Dynamic material sorting system based on stereoscopic warehouse

Abstract

The invention relates to the technical field of stereoscopic warehouse control, and discloses a dynamic material sorting system based on a stereoscopic warehouse, which comprises a graph node mapping unit, a state component processing engine, an addressing scheduling center and a material transfer control module. Calling neighborhood association node data, calculating a neighborhood interference compensation weight based on the logic step number distance, and further correcting the attenuation curve to generate a real-time state tensor containing a maturity component; the addressing scheduling center compares the real-time state tensor with the sorting threshold value and issues an addressing control sequence containing a target goods allocation logic address, a weight diffusion mechanism is used for eliminating hidden state interference errors induced by space proximity, and the sorting matching precision and the scheduling efficiency are improved on the premise that physical sensing hardware is not added.

Description

Technical Field

[0001] This invention relates to the field of automated warehouse control technology, and more specifically, to a dynamic material sorting system based on an automated warehouse. Background Technology

[0002] Current automated storage and retrieval systems (AS / RS) utilize stacker cranes and discrete location coding systems to manage material flow. The mainstream control logic allocates materials to specific three-dimensional spatial coordinates based on their entry order and generates scheduling instructions accordingly. However, for high-temperature rolled parts in scenarios such as metallurgical rolling, the physical properties of the materials change significantly with residence time. Existing technologies treat the materials as rigid bodies with a constant state, and scheduling decisions rely solely on spatial coordinates or entry order. This static management model ignores the correlation between the evolution of the material state and the downstream rolling rhythm, resulting in the sorting instructions output by the system failing to meet the process accuracy requirements and causing energy waste in the production line.

[0003] To address the issue of state deviation, related technologies have attempted to introduce a decay function based on residence time. However, in dense storage environments with compact layouts, adjacent materials generate significant thermal interference. This independent calculation logic, based on pre-defined data silos, cannot capture the coupling effects of neighboring nodes, leading to cumulative offset errors in the system's maintained state data as the storage cycle lengthens. Adding more temperature sensors significantly increases system hardware costs and maintenance complexity. Some scheduling systems introduce software-level multimodal feature fusion to improve management accuracy. For example, Chinese invention patent application CN121639107A discloses an intelligent warehouse management method, system, electronic device, and medium based on multimodal feature fusion. It determines the quality deterioration level of goods and matches storage areas through image and weight feature fusion. When an area is saturated, it dynamically expands by modifying logical tags. However, when dealing with complex thermodynamic evolution conditions, this logic is a static mapping extension, fixing the instantaneous quality status of materials upon entry into the warehouse. It fails to establish a dynamic tracking model for the continuous fluctuations in physical properties during material residence. The system treats each storage location as a logical island and lacks the ability to calculate the field coupling effects of adjacent high-temperature materials in dense storage environments.

[0004] Therefore, the technical problem to be solved by this invention is how to reconstruct the logical topology of an automated warehousing system, eliminate the interference of local environmental disturbances on state calculation by establishing correlation constraints between storage nodes while maintaining the hardware architecture, and integrate spatial distance and physical state parameters to output sorting instructions. Summary of the Invention

[0005] This invention provides a dynamic material sorting system based on an automated warehouse, the system comprising: The graph node mapping unit is used to construct a virtual topology graph in memory based on the topology relationship of the storage locations recorded in the memory, and to obtain the time-varying state parameters of the materials to be sorted in each storage location and the corresponding predicted decay curves. The state component processing engine is used to respond to the update command of the time-varying state parameters of any node, retrieve the dataset of associated nodes in the preset neighborhood of the target node in the virtual topology map, and calculate the neighborhood interference compensation weight for the target node based on the logical step distance between the associated node dataset and the target node, so as to dynamically correct the predicted decay curve and generate the real-time state tensor of the material to be sorted; wherein, the real-time state tensor contains a maturity component that characterizes the degree of physical evolution of the material. The addressing and scheduling center is used to acquire sorting trigger signals and calculate the deviation between the real-time state tensor and the preset sorting threshold. When the deviation meets the preset convergence threshold condition, it generates an addressing control sequence containing the logical address of the target cargo location. The material transfer control module communicates with the addressing and scheduling center and is used to send addressing control sequences to external transfer mechanisms to drive them to trigger extraction actions for the materials to be sorted.

