the QR code printing material can withstand a -30°C frosting environment.

[0020] Preferably, the data acquisition module is composed of an environmental sensor array, a drug batch identification device, and a spatial positioning device deployed on the three-dimensional storage shelves, and further includes:

[0021] The environmental sensing unit consists of a spatial hexahedron grid of sensor nodes, each of which collects temperature T s 、Humidity H s , vibration intensity V s Parameters, perform multi-source data fusion:

[0022]

[0023] Among them, W i is the fusion weight of the i-th sensor node, σ i is the variance of the i-th sensor, σ j is the variance of the jth sensor, d i is the Euclidean distance from the sensor to the monitoring point, α is the distance attenuation coefficient, n is the total number of sensor nodes, and e is the base of the natural logarithm;

[0024] When d i When it is greater than 3, the distance compensation mechanism is triggered;

[0025] Batch identification unit, using RFID tags to store drug codes C m , generate batch attribute data:

[0026] C m =HASH(Manufacturer Code P c l|Date of production p ||Serial Number S q ),

[0027] B k =CRC32(C m ||Storage temperature threshold T th ||Allowable vibration value V max ),

[0028] Among them, C m is the drug code, HASH(·) is the SHA-256 hash function, P cis the manufacturer code, D p is the production date, S q is the serial number, B k is the batch attribute data verification code,

[0029] CRC32(·) is a 32-bit cyclic redundancy check algorithm with the polynomial 0x04C11DB7, T th is the drug storage temperature threshold, V max is the maximum allowable vibration value;

[0030] The spatial positioning unit realizes three-dimensional coordinate solution through UWB base station networking:

[0031]

[0032] Among them, (x, y, z) is the three-dimensional coordinate of the drug storage location, t k is the arrival time of the kth base station signal, c is the speed of electromagnetic waves, (x k ,y k ,z k ) is the base station coordinate, k is the UWB base station number index, (x k+1 ,y k+1 ,z k+1 ) are the coordinates of the k+1th base station;

[0033] Error compensation is enabled when the positioning residual ε>0.1:

[0034]

[0035] Where Δ is , β is the environmental attenuation factor, and ln(·) is the natural logarithm function.

[0036] Preferably, the spatial mapping module receives the three-dimensional coordinates of the drug storage location and the environmental parameters from the data acquisition module, further comprising:

[0037] The coordinate registration unit converts the (x, y, z) coordinates output by the spatial positioning unit into the shelf-relative coordinate system:

[0038]

[0039] Among them, (x0, y0, z0) is the coordinate of the shelf origin, R is the rotation matrix, θ is the horizontal yaw angle, is the vertical pitch angle, ψ is the azimuth rotation angle, and (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system;

[0040] The interpolation calculation unit uses the improved inverse distance weighted method to generate the environmental parameters of the drug points:

[0041]

[0042] Among them, E p is the environmental parameter of the drug point, E i is the reading of the i-th sensor, ΔT i is the time decay term, γ is the temperature gradient coefficient, W i is the fusion weight of the i-th sensor node, n is the total number of sensor nodes;

[0043] Signal compensation unit, calculates the reliability index of environmental parameter mapping:

[0044]

[0045] Among them, Q is the reliability index of environment mapping, E max is the maximum reference value of the environmental parameter, is the historical mean value of the sensor, ζ is the humidity impact factor, H s is humidity, e is the base of natural logarithm;

[0046] When Q < 0.8, the following compensation mechanism is triggered:

[0047]

[0048] Among them, E p ′ is the environmental parameter value after compensation, λ is the compensation intensity coefficient, is the vertical environmental gradient, T s is the real-time temperature measurement value of the sensor, T th The temperature threshold for drug storage.

