Factory logistics allocation system
By combining the intelligent decision-making layer, the digital twin layer, and the end-to-end collaboration layer, dynamic logistics scheduling instructions are generated, solving the problem of adaptability and collaborative optimization of the factory logistics scheduling system in the face of dynamic changes, and realizing efficient and environmentally friendly logistics scheduling.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-31
- Publication Date
- 2026-04-07
AI Technical Summary
Existing factory logistics scheduling systems are poorly adaptable to dynamic changes, lack end-to-end collaborative optimization, cannot proactively predict equipment failures, and fail to effectively consider environmental impacts, resulting in low system efficiency and resource waste.
The system employs an intelligent decision-making layer to integrate multi-source data, a digital twin layer for simulation testing, and a full-link collaboration layer to achieve green optimization, generating dynamic logistics scheduling instructions. By combining reinforcement learning models and predictive maintenance modules, it enables dynamic collaborative scheduling and optimization across factories.
It enables dynamic adaptation to real-time changes, improves system responsiveness and resource utilization, reduces carbon emissions and total costs, reduces the probability of production interruptions, and achieves a balanced optimization of economy and environment.
Smart Images
Figure CN121810152A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of logistics allocation technology and relates to a factory logistics allocation system. Background Technology
[0002] Factory logistics coordination is a core component of modern manufacturing, and its efficiency directly impacts order delivery cycles, production costs, and resource utilization. With the increasing prevalence of multi-factory collaborative production models, logistics scheduling systems need to perform integrated optimization of orders, production capacity, and transportation routes in a complex and dynamic environment.
[0003] Currently, most common factory logistics scheduling solutions are based on rule engines or static optimization algorithms. These systems typically pre-determine fixed scheduling rules or cost parameters based on historical experience during the system initialization phase. In actual operation, the system allocates orders and plans routes according to these static rules. However, existing technologies have at least the following significant drawbacks:
[0004] Static and rigid, with poor adaptability: Systems based on fixed rules cannot effectively adapt to dynamic changes in the production environment. For example, when encountering disturbances such as sudden equipment failure, temporary changes in order priorities, traffic congestion, or supply chain disruptions, the original scheduling rules may immediately become invalid, causing a sharp drop in system efficiency or even stagnation. Manual intervention is required for adjustment, resulting in slow response and an inability to achieve true dynamic optimization.
[0005] Information silos limit optimization: Existing systems are often limited to optimizing a single factory or specific process, lacking deep collaboration with manufacturing execution systems, warehouse management systems, and upstream and downstream supply chain systems. Data barriers prevent holistic optimization from a full-chain perspective; a locally optimal solution may not be a globally optimal solution, limiting the potential for cost reduction and efficiency improvement.
[0006] Passive response, lack of prediction: Most systems only respond passively after a problem occurs, such as rescheduling only after equipment actually crashes. The lack of predictive capabilities for risks such as equipment failure and order fluctuations prevents proactive warnings and avoidance, making it difficult to guarantee the reliability of production plans.
[0007] The optimization model is too narrow in scope and neglects green goals: Traditional optimization models typically focus only on economic benefits, such as minimizing logistics costs or production cycles, while failing to consider environmental impacts, including carbon emissions, as a significant decision-making factor. With increasing global emphasis on sustainable development and the introduction of policies such as carbon tariffs, existing systems are struggling to meet the strategic needs of enterprises for carbon footprint management and green manufacturing.
[0008] Therefore, a factory logistics allocation system is needed to solve the above problems. Summary of the Invention
[0009] To address the aforementioned problems, this invention proposes a factory logistics allocation system that effectively solves the issues in the prior art.
[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0011] A factory logistics dispatching system, comprising:
[0012] The intelligent decision-making layer is used to integrate multi-source data from orders, equipment and the external environment, and to process the multi-source data using an artificial intelligence model to generate dynamic logistics scheduling instructions.
[0013] The digital twin layer, which is communicatively connected to the intelligent decision-making layer, is used to construct a virtual mapping model corresponding to the physical factory entity, and to perform simulation testing and optimization on the logistics scheduling instructions generated by the intelligent decision-making layer based on the virtual mapping model, so as to obtain optimized scheduling instructions.