[0006] Preferably, the addressing and scheduling center also includes a joint cost evaluation component; the joint cost evaluation component is used to obtain real-time location feedback data of external transfer agencies during the addressing control sequence generation process, and calculate the normalized distance cost of external transfer agencies reaching candidate cargo locations; the joint cost evaluation component establishes sorting priority weights by performing negative correlation weighting operations on the normalized distance cost and the maturity component in the real-time state tensor, so as to select the logical node with the highest sorting priority weight as the final scheduling target.

[0007] Preferably, before the addressing control sequence is output by the addressing scheduling center, the path verification module retrieves the state-locked nodes located on the access path of the final scheduling target in the virtual topology map; if there are logical nodes in the access path that are in the state of being occupied, the addressing scheduling center is triggered to logically reconstruct the addressing control sequence to generate avoidance constraint instructions and synchronize them to the material transfer control module.

[0008] Preferably, when the state component processing engine calculates the neighborhood interference compensation weight, the value of the neighborhood interference compensation weight has an inversely proportional monotonic relationship with the number of logical steps between the target node and the associated node, so as to quantitatively characterize the interference intensity of the time-varying state parameter evolution of the spatially adjacent region of the virtual topology map for the material to be sorted.

[0009] Preferably, the system also includes a state closed-loop correction module; the state closed-loop correction module is used to obtain physical sampling data of the corresponding logical address when the material transfer control module arrives at the target location, and calculate the prediction deviation between the physical sampling data and the predicted value of the real-time state tensor; when the prediction deviation exceeds 5%, the addressing scheduling center triggers the addressing sequence overwrite operation and adjusts the global attenuation parameter in the state component processing engine in reverse.

[0010] Preferably, the state closed-loop correction module uses a recursive algorithm to dynamically approximate the global decay parameter in order to achieve online compensation for the parameter temperature drift induced by fluctuations in the external environment of the system, and ensure that the prediction accuracy of the real-time state tensor within a 10ms sampling period is not less than 98%.

[0011] Preferably, the virtual topology map maps the physical addressing space of the memory through a multi-dimensional matrix structure; each data node in the memory corresponds to a unique logical address number and is associated with an entry timestamp and an initial feature vector; the graph node mapping unit maintains the node occupancy characteristics of the virtual topology map by monitoring the status feedback of the material transfer control module.

[0012] Preferably, the state component processing engine uses an asynchronous scheduling mode to process the real-time state tensors of each node; the refresh frequency of the asynchronous scheduling mode is aligned with the synchronization pulse frequency of the downstream consumption end, so that the phase deviation between the issuance time of the addressing control sequence and the material consumption cycle is less than 50ms.

[0013] Preferably, the memory stores a feature benchmark library based on the evolution of historical attributes of materials; the state component processing engine identifies the stage-specific abrupt change points of the time-varying state parameters of the materials to be sorted by calculating the gradient change rate of the real-time state tensor in the virtual topology map, and improves the extraction response level of the addressing scheduling center for the corresponding logical address accordingly.

[0014] The embodiments of the present invention have at least the following beneficial effects: 1. In the dynamic material sorting of automated warehouses, this invention eliminates the data silo effect between storage nodes and builds field perception capabilities. Traditional control logic treats each storage location as an independent data unit, which cannot map the local thermal field interference generated by high-temperature materials in dense storage environments. This invention introduces compensation factors for adjacent materials in the physical neighborhood during the state calculation process, so that the attenuation gradient of the target material can be corrected in real time according to the heat release state of the surrounding environment. This shift from single-point calculation to neighborhood collaborative calculation ensures that the material state variables maintained inside the control system are logically aligned with the complex physical evolution process, eliminating calculation drift caused by environmental interference.

[0015] 2. To achieve decoupling control between physical location and logical attributes of materials, this invention constructs a virtual topology map containing multiple logical sub-regions in memory for materials with non-stationary evolution characteristics. This drives the logical tags of materials to perform cross-region transitions in the map according to the evolution of their computational state. While the materials remain physically stationary and locked, their weight in the sorting sequence dynamically shifts with the maturity of their actual physical attributes. This mechanism ensures that the triggering of sorting instructions directly depends on the real-time state characteristics of the materials, rather than the simple order of entry or static coordinates, effectively solving the matching mismatch problem in the sorting process of special materials.

