Main line taking and conveying integrated material transportation method

Abstract

The invention discloses a trunk line taking and delivering integrated material transportation method, which comprises the following steps: a demand prediction step: training a recurrent neural network order prediction model for predicting an order demand; an intelligent routing string point planning step of determining a package path according to the order demand and the corresponding supplier distribution; a vehicle type selection step: determining the vehicle type of the vehicle according to the number and supply-demand relationship of the goods picking bins on each goods picking path; and a stowage planning step: determining a packing scheme of goods in the goods picking path according to the planned goods picking path and the vehicle type of the vehicle, and when the packing scheme cannot be met, returning to the intelligent routing string point planning step, re-planning the goods picking path and / or re-determining the vehicle type of the vehicle. According to the main line taking and delivering integrated material transportation method, global allocation and stowage, global routing planning and real-time automatic dynamic adjustment are achieved. Agile and flexible development of a supply chain is promoted, and integration and precise cooperation of main line taking and sending modes are promoted and achieved.

Description

Technical Field

[0001] This application relates to the technical field of logistics management, and more specifically, to a material transportation method for integrated trunk pick-up and delivery. Background Art

[0002] In the context of the rapid development of the logistics industry, the efficiency and cost control in the process of material pick-up and delivery have become the core issues that logistics enterprises need to solve. Especially at the supply chain end, ensuring the timely and stable supply of the components required for product production is of great significance for the smooth production of automobiles. Therefore, establishing a flexible and efficient supply chain system has become an urgent need for industry enterprises.

[0003] The supply chain of automobile spare parts has a long chain and high management difficulty, and it is almost one of the most complex and large supply chains in all industrial supply chains. In the routing planning that appears in cooperation with multiple automobile manufacturers, in order to meet the problem of simultaneous pick-up and delivery, the driver's work tasks become extremely complex and heavy. To meet the needs of multiple manufacturers, the driver may need to shuttle between different locations frequently and continuously perform loading and unloading operations. This high-intensity work mode greatly increases the driver's working hours and driving time, resulting in an increase in driver fatigue, and then leading to traffic accidents, causing casualties and property losses.

[0004] In addition, there are currently technical problems in logistics distribution, such as low informatization level, relying on manual experience for decision-making, a long overall circulation chain, and low standardization, informatization, and intelligence levels in each link. Summary of the Invention

[0005] In order to solve the limitations of existing logistics transportation and the technical problems of low informatization level and relying on manual experience for decision-making, the present invention proposes a material transportation method for integrated trunk pick-up and delivery, which can solve the above problems.

[0006] To solve the above technical problems, the present invention is implemented by adopting the following technical solutions:

[0007] A material transportation method for integrated trunk pick-up and delivery, comprising:

[0008] A demand forecasting step of training a recurrent neural network order forecasting model for predicting order demand;

[0009] An intelligent routing and string-point planning step of determining a pick-up path according to the order demand and the corresponding supplier distribution;

[0010] A vehicle type selection step of setting a loading parameter β according to the number of pick-up warehouses and the supply-demand relationship on each pick-up path, calculating the pick-up demand of each pick-up warehouse, and determining the vehicle type with the lowest cost as the goal according to the pick-up demand;

[0011] The stowage planning step determines the packing plan for the goods in the pick-up path according to the planned pick-up path and the vehicle type. When the packing plan cannot be satisfied, return to the intelligent routing string point planning step to re-plan the pick-up path and / or re-determine the vehicle type.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The trunk pick-up and delivery integrated material transportation method of the present invention is based on an intelligent transformation solution of demand forecasting - intelligent routing string point planning - vehicle allocation - stowage planning, accurately schedules transportation capacity resources, reduces in-transit vehicles, improves the loading rate, and realizes the greening of long-distance pre-production logistics of automobiles under the "dual carbon" goal. Build an integrated intelligent logistics information system to achieve an integrated supply chain that can be sensed, visualized, and controlled, realize global allocation and stowage, global routing planning, and real-time automatic dynamic adjustment. Promote the agile and flexible development of the supply chain, and promote and realize the integrated and precise coordination of the trunk pick-up and delivery mode.