[0014] The end-to-end collaboration layer communicates with the intelligent decision-making layer and the external supply chain system, respectively, and is used to realize data exchange between the upstream and downstream of the supply chain through standardized interfaces, and to provide green optimization indicators as constraints to the intelligent decision-making layer.
[0015] The intelligent decision-making layer generates the final logistics scheduling instructions based on the green optimization indicators provided by the end-to-end collaboration layer and the simulation optimization results fed back by the digital twin layer, so as to realize dynamic collaborative scheduling and optimization across factories.
[0016] Preferably, the digital twin layer performs simulation testing and optimization of logistics scheduling instructions, specifically including:
[0017] Before the logistics scheduling instructions are issued to the physical factory for execution, various business scenarios are simulated in the virtual mapping model;
[0018] Evaluate the performance metrics of executing the logistics scheduling instructions under different business scenarios;
[0019] Based on the performance metrics, the optimal scheduling scheme is selected from multiple alternative scheduling schemes as the optimized scheduling instruction.
[0020] Preferably, the green optimization indicators include carbon footprint cost; the end-to-end collaboration layer is also used to calculate carbon emissions from logistics activities according to the Global Logistics Emissions Committee standards;
[0021] The optimization objective function upon which the intelligent decision-making layer generates dynamic logistics scheduling instructions includes a carbon footprint cost term calculated from the carbon emissions.
[0022] Preferably, the artificial intelligence model used in the intelligent decision-making layer is a reinforcement learning model;
[0023] The reward function of the reinforcement learning model includes a basic reward term and a cost penalty term, wherein the cost penalty term includes at least one of the carbon footprint cost, logistics cost, manufacturing cost, and delay cost.
[0024] Preferably, the reward function further includes a dynamic penalty term for sudden equipment disturbances, wherein the weight coefficient of the dynamic penalty term is positively correlated with the equipment failure probability.
[0025] Preferably, it also includes a predictive maintenance module, which is used to predict the probability of equipment failure based on the equipment's historical operating data, and generate a maintenance work order when the probability of failure exceeds a preset threshold;
[0026] The digital twin layer receives maintenance work order information and updates the equipment status in the virtual mapping model. The intelligent decision layer then regenerates logistics scheduling instructions based on the updated equipment status.
[0027] Preferably, the predictive maintenance module uses a long short-term memory neural network model to predict the probability of equipment failure.
[0028] A factory logistics allocation method, applied to the aforementioned factory logistics allocation system, includes the following steps:
[0029] It integrates multi-source data and generates dynamic logistics scheduling instructions through artificial intelligence models;
[0030] The logistics scheduling instructions are simulated, tested, and optimized within the virtual mapping model constructed in the digital twin layer.
[0031] Obtain green optimization indicators through the end-to-end collaboration layer;
[0032] Based on the aforementioned green optimization indicators and simulation optimization results, the final scheduling instructions are generated and executed.
[0033] Compared with the prior art, the present invention has the following beneficial effects:
[0034] This invention features an intelligent decision-making layer, a digital twin layer, and a full-link collaboration layer, constructing a basic architecture for a factory logistics allocation system capable of dynamic, collaborative, and closed-loop optimization. It addresses the fundamental pain point of traditional scheduling systems, which are based on static rules and cannot adapt to real-time changes. Through the interaction of these three layers—intelligent decision-making, digital twin, and full-link collaboration—it achieves a shift from passive response to proactive optimization, providing a system framework for further optimization. Attached Figure Description
[0035] Figure 1 This is a schematic diagram of the overall system architecture of the present invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] The following is in conjunction with the appendix Figure 1 The specific embodiments of the present invention will be described in further detail below.
[0038] Example 1:
[0039] A factory logistics allocation system, consisting of Figure 1 As shown, in order to solve the problem that traditional scheduling systems are based on static rules and cannot adapt to real-time changes, this invention includes:
[0040] The intelligent decision-making layer is used to integrate multi-source data from orders, equipment and the external environment, and to process the multi-source data using an artificial intelligence model to generate dynamic logistics scheduling instructions.
[0041] The digital twin layer, which is communicatively connected to the intelligent decision-making layer, is used to construct a virtual mapping model corresponding to the physical factory entity, and to perform simulation testing and optimization on the logistics scheduling instructions generated by the intelligent decision-making layer based on the virtual mapping model, so as to obtain optimized scheduling instructions.