[0016] 3. Achieving comprehensive optimization of stacker crane addressing path and material state matching: This invention establishes a joint evaluation mechanism that integrates spatial distance cost and material state deviation cost. By calculating the joint cost index when a sorting request is triggered, it achieves collaborative optimization of the stacker crane's mechanical action cost and the degree of achievement of process requirements. When the controller outputs the extraction command, it automatically suppresses invalid addressing operations with excessive spatial span by weighted evaluation of spatial Manhattan distance and state maturity. This improvement enables the system's output control sequence to shorten the operating path of the actuator to the greatest extent while meeting the downstream process temperature threshold constraints, thereby improving the overall command response efficiency of the automated warehouse system. Attached Figure Description

[0017] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings, in which several embodiments of the invention are illustrated by way of example and not limitation, wherein: Figure 1 This is a flowchart of the global state calculation and scheduling control of the dynamic material sorting system of the present invention; Figure 2 This is a diagram showing the interference avoidance and dynamic path reconstruction during the addressing process of the sorting system of the present invention. Detailed Implementation

[0018] The principles and spirit of the present invention will now be described with reference to several exemplary embodiments in conjunction with the accompanying drawings. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0019] A dynamic material sorting system based on an automated warehouse, the system comprising: The graph node mapping unit is used to construct a virtual topology graph in memory based on the topology relationship of the storage locations recorded in the memory, and to obtain the time-varying state parameters of the materials to be sorted in each storage location and the corresponding predicted decay curves. The state component processing engine is used to respond to the update command of the time-varying state parameters of any node, retrieve the dataset of associated nodes in the preset neighborhood of the target node in the virtual topology map, and calculate the neighborhood interference compensation weight for the target node based on the logical step distance between the associated node dataset and the target node, so as to dynamically correct the predicted decay curve and generate the real-time state tensor of the material to be sorted; wherein, the real-time state tensor contains a maturity component that characterizes the degree of physical evolution of the material. The addressing and scheduling center is used to acquire sorting trigger signals and calculate the deviation between the real-time state tensor and the preset sorting threshold. When the deviation meets the preset convergence threshold condition, it generates an addressing control sequence containing the logical address of the target cargo location. The material transfer control module communicates with the addressing and scheduling center and is used to send addressing control sequences to external transfer mechanisms to drive them to trigger extraction actions for the materials to be sorted.

[0020] Preferably, the addressing and scheduling center also includes a joint cost evaluation component; the joint cost evaluation component is used to obtain real-time location feedback data of external transfer agencies during the addressing control sequence generation process, and calculate the normalized distance cost of external transfer agencies reaching candidate cargo locations; the joint cost evaluation component establishes sorting priority weights by performing negative correlation weighting operations on the normalized distance cost and the maturity component in the real-time state tensor, so as to select the logical node with the highest sorting priority weight as the final scheduling target.

[0021] Preferably, before the addressing control sequence is output by the addressing scheduling center, the path verification module retrieves the state-locked nodes located on the access path of the final scheduling target in the virtual topology map; if there are logical nodes in the access path that are in the state of being occupied, the addressing scheduling center is triggered to logically reconstruct the addressing control sequence to generate avoidance constraint instructions and synchronize them to the material transfer control module.

[0022] Preferably, when the state component processing engine calculates the neighborhood interference compensation weight, the value of the neighborhood interference compensation weight has an inversely proportional monotonic relationship with the number of logical steps between the target node and the associated node, so as to quantitatively characterize the interference intensity of the time-varying state parameter evolution of the spatially adjacent region of the virtual topology map for the material to be sorted.

[0023] Preferably, the system also includes a state closed-loop correction module; the state closed-loop correction module is used to obtain physical sampling data of the corresponding logical address when the material transfer control module arrives at the target location, and calculate the prediction deviation between the physical sampling data and the predicted value of the real-time state tensor; when the prediction deviation exceeds 5%, the addressing scheduling center triggers the addressing sequence overwrite operation and adjusts the global attenuation parameter in the state component processing engine in reverse.

[0024] Preferably, the state closed-loop correction module uses a recursive algorithm to dynamically approximate the global decay parameter in order to achieve online compensation for the parameter temperature drift induced by fluctuations in the external environment of the system, and ensure that the prediction accuracy of the real-time state tensor within a 10ms sampling period is not less than 98%.