[0013] After reading the detailed description of the embodiments of the present invention in conjunction with the drawings, other features and advantages of the present invention will become clearer. Brief Description of the Drawings

[0014] Figure 1 is a flowchart of an embodiment of the trunk pick-up and delivery integrated material transportation method proposed by the present invention;

[0015] Figure 2 is a comparison diagram of demand forecasting and actual demand in an embodiment of the trunk pick-up and delivery integrated material transportation method proposed by the present invention;

[0016] Figure 3 is a schematic diagram of the convergence of the model in an embodiment of the trunk pick-up and delivery integrated material transportation method proposed by the present invention;

[0017] Figure 4 is a three-dimensional packing schematic diagram in an embodiment of the trunk pick-up and delivery integrated material transportation method proposed by the present invention;

[0018] Figure 5 is a schematic diagram of the robustness of the model in an embodiment of the trunk pick-up and delivery integrated material transportation method proposed by the present invention. Detailed Embodiments

[0019] The following further describes in detail the specific embodiments of the present invention with reference to the drawings.

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0021] It should be noted that in the description of the present invention, the terms indicating directions or positional relationships such as "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0022] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] Embodiment 1, as shown in Figure 1 the following, a material transportation method integrating trunk pick-up and delivery is proposed in this embodiment, including:

[0024] A demand prediction step of training a recurrent neural network order prediction model for predicting order demand.

[0025] An intelligent routing and string-point planning step of determining the cargo collection path according to the order demand and the corresponding supplier distribution.

[0026] A vehicle type selection step of setting the loading parameter β according to the number of cargo collection warehouses and the supply-demand relationship on each cargo collection path, calculating the cargo collection demand of each cargo collection warehouse, and determining the vehicle type with the lowest cost as the goal according to the cargo collection demand.

[0027] Stowage planning step: According to the planned cargo collection route and the vehicle model, determine the packing plan for the goods in the cargo collection route. When the packing plan cannot be met, return to the intelligent routing string point planning step to re-plan the cargo collection route and / or re-determine the vehicle model.

[0028] The trunk pick-up and delivery integrated material transportation method of this embodiment is based on an intelligent transformation plan of demand forecasting - intelligent routing string point planning - vehicle allocation - stowage planning, accurately schedules transportation capacity resources, reduces in-transit vehicles, improves the loading rate, and realizes the greening of long-distance pre-production logistics of automobiles under the "dual carbon" goal. Build an integrated intelligent logistics information system to realize an integrated supply chain that can be sensed, visualized, and controlled, realize global allocation and stowage, global routing planning, and real-time automatic dynamic adjustment. Promote the agile and flexible development of the supply chain, and promote and realize the precise coordination of the integrated trunk pick-up and delivery mode.

[0029] In the intelligent logistics system, accurately predicting the supply and demand changes of spare parts in the warehouse can help enterprises deploy resources in advance and flexibly respond to market fluctuations, thereby optimizing supply chain management and achieving the goal of cost reduction and efficiency improvement. With the increasing complexity of the types of spare parts, traditional time series forecasting methods such as traditional Recurrent Neural Networks (RNN) are difficult to handle the supply and demand fluctuations of various types of automobile parts in multiple periods. For this reason, this solution proposes a supply and demand forecasting model for various types of spare parts in the warehouse based on LSTM, aiming to anticipate future supply and demand trends in advance through learning historical supply and demand data to help the transportation decision-making layer deploy resources and transportation scheduling in advance.

[0030] In the data preprocessing process, in order to ensure that the training process of the model is more stable, accelerate the convergence speed, and effectively avoid the problem of gradient explosion or gradient disappearance in LSTM, we normalized the data. We used the Min-Max normalization method, and its formula is as follows:

[0031]

[0032] Among them, X n is the normalized data, X is the original data, X min and X max are the minimum and maximum values of the original data respectively.