[0049] Preferably, the time sequence processing module receives the environment mapping data from the space mapping module, further comprising:

[0050] Data reconstruction unit, mapping the environment to data E p Generate a 3D tensor by slicing it into time windows:

[0051]

[0052] in, is the time window tensor, for Environmental parameters of drug points at all times, is the environmental parameter of the drug point at time t, is the history window length, M mask is the reliability mask matrix, is the tensor element-wise multiplication operator, is the environmental parameter of the drug point at time t-1;

[0053] When the reliability index Q of the calculation environment parameter mapping is less than 0.7, the corresponding element is set to zero;

[0054] Feature coupling unit, extracting spatiotemporal correlation features:

[0055]

[0056] Where F is the space-time coupling eigenvector, x i is the spatial dimension, M i is the spatial weight, Υ is the temperature change suppression factor, is the spatial direction partial derivative, is the second-order derivative of time, m is the characteristic dimension parameter, e is the base of natural logarithm, and ΔT is the temperature change;

[0057] GUR prediction unit, builds a bidirectional gating network to calculate the predicted value:

[0058]

[0059] Among them, τ is the prediction window, σ is the Sigmoid activation function, Q t is the reliability index, is the forward GUR hidden state at time t, is the reverse GRU hidden state at time t, GRU(·) is the gated recurrent unit calculation function, W f is the forward hidden state weight matrix, is the predicted value of the environmental parameters in the future τ time window, is the hidden state at time t+1, For W b is the reverse hidden state weight matrix, b is the bias vector, F t is the space-time coupling characteristic at time t.

[0060] Preferably, the time series processing module includes: constructing three-dimensional space-time tensor data, generating a data mask matrix using reliability indicators; extracting coupled eigenvectors containing spatial partial derivatives and time second-order derivatives; when the data missing rate exceeds a threshold, starting a historical data backtracking completion mechanism based on exponential decay.

[0061] Preferably, the risk assessment module receives the environmental parameter prediction value and the drug thermal sensitivity coefficient of the time sequence processing module, further comprising:

[0062] Deviation calculation unit, calculates the dynamic deviation between the temperature prediction value and the threshold:

[0063]

[0064] Among them, κ is the time decay coefficient, τ is the prediction window, D t is the integral value of temperature dynamic deviation, is the predicted temperature value at time t, T th(t) is the dynamic temperature threshold function, t is the current time point, t0 is the starting time of risk assessment, and dt is the time differential operator;

[0065] When D t >0.3 triggers a level 1 warning;

[0066] Rate of change analysis unit, assessing the risk of temperature change acceleration:

[0067]

[0068] Among them, ρ is the Sigmoid steepness coefficient, A t is the temperature change acceleration risk value, is the second-order time derivative of the temperature prediction value, and e is the natural logarithm base;

[0069] When A t >2, activate the second level warning;

[0070] Comprehensive assessment unit, integrating thermal sensitivity coefficient to calculate risk value:

[0071]

[0072] Among them, S c is the drug thermal sensitivity coefficient, ω1 and ω2 are weight factors, is the nonlinear amplification coefficient, R is the comprehensive risk value, D max is the maximum allowable deviation value, A max is the maximum allowable value of the rate of change.

[0073] Preferably, the decision execution module further includes matching the preset strategy library according to the comprehensive risk value of the risk assessment module:

[0074] The strategy matching unit uses fuzzy decision tree to achieve risk-strategy mapping:

[0075]

[0076] Among them, P k is the serial number of the kth control strategy in the strategy set Ω, υ is the decision steepness coefficient, θ p is the activation threshold of policy p, W p is the policy priority weight, D p is the strategy execution delay time, D max is the maximum allowed strategy execution delay time, R is the comprehensive risk value;

[0077] Instruction optimization unit, generates device control quantity:

[0078]

[0079] Among them, C cool is the output power control quantity of the refrigeration unit, V AVG is the driving speed of the AGV, R max Design the maximum risk threshold for the system, C max is the maximum rated power of the refrigeration unit, ΔT is the real-time temperature deviation, Q is the environmental mapping reliability index, V max is the maximum safe speed of AVG, β is the temperature adjustment coefficient, η is the speed adjustment factor, Q min is the minimum reliability threshold, θ v is the speed adjustment threshold;

[0080] Data encapsulation unit, generates traceability data packet:

[0081]

[0082] Among them, τ is the prediction window, || is data splicing, SHA256(·) is the hash algorithm, C m is the drug code, (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system, is the predicted temperature value at time t, H sig is the secure hash value of the data packet, R(t) is the time series risk function, is the value collected by the i-th vibration sensor, T th The temperature threshold for drug storage.