[0025] Preferably, the sorting priority weight is calculated according to the following formula: Where H is the sorting priority weight, L is the normalized distance cost, and T is the dimensionless scalar value of the real-time state tensor. Assigning factor to distance Assign maturity factors; the addressing and scheduling center adjusts... and The ratio achieves coordinated control of mechanical operation energy consumption indicators and material process requirements.

[0026] Preferably, the virtual topology map maps the physical addressing space of the memory through a multi-dimensional matrix structure; each data node in the memory corresponds to a unique logical address number and is associated with an entry timestamp and an initial feature vector; the graph node mapping unit maintains the node occupancy characteristics of the virtual topology map by monitoring the status feedback of the material transfer control module.

[0027] Preferably, the state component processing engine uses an asynchronous scheduling mode to process the real-time state tensors of each node; the refresh frequency of the asynchronous scheduling mode is aligned with the synchronization pulse frequency of the downstream consumption end, so that the phase deviation between the issuance time of the addressing control sequence and the material consumption cycle is less than 50ms.

[0028] Preferably, the memory stores a feature benchmark library based on the evolution of historical attributes of materials; the state component processing engine identifies the stage-specific abrupt change points of the time-varying state parameters of the materials to be sorted by calculating the gradient change rate of the real-time state tensor in the virtual topology map, and improves the extraction response level of the addressing scheduling center for the corresponding logical address accordingly.

[0029] Example 1: In the intensive dynamic material sorting process of a continuous metallurgical manufacturing enterprise, high-temperature rolled parts in tens of thousands of storage locations continuously release heat energy and cause local thermal field interference. The static control logic, by setting the storage locations as isolated data units, fails to cover the coupling effects of adjacent nodes, causing a cumulative offset error between the material calculation state maintained by the system and the physical evolution state. This results in the extracted materials not meeting the temperature threshold requirements of downstream processes. Under this condition, the graph node mapping unit constructs a virtual topology graph in memory based on the storage location topology relationships recorded in the memory, and obtains the time-varying state parameters of the materials to be sorted in each storage location and the corresponding predicted decay curves. Independent of external sensing hardware, the state component processing engine responds to the update command of the time-varying state parameters of any node by retrieving... In the virtual topology graph, a dataset of associated nodes within the preset neighborhood of the target node is generated. Based on the logical step distance between the associated node dataset and the target node, a neighborhood interference compensation weight is calculated for the target node. This dynamically corrects the predicted decay curve and generates a real-time state tensor for the material to be sorted. This real-time state tensor includes a maturity component that characterizes the physical evolution of the material. In the control loop, the virtual topology graph established by the graph node mapping unit defines the search boundary for the extraction of associated node data. The state component processing engine eliminates parameter drift errors based on the neighborhood interference compensation weight. Together, they enable the control system to compensate for local micro-environment fluctuations within the system clock cycle. The addressing and scheduling center obtains the sorting trigger signal and calculates the deviation between the real-time state tensor and the preset sorting threshold.

[0030] When the deviation meets the preset convergence threshold, the joint cost evaluation component inside the addressing scheduling center obtains the real-time position feedback data of the external transfer mechanism and calculates the normalized distance cost for the external transfer mechanism to reach the candidate storage location. Since the mechanical transmission response speed and dynamic boundary of the stacker crane's horizontal operating axis, lifting axis, and fork extension axis in the automated warehouse differ physically from each other, the three-dimensional straight-line distance cannot objectively represent the actual mechanical addressing action delay. The joint cost evaluation component extracts the three-dimensional spatial coordinates of the candidate storage location and calculates the absolute physical displacement from the current position along three independent axes. The joint cost evaluation component divides the absolute physical displacement along each axis by the preset maximum mechanical operating speed of the corresponding transmission axis, outputting the expected operating time for each axis. The joint cost evaluation component extracts the maximum value among the three expected operating times as the time-equivalent distance and divides the time-equivalent distance by the system-set global maximum allowable addressing time to generate the normalized distance cost for a specific external transfer mechanism. The joint cost evaluation component combines the normalized distance cost with the maturity component to establish sorting priority weights; the specific calculation follows the formula... Where H is the sorting priority weight, L is the normalized distance cost, and T is the dimensionless scalar value of the real-time state tensor. Assigning factor to distance A maturity factor is assigned; the addressing scheduling center selects the logical node with the highest sorting priority weight as the final scheduling target, generates an addressing control sequence containing the logical address of the target location, and sends the addressing control sequence to the external transfer mechanism, driving the external transfer mechanism to trigger the extraction action for the material to be sorted; the material status parameters output by the external transfer mechanism meet the process range requirements, and the logical topology calculation of the control unit achieves constraint synchronization with the physical space addressing requirements.