[0042] The end-to-end collaboration layer communicates with the intelligent decision-making layer and the external supply chain system, respectively, and is used to realize data exchange between the upstream and downstream of the supply chain through standardized interfaces, and to provide green optimization indicators as constraints to the intelligent decision-making layer.
[0043] The intelligent decision-making layer generates the final logistics scheduling instructions based on the green optimization indicators provided by the end-to-end collaboration layer and the simulation optimization results fed back by the digital twin layer, so as to realize dynamic collaborative scheduling and optimization across factories.
[0044] It should be noted that the core of this embodiment lies in constructing a three-layer architecture of intelligent decision-making, digital twin, and end-to-end collaboration, as well as their closed-loop interactive relationship.
[0045] The "multi-source data" integrated by the intelligent decision-making layer may include: order attributes such as product model and delivery date, real-time factory capacity utilization, equipment health index (such as AGV battery life, whose normal range is 0.8-1.0), and external factors. The equipment health index includes AGV battery life, whose normal range is 0.8-1.0, and external factors include traffic congestion index.
[0046] In this embodiment, when constructing the virtual mapping model of the digital twin layer, the Unity engine is used, and the Apache Kafka streaming platform is connected to collect IoT sensor data in real time at a frequency of 100Hz to ensure that the virtual model is synchronized with the physical entity.
[0047] The "standardized interface" of the end-to-end collaboration layer is specifically implemented as a RESTful API, with JSON-LD as the data format, to achieve semantic data interoperability with upstream and downstream supply chain systems such as Manufacturing Execution System (MES) and Warehouse Management System (WMS).
[0048] Ultimately, the intelligent decision-making layer generates the "final logistics scheduling instruction," which can be specifically represented as a decision matrix. In one embodiment, order A is assigned to production line Y of factory X, and AGV is instructed to transport it along path Z, thereby achieving dynamic collaborative scheduling and optimization across factories.
[0049] Furthermore, by Figure 1 As shown, in order to improve the success rate and economy of the scheduling scheme, the digital twin layer performs simulation testing and optimization of logistics scheduling instructions, specifically including:
[0050] Before the logistics scheduling instructions are issued to the physical factory for execution, various business scenarios are simulated in the virtual mapping model;
[0051] Evaluate the performance metrics of executing the logistics scheduling instructions under different business scenarios;
[0052] Based on the performance metrics, the optimal scheduling scheme is selected from multiple alternative scheduling schemes as the optimized scheduling instruction.
[0053] It should be noted that various business scenarios include, but are not limited to: simulating a 20% surge in order demand, simulating AGV path conflicts, and simulating sudden failures of critical equipment. In one embodiment, when the system receives an order that increases total demand by 20%, the digital twin layer will test at least three different scheduling schemes in advance in a virtual environment.
[0054] Performance metrics include total cost, order completion time, and equipment utilization, where total cost equals logistics cost, manufacturing cost, delay cost, and carbon footprint cost. By comparing the performance metrics of different solutions, the system can select the solution with the lowest total cost as the optimized scheduling instruction, thereby mitigating potential risks before actual execution.
[0055] Furthermore, to make decision-making more economical and environmentally friendly, the green optimization indicators include carbon footprint costs; the end-to-end collaboration layer is also used to calculate carbon emissions in logistics activities according to the Global Logistics Emissions Committee standards.
[0056] The optimization objective function upon which the intelligent decision-making layer generates dynamic logistics scheduling instructions includes a carbon footprint cost item calculated from the carbon emissions.
[0057] It should be noted that the calculation of carbon emissions in logistics activities according to the Global Logistics Emissions Council (GLEC) standard is implemented by using different emission coefficients for different modes of transport and routes. For example, an electric AGV emits 0.05 kg CO2 per kilometer, while a diesel truck emits 0.15 kg CO2 per kilometer.
[0058] In the intelligent decision-making layer, the optimization objective function is specifically manifested as part of the reward function or cost function. For example, in the optimization objective, the carbon footprint cost will be quantified as a cost item and participate in the optimization calculation along with other costs, thereby guiding the system to prioritize low-carbon emission scheduling schemes.
[0059] Furthermore, in order to achieve a balance between economic and environmental costs, the artificial intelligence model used in the intelligent decision-making layer is a reinforcement learning model.