[0083] The present invention provides a logistics warehousing monitoring system based on artificial intelligence. It has the following beneficial effects:

[0084] 1. This invention establishes a dynamic association model between drug batches and environmental sensors, combining spatial interpolation algorithms with three-dimensional coordinate mapping technology to accurately link abnormal events to drug batches. When an environmental anomaly occurs in a certain area, the system can quickly locate the affected drug batches and generate a traceability map. Compared to existing solutions that disconnect environmental data from drug locations, this approach addresses the issues of inefficient abnormality tracing and the perishable nature of high-value drugs.

[0085] 2. The present invention adopts the GUR time series prediction network and the thermal sensitivity coefficient dynamic risk assessment model. By analyzing the time series characteristics, it predicts the risk of environmental parameter crossing the boundary in advance and generates a gradient response strategy. Compared with the passive threshold alarm mechanism in the existing technology, it solves the problem of early warning lag, reserves buffer time for refrigeration adjustment and drug transfer operations, and avoids irreversible loss of biological preparations caused by the destruction of the constant temperature environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a system diagram of the present invention. DETAILED DESCRIPTION

[0087] To help those skilled in the art understand the present invention, the following will provide a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only partial embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0088] The present invention is described in detail below with reference to the accompanying drawings:

[0089] Example:

[0090] Please see the attached Figure 1 The embodiment of the present invention provides a logistics and warehousing monitoring system based on artificial intelligence, and the logistics and warehousing monitoring system includes:

[0091] The data acquisition module consists of an environmental sensor array, a drug batch identification device, and a spatial positioning device deployed on the warehouse's three-dimensional shelves;

[0092] The multi-parameter environmental sensor array is deployed in a cellular topology with node spacing less than 1.5 meters. It includes temperature sensors, humidity sensors, and vibration sensors, and has an operating temperature range of -40°C to 60°C.

[0093] The environmental sensor array collects temperature, humidity, and vibration parameters in real time. The drug batch identification unit stores the drug's unique code and batch attribute data. The spatial positioning device uses ultra-wideband technology to obtain the three-dimensional coordinates of the drug storage location.

[0094] The batch identification unit uses a dual identification of ultra-high frequency radio frequency identification tags and low-temperature resistant QR codes. The lower limit of the tag's operating temperature is -40°C, and the QR code printing material can withstand a frosting environment of -30°C.

[0095] The environmental sensing unit consists of a spatial hexahedron grid of sensor nodes, each of which collects temperature T s 、Humidity H s , vibration intensity V s Parameters, perform multi-source data fusion:

[0096]

[0097] Among them, W i is the fusion weight of the i-th sensor node, σ i is the variance of the i-th sensor, σ j is the variance of the jth sensor, d i is the Euclidean distance from the sensor to the monitoring point, α is the distance attenuation coefficient, n is the total number of sensor nodes, and e is the base of the natural logarithm;

[0098] When di When it is greater than 3, the distance compensation mechanism is triggered;

[0099] Batch identification unit, using RFID tags to store drug codes C m , generate batch attribute data:

[0100] C m =HASH(Manufacturer Code P c l|Date of production p ||Serial Number S q ),

[0101] B k =CRC32(C m ||Storage temperature threshold T th ||Allowable vibration value V max ),

[0102] Among them, C m is the drug code, HASH(·) is the SHA-256 hash function, P c is the manufacturer code, D p is the production date, S q is the serial number, B k is the batch attribute data verification code,

[0103] CRC32(·) is a 32-bit cyclic redundancy check algorithm with the polynomial 0x04C11DB7, T th is the drug storage temperature threshold, V max is the maximum allowable vibration value;

[0104] The spatial positioning unit realizes three-dimensional coordinate solution through UWB base station networking:

[0105]

[0106] Among them, (x, y, z) is the three-dimensional coordinate of the drug storage location, t k is the arrival time of the kth base station signal, c is the speed of electromagnetic waves, (x k ,y k ,z k ) is the base station coordinate, k is the UWB base station number index, (x k+1 ,y k+1 ,z k+1 ) are the coordinates of the k+1th base station;

[0107] Error compensation is enabled when the positioning residual ε>0.1:

[0108]

[0109] Where Δ is, β is the environmental attenuation factor, and ln(·) is the natural logarithm function;