[0031] Example 2: This example aims to quantitatively verify the addressing error convergence performance of the neighborhood interference compensation mechanism based on virtual topology maps in a multi-heat-source strongly coupled environment. The experiment relies on a physically scaled-down storage model with a built-in aerodynamic disturbance generator and is equipped with an infrared thermal imager to obtain real thermal field distribution data. To realistically reproduce the irregular turbulence caused by moving mechanisms in industrial settings, low-frequency Gaussian white noise with a peak-to-peak value of 5% is injected into the system control terminal as a temperature feedback disturbance source. The state component processing engine needs to balance feature capture granularity and processor bus load when setting the sampling update cycle. When the first derivative of the temperature of spatially associated nodes exceeds a preset threshold, the system increases the sampling frequency to prevent spectral aliasing of high-frequency interference signals; conversely, it decreases the sampling frequency to release computing resources. Based on this rule, the initial average temperature of the test location is kept constant at 600°C. Furthermore, under the strongly coupled benchmark operating condition with a full load rate of 85.0%, the sampling update cycle was determined to be 1.5s.

[0032] The experiment constructed three spatial distance gradients to characterize the interference intensity, setting the physical distance between the centers of the associated nodes and the target node to be 0.5m, 1.0m, and 1.5m, respectively. Control group one used a static cooling coefficient and did not perform data retrieval and attenuation correction for neighboring nodes. Control group two extracted the associated node dataset but added the weighting factor for the maturity component in the joint cost evaluation component. The value was forcibly fixed at 0.1 to weaken the control weight of the material evolution state on the final addressing decision. The experimental group executed the complete correction logic including neighborhood interference compensation weights, and based on the formula... Establish sorting priority weights, where H represents the sorting priority weight, L represents the normalized distance cost, and T represents the dimensionless scalar value of the real-time state tensor. Representing the distance allocation factor, under the extreme condition of 0.5m with the shortest interference spacing, after 45.0 minutes of physical evolution, the calculated temperature maintained in the system memory of Control Group 1 deviated from the actual infrared measurement temperature by 18.7℃ due to the lack of a neighborhood interference compensation mechanism. Control Group 2, due to the serious bias of state weights, had a material temperature that deviated from the target temperature by 12.4℃ due to the scheduling center scheduling, indicating that there was a substantial misalignment between the physical state and the scheduling sequence. The graph node mapping unit of the experimental group converted the temperature rise effect of adjacent materials into compensation weights and fed them back to the state tensor. The absolute deviation between the final material temperature and the target setting value of 550.0℃ was only 1.2℃.

[0033] The cross-gradient comparison data exhibited nonlinear evolution characteristics. As the node spacing decreased from 1.5m to 1.0m, the measured temperature deviation of control group 1 increased from 8.5℃ to 11.3℃. When the spacing was further compressed to 0.5m, the superposition of the radiation fields from multiple heat sources caused the local temperature rise to exceed the linear boundary, and the measured deviation of control group 1 surged nonlinearly to 24.5℃. In this range, the system completely lost its ability to effectively track the material state. The temperature measurement deviation of the experimental group converged to 0.8℃, 1.1℃, and 1.4℃ under the above three spatial gradients, respectively. This set of data proves that the neighborhood interference compensation... The compensation weight effectively suppressed the second-order state drift error caused by the nonlinear deterioration of the thermal field. The above objective experimental data confirmed that the interference compensation path jointly constructed by the graph node mapping unit and the state component processing engine cut off the interference transmission of physical space interference to the logical state calculation under complex working conditions with airflow disturbance and strong thermal field coupling. The addressing and scheduling center established the final scheduling target by jointly normalizing the distance cost and maturity component, and established an engineering-feasible collaborative control window between mechanical addressing lag and material state dissipation, so that the physical evolution degree of the extracted material matches the subsequent process range standard.