[0033] After the prediction model finishes predicting, in order to make the prediction results return to the magnitude of the actual supply and demand data and facilitate subsequent path optimization, we need to perform anti-normalization processing on the output results of the model, so as to better serve the actual application scenarios of supply and demand forecasting.

[0034] After normalization is completed, the historical supply and demand data is first divided into a training set and a validation set. This model uses 80% of the data for training and 20% for validation. This can ensure that the model has enough historical data for learning, and at the same time, the generalization ability of the model can be tested with the validation set. Secondly, in order to capture the time series pattern of the supply and demand quantity, a sliding window technique is adopted, and the data of the past 24 hours is used as input to predict future demand and supply. Each input sample contains the historical supply and demand data of different types of spare parts so that the model can identify the supply and demand change trends of various types of spare parts.

[0035] To ensure that the model can accurately predict the changes in warehouse supply and demand, the model is trained through the Back Propagation Through Time (BPTT) algorithm. The model parameters are adjusted through backpropagation to minimize the error between the prediction result and the actual supply and demand data. The training steps of BPTT are as follows:

[0036] Step1 Forward propagation: The input data is passed layer by layer to the neural network units to generate the prediction result.

[0037] Step2 Calculate the error: Compare the model prediction result with the actual supply and demand data, and use the Mean Squared Error (MSE) as the loss function to calculate the prediction error.

[0038] Step3 Error backpropagation: According to the calculated error, adjust the weights in the neural network through the backpropagation algorithm. Specifically, the Adam Adaptive Moment Estimation Algorithm (Adam) is used to update the weights.

[0039] Step4 Iterative update: Continuously repeat the above process until the model converges and the prediction error reaches the minimum.

[0040] During the gradient optimization process, the model uses the Adam optimizer, which can dynamically adjust the learning rate of each parameter and combine momentum to accelerate convergence. To ensure the stability of the model, we also combine a learning rate decay strategy. As the number of training rounds increases, the learning rate gradually decreases to prevent the model from having too large a step size in the later stage of training, thus ensuring the convergence of the model.

[0041] To further prevent the model from overfitting, the Dropout regularization technique is introduced. During the training process, some neural network units are randomly discarded to avoid the model from overfitting to the training data and improve the generalization ability of the model on unseen data. This technique effectively solves the problem that the model performs well on the training data but poorly on the test data.

[0042] By using a neural network model to predict future supply and demand, the actual demand of the warehouse (blue line) and the predicted value of the model (red line) are as follows Figure 2 shown. The basic data for this example is only 750 pieces, but it still shows a good fitting effect. It can be seen that the LSTM model can not only capture the overall trend of the supply and demand changes in the warehouse, but also has a certain ability to predict large-scale supply and demand fluctuations, demonstrating good prediction ability. In addition, Figure 3 shows the convergence of the model. The key to judging whether the model converges is to observe the changes in the loss value (Loss) and the validation loss value (Validation Loss) curves. Among them, if the training loss gradually decreases as the number of training rounds increases and tends to be stable after a certain round, it means that the model is learning and converging, proving that the model has found a relatively optimal solution. If the validation loss is the same as the training loss and also gradually decreases and tends to be stable, it means that the model can not only remember the patterns of the training data but also generalize to new data. If the model is overfitting, the training loss should continue to decrease, while the validation loss starts to increase at a certain moment; if the model is underfitting, the training loss and the validation loss should remain unchanged or decrease slowly at a relatively high level. So from Figure 3 it can be seen that the LSTM model proposed in this paper has good convergence.

[0043] It is crucial to implement the vehicle routing planning and scheduling for multi-stop less-than-truckload logistics with pick-up and delivery integration on trunk and feeder lines. According to the supply and demand of each cargo collection warehouse, while satisfying the possibility of multi-stop generated by picking up goods and cross-distributing in various regions across the country, carefully design and select the transportation network to ensure the maximization of cost-effectiveness while meeting customer needs. By optimizing the driving routes and scheduling plans of vehicles, significantly reduce the logistics costs of enterprises, improve the overall logistics service quality, and thus enhance the competitiveness of enterprises.