[0110] The spatial mapping module receives the three-dimensional coordinates of the drug storage location and environmental parameters from the data acquisition module, establishes a drug-sensor association model based on a spatial interpolation algorithm, and outputs environmental mapping data with distance weights and signal compensation coefficients;

[0111] The coordinate registration unit converts the (x, y, z) coordinates output by the spatial positioning unit into the shelf-relative coordinate system:

[0112]

[0113] Among them, (x0, y0, z0) is the coordinate of the shelf origin, R is the rotation matrix, θ is the horizontal yaw angle, is the vertical pitch angle, ψ is the azimuth rotation angle, and (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system;

[0114] The interpolation calculation unit uses the improved inverse distance weighted method to generate the environmental parameters of the drug points:

[0115]

[0116] Among them, E p is the environmental parameter of the drug point, E i is the reading of the i-th sensor, ΔT i is the time decay term, γ is the temperature gradient coefficient, W i is the fusion weight of the i-th sensor node, n is the total number of sensor nodes;

[0117] Signal compensation unit, calculates the reliability index of environmental parameter mapping:

[0118]

[0119] Among them, Q is the reliability index of environment mapping, E max is the maximum reference value of the environmental parameter, is the historical mean value of the sensor, ζ is the humidity impact factor, H s is humidity, e is the base of natural logarithm;

[0120] When Q < 0.8, the following compensation mechanism is triggered:

[0121]

[0122] Among them, E p ′ is the environmental parameter value after compensation, λ is the compensation intensity coefficient, is the vertical environmental gradient, T s is the real-time temperature measurement value of the sensor, T th The temperature threshold for drug storage;

[0123] The time series processing module receives the environmental mapping data from the spatial mapping module, processes the time series data through the gated recurrent unit network, and outputs the predicted values ​​of the environmental parameters in the future time window;

[0124] Data reconstruction unit, mapping the environment to data E p Generate a 3D tensor by slicing it into time windows:

[0125]

[0126] in, is the time window tensor, for Environmental parameters of drug points at all times, is the environmental parameter of the drug point at time t, is the history window length, M mask is the reliability mask matrix, is the tensor element-wise multiplication operator, is the environmental parameter of the drug point at time t-1;

[0127] When the reliability index Q of the calculation environment parameter mapping is less than 0.7, the corresponding element is set to zero;

[0128] Feature coupling unit, extracting spatiotemporal correlation features:

[0129]

[0130] Where F is the space-time coupling eigenvector, x i is the spatial dimension, M i is the spatial weight, Υ is the temperature change suppression factor, is the spatial direction partial derivative, is the second-order derivative of time, m is the characteristic dimension parameter, e is the base of natural logarithm, and ΔT is the temperature change;

[0131] GUR prediction unit, builds a bidirectional gating network to calculate the predicted value:

[0132]

[0133] Among them, τ is the prediction window, σ is the Sigmoid activation function, Q t is the reliability index, is the forward GUR hidden state at time t, is the reverse GRU hidden state at time t, GRU(·) is the gated recurrent unit calculation function, W f is the forward hidden state weight matrix, is the predicted value of the environmental parameters in the future τ time window, is the hidden state at time t+1, For Wb is the reverse hidden state weight matrix, b is the bias vector, F t is the spatiotemporal coupling characteristic at time t;

[0134] The time series processing module includes: constructing three-dimensional space-time tensor data, generating a data mask matrix using reliability indicators; extracting coupled eigenvectors containing spatial partial derivatives and temporal second-order derivatives; and activating a historical data backtracking and completion mechanism based on exponential decay when the data missing rate exceeds a threshold.

[0135] The risk assessment module receives the environmental parameter prediction value from the time sequence processing module and the preset drug thermal sensitivity coefficient, and calculates the comprehensive risk value including the temperature deviation and change rate;

[0136] Deviation calculation unit, calculates the dynamic deviation between the temperature prediction value and the threshold:

[0137]

[0138] Among them, κ is the time decay coefficient, τ is the prediction window, D t is the integral value of temperature dynamic deviation, is the predicted temperature value at time t, T th (t) is the dynamic temperature threshold function, t is the current time point, t0 is the starting time of risk assessment, and dt is the time differential operator;

[0139] When D t >0.3 triggers a level 1 warning;

[0140] Rate of change analysis unit, assessing the risk of temperature change acceleration:

[0141]