[0060] The reward function of the reinforcement learning model includes a basic reward term and a cost penalty term, wherein the cost penalty term includes at least one of the carbon footprint cost, logistics cost, manufacturing cost, and delay cost;
[0061] It should be noted that the state space of the reinforcement learning model is designed to include multi-dimensional inputs ranging from 1 to 1000 units, such as order quantity, equipment health index, and capacity utilization rate. The action space is defined as the order allocation decision matrix. The basic form of the reward function is: Reward = Basic Reward - (Logistics Cost + Manufacturing Cost + Delay Cost + Carbon Footprint Cost). This design ensures that the model, while pursuing efficiency, can comprehensively balance economic and environmental costs.
[0062] Furthermore, in order to add a dynamic penalty term for sudden equipment disturbances, the reward function also includes a dynamic penalty term for sudden equipment disturbances, the weight coefficient of which is positively correlated with the probability of equipment failure.
[0063] In this implementation, the probability of equipment failure is provided by the predictive maintenance module.
[0064] The dynamic penalty weight is set as follows: for every 1% increase in the failure rate, the penalty coefficient increases by 0.05. For example, when the predicted failure probability of a device increases from the base value of 5% to 6%, the penalty coefficient increases accordingly, making the reward function more inclined to avoid using the device, thereby enhancing the robustness of the scheduling scheme.
[0065] Furthermore, in order to reduce the probability of production interruption due to sudden failures, a predictive maintenance module is also included, which is used to predict the probability of equipment failure based on the historical operating data of the equipment, and generate a maintenance work order when the failure probability exceeds a preset threshold.
[0066] The digital twin layer receives maintenance work order information and updates the equipment status in the virtual mapping model. The intelligent decision layer regenerates logistics scheduling instructions based on the updated equipment status.
[0067] It should be noted that the predictive maintenance module predicts the probability of failure by analyzing data such as equipment vibration frequency and temperature history records over the past 30 days. The preset failure probability threshold can be set to 0.7. Once this threshold is exceeded, the system automatically generates a maintenance work order. After receiving this information, the digital twin layer immediately marks the equipment status as "under maintenance" or "reliability degraded" in the virtual model. The intelligent decision-making layer then recalculates the scheduling based on this updated global status, thus achieving a complete closed loop from prediction to maintenance to scheduling adjustment.
[0068] Furthermore, in order to improve the accuracy of fault prediction, the predictive maintenance module uses a long short-term memory (LSTM) neural network model to predict the probability of equipment failure.
[0069] It's important to note that LSTM models are particularly well-suited for processing time-series data, such as equipment operation data. The model takes sensor data (e.g., vibration, temperature) from the past 30 days as input and outputs the probability of equipment failure within a future period. Trained on historical data, this model effectively captures trends in equipment performance degradation, enabling accurate predictive maintenance.
[0070] Example 2:
[0071] This embodiment provides a factory logistics allocation method applied to the system described in Embodiment 1, including the following steps:
[0072] It integrates multi-source data and generates dynamic logistics scheduling instructions through artificial intelligence models;
[0073] The logistics scheduling instructions are simulated, tested, and optimized within the virtual mapping model constructed in the digital twin layer.
[0074] Obtain green optimization indicators through the end-to-end collaboration layer;
[0075] Based on the aforementioned green optimization indicators and simulation optimization results, the final scheduling instructions are generated and executed.
[0076] It should be noted that this method embodies the collaborative working logic of the three-layer architecture of this invention:
[0077] Integration and Decision-Making: The system first integrates multi-source data from orders, equipment and environment across the entire chain, and then generates a preliminary dynamic logistics scheduling instruction by the reinforcement learning model of the intelligent decision-making layer.
[0078] Simulation and Optimization: The initial instruction is sent to the digital twin layer. In the virtual mapping model, the system simulates various possible business scenarios, such as equipment failure and order surges, to conduct a "sandbox simulation" of the instruction, evaluate its performance, and provide feedback on optimization results. Simultaneously, early warning information from the predictive maintenance module is also incorporated into this simulation process.
[0079] Collaboration and Constraints: The end-to-end collaboration layer obtains information from external systems and provides green optimization indicators such as carbon footprint as hard constraints or optimization targets to the intelligent decision-making layer.
[0080] Final decision and execution: The intelligent decision-making layer integrates the simulation optimization results and green constraints to generate the final optimal scheduling instruction, which is then sent to the physical factory for execution.