[0034] Example 3: When concurrent access operations or local high temperatures cause spatial node overlap in a high-density automated storage and retrieval warehouse, the waiting time for target materials is extended and the physical thermal parameters deviate from the set process range. Before outputting the addressing control sequence, the addressing scheduling center starts the path verification module. The path verification module reads the initial access path generated by the addressing scheduling center and traverses the logical nodes in the virtual topology map level by level along the initial access path. The path verification module obtains the real-time status flag of each logical node. When a logical node with physical space occupation or temperature exceeding the limit is identified, the logical node is set as a state-locked node. The state component processing engine takes the state-locked node as the origin and extracts the set of candidate avoidance nodes in the idle state according to the breadth-first search algorithm. For any candidate avoidance node in the candidate avoidance node set, the path verification module calculates the topology offset steps of its deviation from the initial access path and establishes the time compensation variable based on the external transfer mechanism's reference moving speed of 2.5m / s and the topology offset steps.

[0035] The state component processing engine inputs the time compensation variable into the predicted decay curve of the target material, extracts the corrected maturity component of the target material after superimposing the avoidance delay, and the addressing scheduling center calculates the comprehensive passage cost of each candidate avoidance node. The specific calculation logic is as follows: Structure, where C is the synthesis cost and D is the topology offset step number. ΔT is the step penalty coefficient, and ΔT is the absolute difference between the corrected maturity component and the preset sorting threshold. The maturity penalty coefficient is determined based on the maximum no-load acceleration of the external transfer mechanism and the current cooling rate of the target material. The addressing scheduling center selects the candidate avoidance node with the lowest comprehensive passage cost and splices it to the initial access path to generate a new path structure that bypasses the state-locked node to reconstruct the addressing control sequence. The addressing control sequence containing avoidance constraint instructions is sent to the material transfer control module. The material transfer control module issues the addressing control sequence to the external transfer mechanism, driving the external transfer mechanism to move along the new path structure. The external transfer mechanism avoids interference areas and maintains continuous physical movement. The physical parameters of the material to be sorted extracted converge within the set process range. The spatial addressing smoothness of the system and the material thermal state control requirements are synchronously constrained.

[0036] Example 4: When the system faces the initial deployment of multiple batches of heterogeneous materials, the graph node mapping unit continuously collects the natural cooling temperature data of the target alloy material in a static state through an infrared temperature measurement device, calculates the temperature drop gradient within adjacent sampling periods, multiplies the temperature drop gradient with the three-dimensional size characteristic value of the material to calculate the baseline decay slope, and the state component processing engine writes the baseline decay slope into the system memory to construct a discrete data matrix characterizing the heat dissipation process. The graph node mapping unit extracts the predicted decay curve based on the discrete data matrix.

[0037] When the system connects to a newly deployed external transport facility, the addressing and scheduling center sends a full-load acceleration / deceleration control sequence to the external transport facility. Position sensors collect the actual hysteresis time of the external transport facility's response to the full-load acceleration / deceleration control sequence. The joint cost evaluation component multiplies the actual hysteresis time by the system communication cycle value to calculate the spatial offset tolerance. Based on the spatial offset tolerance, the addressing and scheduling center determines the comprehensive cost calculation formula through a monotonically decreasing function mapping. Step penalty coefficient Where C is the overall throughput cost, D is the topology offset steps, and ΔT is the absolute difference between the corrected maturity component and the preset sorting threshold. Based on the maturity penalty coefficient determined by the baseline decay slope, the addressing scheduling center generates an addressing control sequence based on quantitative evaluation parameters including the step penalty coefficient. The material transfer control module issues the addressing control sequence to limit the spatial addressing actions of the external transfer mechanism.

[0038] Example 5: When the system faces the initial deployment of multiple batches of heterogeneous materials, the graph node mapping unit continuously collects the natural cooling temperature data of the target alloy material in a static state using an infrared thermometer at a sampling frequency of 10Hz; the graph node mapping unit calculates the temperature drop gradient within adjacent sampling periods; the temperature drop gradient is multiplied by the three-dimensional dimensional characteristic values ​​of the target alloy material to determine the baseline decay slope; based on Newton's law of cooling, the convective heat release rate of high-temperature materials has an objective physical mapping relationship with the effective heat dissipation surface area, convective heat transfer coefficient, and temperature difference with the surrounding environment; the graph node mapping unit obtains the target alloy material... The material's specific heat capacity parameter, mass value, and initial background temperature of the storage space are combined with continuously collected natural cooling temperature data. The equivalent convective heat transfer coefficient under the current fluid environment is obtained through first-order discrete difference operation. The state component processing engine uses the equivalent convective heat transfer coefficient and the ratio of the material's effective surface area to its mass as core input variables to calculate the nonlinear heat dissipation mapping benchmark for a specific batch of material under different ambient temperature zones. The state component processing engine writes the benchmark decay slope into the system memory to construct a discrete data matrix characterizing the heat dissipation process. The graph node mapping unit extracts the predicted decay curve corresponding to the measured characteristics based on the discrete data matrix.