[0142] Among them, ρ is the Sigmoid steepness coefficient, A t is the temperature change acceleration risk value, is the second-order time derivative of the temperature prediction value, and e is the natural logarithm base;

[0143] When A t >2, activate the second level warning;

[0144] Comprehensive assessment unit, integrating thermal sensitivity coefficient to calculate risk value:

[0145]

[0146] Among them, S c is the drug thermal sensitivity coefficient, ω1 and ω2 are weight factors, is the nonlinear amplification coefficient, R is the comprehensive risk value, D max is the maximum allowable deviation value, A maxis the maximum allowable value of the rate of change;

[0147] The decision execution module matches the preset strategy library according to the comprehensive risk value of the risk assessment module, sends control instructions to the refrigeration unit, automatic guided vehicle, and alarm device, and generates a traceability data package containing drug codes, location coordinates, and environmental curves;

[0148] The strategy matching unit uses fuzzy decision tree to achieve risk-strategy mapping:

[0149]

[0150] Among them, P k is the serial number of the kth control strategy in the strategy set Ω, υ is the decision steepness coefficient, θ p is the activation threshold of policy p, W p is the policy priority weight, D p is the strategy execution delay time, D max is the maximum allowed strategy execution delay time, R is the comprehensive risk value;

[0151] Instruction optimization unit, generates device control quantity:

[0152]

[0153] Among them, C cool is the output power control quantity of the refrigeration unit, V AVG is the driving speed of the AGV, R max Design the maximum risk threshold for the system, C max is the maximum rated power of the refrigeration unit, ΔT is the real-time temperature deviation, Q is the environmental mapping reliability index, V max is the maximum safe speed of AVG, β is the temperature adjustment coefficient, η is the speed adjustment factor, Q min is the minimum reliability threshold, θ v is the speed adjustment threshold;

[0154] Data encapsulation unit, generates traceability data packet:

[0155]

[0156] Among them, τ is the prediction window, || is data splicing, SHA256(·) is the hash algorithm, C m is the drug code, (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system, is the predicted temperature value at time t, H sig is the secure hash value of the data packet, R(t) is the time series risk function, is the value collected by the i-th vibration sensor, T th The temperature threshold for drug storage;

[0157] The feedback calibration module receives the control instruction execution status data and traceability data packet from the decision execution module, calculates the deviation between the actual effect of the thermal sensitivity coefficient and the expected value, and generates the drug thermal sensitivity coefficient correction parameter.

[0158] Environmental sensor deployment: Multi-parameter sensor nodes are deployed in a cellular topology on the warehouse's three-dimensional shelves, with node spacing less than 1.5 meters, forming a hexahedral monitoring grid. Each node integrates temperature, humidity, and vibration sensors, operating in temperatures ranging from -50°C to 50°C. Sensor data is uploaded via the LoRa wireless network, using a time-delayed multiple access (TDMA) mechanism to avoid signal conflicts.

[0159] Drug batch identification: Drug packaging is dually labeled with an ultra-high frequency (UHF) RFID tag and a low-temperature-resistant QR code. The RFID tag has an embedded metal-resistant antenna and operates at a temperature lower than -40°C. It stores the drug's unique identifier (manufacturer code, production date, and serial number). The QR code is printed with frost-resistant ink and can be read by industrial scanners in temperatures as low as -30°C. The dual identification provides redundancy, ensuring data readability even under extreme conditions.

[0160] Spatial positioning: Ultra-wideband positioning base stations are deployed in the shelf area, using the TDOA algorithm to calculate the three-dimensional coordinates of the drug storage location. Positioning accuracy reaches ±10cm, and dynamic compensation for signal attenuation caused by metal shelves is supported. Coordinate data is aligned with the shelf's relative coordinate system.

[0161] Coordinate registration and interpolation: The absolute coordinates output by the UWB are converted to relative coordinates based on the shelf origin, and rotation matrix correction is performed to account for shelf tilt. An improved inverse distance weighted method is used to fuse data from neighboring sensors to calculate environmental parameters for the drug point. A time decay factor is introduced into the interpolation process to reduce the weight of historical data and enhance real-time performance.