[0044] In some embodiments, in the intelligent routing multi-stop planning step, it includes establishing a total cost minimization model, and the objective function of the total cost minimization model is as follows

[0045]

[0046] where S represents the set of all vehicle types, K represents the set of all vehicle quantities of model s, V represents the set of all cargo collection warehouses, F s represents the total fixed cost of vehicles of type s, represents that the vehicle travels from node i to node j, otherwise C s represents the unit mileage cost of vehicle activation, d ij represents the distance from node i to node j, C6 represents the time penalty cost per unit volume of goods, Denotes the unloading volume of the vehicle at node j. Denotes the time taken for the vehicle to reach node j. t1 represents the soft time window requirement, σ represents the carbon emission penalty coefficient, e s Denotes the amount of carbon dioxide emitted per kilometer traveled by a vehicle of model s.

[0047] According to the Interim Regulations on the Administration of Carbon Emission Trading issued on January 25, 2024, key carbon emission units should purchase carbon emission allowances through the national carbon emission trading market to clear the greenhouse gas emissions of the previous year, with the unit being yuan / ton.

[0048] Among the characters with the sk label, the meaning of sk represents the k-th vehicle of model s.

[0049] The constraint function of the total cost minimization model is:

[0050] This constraint ensures that the vehicle has operations at each point;

[0051] This constraint guarantees the balance between the passing points and the operating points.

[0052] This constraint guarantees that the maximum volume of goods obtained by the vehicle during transportation is less than β times the rated volume N of the k-th vehicle of model s s .

[0053] This constraint guarantees that the maximum weight of goods obtained by the vehicle during transportation is less than the rated load G of the k-th vehicle of model s s .

[0054] This constraint ensures that the pickup volume of each vehicle at node i meets the supply volume (volume) outward from this point.

[0055] This constraint ensures that the unloading volume of each vehicle at node i meets the demand volume (volume) inward from this point.

[0056] This constraint represents the time when the k-th vehicle of model s reaches each point, where is the time required for the vehicle to travel from node a to node j, v s is the average operating speed of model s, represents the time experienced by the vehicle when reaching node i.

[0057] This constraint ensures that the hard time window requirement is met. t2 represents the hard time window requirement.

[0058] This constraint represents the pick-up volume (in terms of cubic capacity) of the k-th vehicle of model s at node i, which is expressed by the length of a unit m part picked up at node i. Width Height to represent. It represents the quantity of m parts picked up by the k-th vehicle of model s at node i.

[0059] This constraint represents the delivery volume (in terms of cubic capacity) of the k-th vehicle of model s at node i, which is expressed by the length of a unit m part unloaded at node i. Width Height to represent. It represents the quantity of m parts unloaded by the k-th vehicle of model s at node i.

[0060] This constraint represents the weight of the goods picked up by the k-th vehicle of model s at node i.

[0061] This constraint represents the weight of the goods unloaded by the k-th vehicle of model s at node i.

[0062] Among them, represents that the vehicle delivers goods at node i, otherwise represents that the vehicle delivers goods at node j, otherwise represents the pick-up volume of the vehicle at node i, represents the delivery volume of the vehicle at node i, represents the pick-up volume of the vehicle at node j, represents the pick-up weight of the vehicle at node i, represents the delivery weight of the vehicle at node i, represents the pick-up weight of the vehicle at node j, represents the delivery weight of the vehicle at node j, represents the set of node numbers that the vehicle has passed through when it arrives at point i. β represents the maximum ratio of the goods in the vehicle to the rated volume of the vehicle, N S represents the rated volume of the vehicle of model s, G s represents the rated load of the vehicle of model s, D i represents the total volume of goods sent out from the collection warehouse node i, E i represents the total volume of goods required by the collection warehouse node i, represents the time required for the vehicle to travel from node a to node j, represents that the vehicle travels from node a to node j, otherwise $t_1$ represents the time it takes for the vehicle to reach node $i$, and $t_2$ represents the hard time window requirement. represents the length of the goods taken at node $i$. represents the width of the goods taken at node $i$. represents the height of the goods taken at node $i$. represents the length of the goods unloaded at node $i$. represents the width of the goods unloaded at node $i$. represents the height of the goods unloaded at node $i$. represents the quantity of part $m$ taken by the $k$-th vehicle of model $s$ at node $i$. represents the quantity of part $m$ unloaded by the $k$-th vehicle of model $s$ at node $i$. represents the unit weight of part $m$ taken by the $k$-th vehicle of model $s$ at node $i$. represents the unit weight of part $m$ unloaded by the $k$-th vehicle of model $s$ at node $i$. is the set of the number of parts taken by the $k$-th vehicle of model $s$ at node $i$.