[0039] When the system faces electromechanical commissioning of a newly deployed external transfer mechanism, the addressing and dispatching center issues a full-load acceleration / deceleration control sequence to the external transfer mechanism; the position sensor collects the actual hysteresis time of the external transfer mechanism in response to the full-load acceleration / deceleration control sequence; the joint cost evaluation component multiplies the actual hysteresis time by the system communication cycle value with a time length of 10ms to determine the spatial offset tolerance; the addressing and dispatching center determines the comprehensive cost calculation formula based on the spatial offset tolerance through a monotonically decreasing function mapping. Step penalty coefficient Where C represents the overall throughput cost, D represents the topology offset steps, and ΔT represents the absolute difference between the corrected maturity component and the preset sorting threshold. The maturity penalty coefficient is determined based on the baseline attenuation slope. The addressing scheduling center generates an addressing control sequence based on quantitative evaluation parameters including the step penalty coefficient. The material transfer control module issues the addressing control sequence to limit the spatial addressing actions of the external transfer mechanism. When the system faces a multi-heat source spatial radiation gradient distribution, the state component processing engine determines the logical steps based on the number of physical shelf barriers between the target node and associated nodes. The state component processing engine extracts the instantaneous temperature offset of the associated node relative to the environmental baseline value. The state component processing engine divides the instantaneous temperature offset by the logical steps to calculate the neighborhood interference compensation weight. The specific calculation structure follows the formula. Where W represents the neighborhood interference compensation weight, The instantaneous temperature bias of the associated node is represented by S, which represents the logical step number. The state component processing engine multiplies the neighborhood interference compensation weight into the predicted attenuation curve of the target material to correct the temperature drop slope. The thermal interference effect of the far-end heat source on the target material is transformed into a deterministic attenuation operator that is inversely proportional and monotonic to the spatial barrier level.

[0040] When the material transfer control module drives the external transfer mechanism to the target location, the system faces a situation where the predicted decay curve maintained in memory and the actual physical evolution trajectory caused by environmental changes result in cumulative drift. The state closed-loop correction module acquires the target material's physical characteristic variables collected by the temperature measuring component located at the contact end of the external transfer mechanism, extracts the calculated temperature value at the corresponding moment from the real-time state tensor, calculates the algebraic absolute difference between the physical characteristic variables and the calculated temperature value to generate evolution residuals, and the state component processing engine extracts the cooling coefficient of the predicted decay curve of the same specification material in the system memory and injects the evolution residual as a compensation term into the cooling coefficient. The state closed-loop correction module has built-in discrete Kalman filter recursive operation logic. During the system power-on self-test phase, the module continuously acquires 100 sets of sampling data from the temperature measuring component on the constant temperature reference block and calculates the sample variance, calibrating the measurement noise variance R to 0.06 and simultaneously calibrating the process noise variance Q to 0.02. The initial diagonal elements of the posterior error covariance matrix are set to 1.00. During the recursive calculation, the module uses the predicted temperature at the current moment as the state variable and performs a correction on the cooling coefficient in steps of 0.005 using the calculated Kalman gain. The state closed-loop correction module sets the evolution residual as the direct input variable for the measurement update step, extracts the posterior error covariance matrix of the previous sampling period, and calculates the Kalman gain parameter for the current sampling period by combining it with the current measurement noise variance. The state component processing engine calculates the algebraic product of the evolution residual and the Kalman gain parameter, and superimposes the algebraic product as the dynamic compensation step size onto the prior cooling coefficient of the current period to generate the updated posterior cooling coefficient, which overwrites the discrete value in the original memory address. The control system suppresses the global parameter drift caused by high-frequency environmental noise through fixed-cycle feedforward and measurement recursion procedures. The discrete inference data stream in the virtual topology map and the objective physical evolution field of the industrial site achieve closed-loop convergence mapping.