[0081] Through the above steps, this method achieves a leap from static rules to dynamic optimization, from single-plant optimization to end-to-end collaboration, and from passive response to proactive prediction, ultimately achieving the technical effects of shortening response time, reducing total cost and carbon emissions, and improving system reliability.
[0082] This invention features a novel structure, ingenious design, and simple and convenient operation. It effectively solves the problem that traditional scheduling systems, based on static rules, cannot adapt to real-time changes, thus improving the success rate and economy of scheduling schemes. This makes decision-making more economical and environmentally friendly, achieving a balance between economic and environmental costs. It also adds dynamic penalty items for sudden equipment disturbances, reducing the probability of production interruptions caused by sudden failures and improving the accuracy of fault prediction.
[0083] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A factory logistics allocation system, characterized in that, include: The intelligent decision-making layer is used to integrate multi-source data from orders, equipment and the external environment, and to process the multi-source data using an artificial intelligence model to generate dynamic logistics scheduling instructions. The digital twin layer, which is communicatively connected to the intelligent decision-making layer, is used to construct a virtual mapping model corresponding to the physical factory entity, and to perform simulation testing and optimization on the logistics scheduling instructions generated by the intelligent decision-making layer based on the virtual mapping model, so as to obtain optimized scheduling instructions. The end-to-end collaboration layer communicates with the intelligent decision-making layer and the external supply chain system, respectively, and is used to realize data exchange between the upstream and downstream of the supply chain through standardized interfaces, and to provide green optimization indicators as constraints to the intelligent decision-making layer. The intelligent decision-making layer generates the final logistics scheduling instructions based on the green optimization indicators provided by the end-to-end collaboration layer and the simulation optimization results fed back by the digital twin layer, so as to realize dynamic collaborative scheduling and optimization across factories.
2. The factory logistics allocation system according to claim 1, characterized in that: The digital twin layer performs simulation testing and optimization of logistics scheduling instructions, specifically including: Before the logistics scheduling instructions are issued to the physical factory for execution, various business scenarios are simulated in the virtual mapping model; Evaluate the performance metrics of executing the logistics scheduling instructions under different business scenarios; Based on the performance metrics, the optimal scheduling scheme is selected from multiple alternative scheduling schemes as the optimized scheduling instruction.
3. The factory logistics allocation system according to claim 2, characterized in that: The green optimization indicators include carbon footprint cost; the end-to-end collaboration layer is also used to calculate carbon emissions from logistics activities according to the Global Logistics Emissions Committee standards. The optimization objective function upon which the intelligent decision-making layer generates dynamic logistics scheduling instructions includes a carbon footprint cost term calculated from the carbon emissions.
4. A factory logistics allocation system according to claim 3, characterized in that: The artificial intelligence model used in the intelligent decision-making layer is a reinforcement learning model; The reward function of the reinforcement learning model includes a basic reward term and a cost penalty term, wherein the cost penalty term includes at least one of the carbon footprint cost, logistics cost, manufacturing cost, and delay cost.
5. A factory logistics allocation system according to claim 4, characterized in that: The reward function also includes a dynamic penalty term for sudden equipment disturbances, the weight coefficient of which is positively correlated with the probability of equipment failure.
6. A factory logistics allocation system according to claim 1, characterized in that: It also includes a predictive maintenance module, which is used to predict the probability of equipment failure based on the equipment's historical operating data, and generate a maintenance work order when the failure probability exceeds a preset threshold; The digital twin layer receives maintenance work order information and updates the equipment status in the virtual mapping model. The intelligent decision layer then regenerates logistics scheduling instructions based on the updated equipment status.
7. A factory logistics allocation system according to claim 6, characterized in that: The predictive maintenance module uses a long short-term memory neural network model to predict the probability of equipment failure.
8. A factory logistics allocation method, applied to a factory logistics allocation system according to any one of claims 1-7, characterized in that, Includes the following steps: It integrates multi-source data and generates dynamic logistics scheduling instructions through artificial intelligence models; The logistics scheduling instructions are simulated, tested, and optimized within the virtual mapping model constructed in the digital twin layer. Obtain green optimization indicators through the end-to-end collaboration layer; Based on the aforementioned green optimization indicators and simulation optimization results, the final scheduling instructions are generated and executed.