[0162] Signal Reliability Assessment: Dynamically calculates the reliability of environmental parameters based on the sensor's historical data average, environmental gradient changes, and humidity. When the reliability falls below a threshold, a vertical gradient compensation mechanism is activated, and the measured value is corrected based on the drug storage temperature threshold to avoid misjudgments due to sensor failure or signal interference.

[0163] Spatiotemporal Data Reconstruction: Environmental parameters are sliced ​​into time windows to construct a three-dimensional tensor containing spatial location, time series, and reliability masks. The mask matrix automatically filters out low-reliability data, and missing data is supplemented with exponentially weighted historical values ​​to ensure input integrity.

[0164] Feature Extraction and Prediction: A bidirectional gated recurrent unit network is used to extract spatiotemporal correlation features, including spatial temperature gradients and temporal acceleration. The network outputs predicted environmental parameters for the next 15 minutes. The abnormal fluctuation detection module monitors hidden states in real time and triggers data backtesting when the prediction deviation exceeds a threshold.

[0165] Dynamic Deviation Analysis: Calculates the cumulative deviation between the predicted temperature and the drug storage threshold, introducing a time decay factor to weight recent deviations. When the deviation integral exceeds the limit, a Level 1 alert is generated and the location of high-risk drugs is marked.

[0166] Rate of change risk assessment: Analyze the acceleration of temperature changes to identify sudden rise or fall trends. Combined with the drug's thermal sensitivity coefficient, calculate the weight of the rate of change's impact on drug stability and initiate an emergency response when a Level 2 alert is triggered.

[0167] Comprehensive risk decision-making: This system integrates deviation, rate of change, and drug characteristics to calculate a comprehensive risk value. It uses a fuzzy decision tree to match a pre-set strategy library. For example, it can adjust local cooling power when the risk is low, dispatch AGVs to transfer drugs when the risk is medium, and activate audible and visual alarms and notify human intervention when the risk is high.

[0168] Control instruction generation, generate differentiated instructions based on risk level:

[0169] Refrigeration unit: Adjust the air supply volume according to the temperature deviation ratio, giving priority to high-risk areas.

[0170] AGV scheduling: Plan the optimal path to transfer high-risk drugs and avoid congested shelf areas.

[0171] Alarm system: hierarchical triggering of buzzer, SMS notification and platform pop-up alarm.

[0172] Traceability data packaging: Generates a data package containing drug codes, location coordinates, environmental profiles, and operation logs, encrypted using the SHA-256 algorithm. Data packages are archived by timestamp and supported on blockchain for auditing purposes.

[0173] Dynamic correction of thermal coefficients: Compare drug stability data before and after the control strategy is implemented to calculate the actual effect of the thermal coefficient. When the deviation exceeds 5%, the Bayesian optimization algorithm is activated to adjust the coefficient weight and update the risk assessment model.

[0174] Strategy library self-learning: Records the execution results of historical control strategies and builds a case library. When encountering new drugs or environmental patterns, similarity matching is used to recommend the historically optimal strategy, and reinforcement learning is used to iterate and optimize.

[0175] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A logistics warehousing monitoring system based on artificial intelligence, characterized in that: The logistics warehousing monitoring system includes: The data acquisition module consists of an environmental sensor array, a drug batch identification device, and a spatial positioning device deployed on the warehouse's three-dimensional shelves; The spatial mapping module receives the three-dimensional coordinates of the drug storage location and environmental parameters from the data acquisition module, establishes a drug-sensor association model based on a spatial interpolation algorithm, and outputs environmental mapping data with distance weights and signal compensation coefficients; The time series processing module receives the environmental mapping data from the spatial mapping module, processes the time series data through the gated recurrent unit network, and outputs the predicted values ​​of the environmental parameters in the future time window; The risk assessment module receives the environmental parameter prediction value from the time sequence processing module and the preset drug thermal sensitivity coefficient, and calculates the comprehensive risk value including the temperature deviation and change rate; The decision execution module matches the preset strategy library according to the comprehensive risk value of the risk assessment module, sends control instructions to the refrigeration unit, automatic guided vehicle, and alarm device, and generates a traceability data package containing drug codes, location coordinates, and environmental curves; The feedback calibration module receives the control instruction execution status data and traceability data packet from the decision execution module, calculates the deviation between the actual effect of the thermal sensitivity coefficient and the expected value, and generates the drug thermal sensitivity coefficient correction parameter.

2. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The multi-parameter environmental sensor array is deployed in a cellular topology with a node spacing of less than 1.5 meters. It includes temperature sensors, humidity sensors, and vibration sensors, and has an operating temperature range of -40°C to 60°C.

3. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The environmental sensor array collects temperature, humidity, and vibration parameters in real time, the drug batch identification unit stores the drug's unique code and batch attribute data, and the spatial positioning device obtains the three-dimensional coordinates of the drug storage location through ultra-wideband technology.

4. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The batch identification unit adopts dual identification of ultra-high frequency radio frequency identification tag and low-temperature resistant QR code. The lower limit of the label working temperature is -40°C, and the QR code printing material can withstand a frosting environment of -30°C.

5. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The data acquisition module is composed of an environmental sensor array, a drug batch identification device, and a spatial positioning device deployed on the warehouse three-dimensional shelves, and further includes: The environmental sensing unit consists of a spatial hexahedron grid of sensor nodes, each of which collects temperature T s 、Humidity H s , vibration intensity V s Parameters, perform multi-source data fusion: Among them, W i is the fusion weight of the i-th sensor node, σ i is the variance of the i-th sensor, σ j is the variance of the jth sensor, d i is the Euclidean distance from the sensor to the monitoring point, α is the distance attenuation coefficient, n is the total number of sensor nodes, and e is the base of the natural logarithm; When d i When it is greater than 3, the distance compensation mechanism is triggered; Batch identification unit, using RFID tags to store drug codes C m , generate batch attribute data: C m =HASH(Manufacturer Code P c l|Date of production p ||Serial Number S q ), B k =CRC32(C m ||Storage temperature threshold T th ||Allowable vibration value V max ), Among them, C m is the drug code, HASH(·) is the SHA-256 hash function, P c is the manufacturer code, D p is the production date, S q is the serial number, B k is the batch attribute data verification code, CRC32(·) is a 32-bit cyclic redundancy check algorithm with the polynomial 0x04C11DB7, T th is the drug storage temperature threshold, V max is the maximum allowable vibration value; The spatial positioning unit realizes three-dimensional coordinate solution through UWB base station networking: Among them, (x, y, z) is the three-dimensional coordinate of the drug storage location, t k is the arrival time of the kth base station signal, c is the speed of electromagnetic waves, (x k ,y k ,z k ) is the base station coordinate, k is the UWB base station number index, (x k+1 ,y k+1 ,z k+1 ) are the coordinates of the k+1th base station; Error compensation is enabled when the positioning residual ε>0.1: Where Δ is , β is the environmental attenuation factor, and ln(·) is the natural logarithm function.

6. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The spatial mapping module receives the three-dimensional coordinates of the drug storage location and the environmental parameters from the data acquisition module, further comprising: The coordinate registration unit converts the (x, y, z) coordinates output by the spatial positioning unit into the shelf-relative coordinate system: Among them, (x0, y0, z0) is the coordinate of the shelf origin, R is the rotation matrix, θ is the horizontal yaw angle, is the vertical pitch angle, ψ is the azimuth rotation angle, and (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system; The interpolation calculation unit uses the improved inverse distance weighted method to generate the environmental parameters of the drug points: Among them, E p is the environmental parameter of the drug point, E i is the reading of the i-th sensor, ΔT i is the time decay term, γ is the temperature gradient coefficient, W i is the fusion weight of the i-th sensor node, n is the total number of sensor nodes; Signal compensation unit, calculates the reliability index of environmental parameter mapping: Among them, Q is the reliability index of environment mapping, E max is the maximum reference value of the environmental parameter, is the historical mean value of the sensor, ζ is the humidity impact factor, H s is humidity, e is the base of natural logarithm; When Q < 0.8, the following compensation mechanism is triggered: Among them, E p ′ is the environmental parameter value after compensation, λ is the compensation intensity coefficient, is the vertical environmental gradient, T s is the real-time temperature measurement value of the sensor, T th The temperature threshold for drug storage.

7. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The time sequence processing module receives the environment mapping data of the space mapping module, further comprising: Data reconstruction unit, mapping the environment to data E p Generate a 3D tensor by slicing it into time windows: in, is the time window tensor, for Environmental parameters of drug points at all times, is the environmental parameter of the drug point at time t, is the history window length, M mask is the reliability mask matrix, is the tensor element-wise multiplication operator, is the environmental parameter of the drug point at time t-1; When the reliability index Q of the calculation environment parameter mapping is less than 0.7, the corresponding element is set to zero; Feature coupling unit, extracting spatiotemporal correlation features: Where F is the space-time coupling eigenvector, x i is the spatial dimension, M i is the spatial weight, Υ is the temperature change suppression factor, is the spatial direction partial derivative, is the second-order derivative of time, m is the characteristic dimension parameter, e is the base of natural logarithm, and ΔT is the temperature change; GUR prediction unit, builds a bidirectional gating network to calculate the predicted value: Among them, τ is the prediction window, σ is the Sigmoid activation function, Q t is the reliability index, is the forward GUR hidden state at time t, is the reverse GRU hidden state at time t, GRU(·) is the gated recurrent unit calculation function, W f is the forward hidden state weight matrix, is the predicted value of the environmental parameters in the future τ time window, is the hidden state at time t+1, For W b is the reverse hidden state weight matrix, b is the bias vector, F t is the space-time coupling characteristic at time t.

8. The artificial intelligence-based logistics warehousing monitoring system according to claim 7, characterized in that: The time series processing module includes: constructing three-dimensional spatiotemporal tensor data and generating a data mask matrix using reliability indicators; extracting coupled eigenvectors containing spatial partial derivatives and temporal second-order derivatives; and when the data missing rate exceeds a threshold, initiating a historical data backtracking and completion mechanism based on exponential decay.

9. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: The risk assessment module receives the environmental parameter prediction value and the drug thermal sensitivity coefficient of the time sequence processing module, further comprising: Deviation calculation unit, calculates the dynamic deviation between the temperature prediction value and the threshold: Among them, κ is the time decay coefficient, τ is the prediction window, D t is the integral value of temperature dynamic deviation, is the predicted temperature value at time t, T th (t) is the dynamic temperature threshold function, t is the current time point, t0 is the starting time of risk assessment, and dt is the time differential operator; When D t >0.3 triggers a level 1 warning; Rate of change analysis unit, assessing the risk of temperature change acceleration: Among them, ρ is the Sigmoid steepness coefficient, A t is the temperature change acceleration risk value, is the second-order time derivative of the temperature prediction value, and e is the natural logarithm base; When A t >2, activate the second level warning; Comprehensive assessment unit, integrating thermal sensitivity coefficient to calculate risk value: Among them, S c is the drug thermal sensitivity coefficient, ω1 and ω2 are weight factors, is the nonlinear amplification coefficient, R is the comprehensive risk value, D max is the maximum allowable deviation value, A max is the maximum allowable value of the rate of change.

10. The artificial intelligence-based logistics warehousing monitoring system according to claim 1, characterized in that: In the decision execution module, matching the preset strategy library according to the comprehensive risk value of the risk assessment module further includes: The strategy matching unit uses fuzzy decision tree to achieve risk-strategy mapping: Among them, P k is the serial number of the kth control strategy in the strategy set Ω, υ is the decision steepness coefficient, θ p is the activation threshold of policy p, W p is the policy priority weight, D p is the strategy execution delay time, D max is the maximum allowed strategy execution delay time, R is the comprehensive risk value; Instruction optimization unit, generates device control quantity: Among them, C cool is the output power control quantity of the refrigeration unit, V AVG is the driving speed of the AGV, R max Design the maximum risk threshold for the system, C max is the maximum rated power of the refrigeration unit, ΔT is the real-time temperature deviation, Q is the environmental mapping reliability index, V max is the maximum safe speed of AVG, β is the temperature adjustment coefficient, η is the speed adjustment factor, Q min is the minimum reliability threshold, θ v is the speed adjustment threshold; Data encapsulation unit, generates traceability data packet: Among them, τ is the prediction window, || is data splicing, SHA256(·) is the hash algorithm, C m is the drug code, (X, Y, Z) is the three-dimensional coordinate of the shelf relative coordinate system, is the predicted temperature value at time t, H sig is the secure hash value of the data packet, R(t) is the time series risk function, is the value collected by the i-th vibration sensor, T th The temperature threshold for drug storage.

Images