[0041] The above description is only a few preferred embodiments of the present invention and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, technical solutions formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A dynamic material sorting system based on an automated warehouse, characterized in that, The system includes: The graph node mapping unit is used to construct a virtual topology graph in memory based on the topology relationship of the storage locations recorded in the memory, and to obtain the time-varying state parameters of the materials to be sorted in each storage location and the corresponding predicted decay curves. The state component processing engine is used to respond to the update command of the time-varying state parameters of any node, retrieve the dataset of associated nodes in the preset neighborhood of the target node in the virtual topology map, and calculate the neighborhood interference compensation weight for the target node based on the logical step distance between the associated node dataset and the target node, so as to dynamically correct the predicted decay curve and generate the real-time state tensor of the material to be sorted; wherein, the real-time state tensor contains a maturity component that characterizes the degree of physical evolution of the material. The addressing and scheduling center is used to acquire sorting trigger signals and calculate the deviation between the real-time state tensor and the preset sorting threshold. When the deviation meets the preset convergence threshold condition, it generates an addressing control sequence containing the logical address of the target cargo location. The material transfer control module communicates with the addressing and scheduling center and is used to send addressing control sequences to external transfer mechanisms to drive them to trigger extraction actions for the materials to be sorted.

2. The dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, The addressing and scheduling center also includes a joint cost evaluation component; the joint cost evaluation component is used to obtain real-time location feedback data of external transfer agencies during the addressing control sequence generation process, and to calculate the normalized distance cost of external transfer agencies reaching candidate cargo locations. The joint cost evaluation component establishes sorting priority weights by negatively weighting the normalized distance cost with the maturity component in the real-time state tensor, and selects the logical node with the highest sorting priority weight as the final scheduling target.

3. A dynamic material sorting system based on an automated warehouse according to claim 2, characterized in that, Before the addressing control sequence is output by the addressing scheduling center, the path verification module retrieves the status-locked nodes on the access path of the final scheduling target in the virtual topology map. If there are logical nodes in the access path that are in the occupied state, the addressing scheduling center is triggered to perform logical reconstruction of the addressing control sequence to generate avoidance constraint instructions and synchronize them to the material transfer control module.

4. A dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, When calculating the neighborhood interference compensation weight, the state component processing engine shows an inversely proportional monotonic relationship between the value of the neighborhood interference compensation weight and the number of logical steps from the target node to the associated node, so as to quantitatively characterize the interference intensity of the time-varying state parameter evolution of the spatially adjacent region of the virtual topology map for the material to be sorted.

5. A dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, The system also includes a state closed-loop correction module; the state closed-loop correction module is used to obtain physical sampling data of the corresponding logical address when the material transfer control module arrives at the target location, and calculate the prediction deviation between the physical sampling data and the predicted value of the real-time state tensor; when the prediction deviation exceeds 5%, the addressing scheduling center triggers the addressing sequence overwrite operation and adjusts the global attenuation parameter in the state component processing engine in reverse.

6. A dynamic material sorting system based on an automated warehouse according to claim 5, characterized in that, The state closed-loop correction module uses a recursive algorithm to dynamically approximate the global decay parameter in order to achieve online compensation for parameter temperature drift induced by fluctuations in the external environment of the system, ensuring that the prediction accuracy of the real-time state tensor within a 10ms sampling period is not less than 98%.

7. A dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, The virtual topology map maps the physical addressing space of the memory through a multi-dimensional matrix structure; each data node in the memory corresponds to a unique logical address number and is associated with an entry timestamp and an initial feature vector; the graph node mapping unit maintains the node occupancy characteristics of the virtual topology map by monitoring the status feedback of the material transfer control module.

8. A dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, The state component processing engine uses an asynchronous scheduling mode to process the real-time state tensors of each node; the refresh frequency of the asynchronous scheduling mode is aligned with the synchronization pulse frequency of the downstream consumption end, so that the phase deviation between the issuance time of the addressing control sequence and the material consumption cycle is less than 50ms.

9. A dynamic material sorting system based on an automated warehouse according to claim 1, characterized in that, The memory stores a feature benchmark library based on the evolution of historical material attributes; the state component processing engine identifies the stage-specific abrupt change points of the time-varying state parameters of the materials to be sorted by calculating the gradient change rate of the real-time state tensor in the virtual topology map, and improves the extraction response level of the addressing scheduling center for the corresponding logical address accordingly.

Images