[0063] The total distribution cost consists of vehicle fixed cost, vehicle cost per unit mileage consumption, time penalty cost, and carbon emission cost. Among them, the vehicle cost per unit mileage consumption mainly consists of fuel consumption cost, air-conditioning fuel consumption cost, urea consumption cost, and highway tolls. In the vehicle fixed cost, the costs related to the driver are elaborated.

[0064] In some embodiments, the cost per unit mileage $C$ s of the vehicle is calculated as follows:

[0065]

[0066] Among them, $C_1$ represents the highway toll per kilometer of the vehicle. represents the fuel cost per kilometer consumed by the vehicle of model $s$. represents the average vehicle maintenance cost per kilometer for the vehicle of model $s$ during driving. represents the urea cost per kilometer consumed by the vehicle of model $s$. represents the average air-conditioning fuel consumption cost per kilometer for the vehicle of model $s$.

[0067] During the transportation process, the fuel consumption cost of the vehicle is a key economic factor, which directly affects the transportation cost and operation efficiency. In this paper, the load estimation method is adopted, and we consider the fuel consumption rate per unit mileage of the vehicle in the unloaded and fully loaded states. When the vehicle is unloaded, due to the lighter load, its fuel consumption is relatively low; while when it is fully loaded, due to the vehicle needing to overcome greater gravity, the fuel consumption increases accordingly. This relationship can usually be approximated as linear.

[0068] To calculate the fuel consumption cost, we first determine the fuel consumption rate per unit mileage when the vehicle is unloaded and fully loaded. Then, based on the actual load of the vehicle, we can calculate the actual fuel consumption rate per unit mileage. Finally, multiplying this consumption rate by the fuel cost per unit mileage gives us the fuel consumption cost of the vehicle during transportation. The specific calculation formula is as follows:

[0069]

[0070] o s represents the basic fuel consumption cost per unit mileage of the vehicle of model s, represents the influence rate of the unloaded vehicle of model s on the basic fuel consumption, represents the influence rate of the fully loaded vehicle of model s on the fuel consumption, represents the ratio of the weight of the goods in the vehicle to the rated load of the vehicle of model s when the vehicle travels from node i to node j.

[0071] The operation of the air-conditioning compressor consumes about 20% of the engine's power. This additional power consumption causes an increase in the engine's load, resulting in an increase in fuel consumption. In addition, the energy consumption of the air conditioner is also affected by temperature and humidity. As these two environmental factors increase, the energy consumption of the air conditioner also increases accordingly. Usually, this increase in fuel consumption ratio is between 10% and 20%.

[0072] Therefore, considering the fuel cost consumed after the driver turns on the air conditioner, the cost consumed by the air conditioner per hour for each section of the road is calculated according to the temperature of each section of the path. At the same time, the longer the air conditioner is turned on, the higher the cost. The duration of the air conditioner being turned on is the driving duration on that section of the road.

[0073]

[0074] Among them, u ij represents the basic time taken from node i to node j without interference; represents the influence rate of the unloaded vehicle of model s on the driving time, represents the influence rate of the fully loaded vehicle of model s on the driving time, ε ij is the fuel consumption of the air conditioner turned on for the section of the road from node i to node j. By default, the air conditioner is not turned on, that is, ε ij = 0, δ ij represents the road condition influence factor on the path from node i to node j.

[0075]

[0076] Urea consumption cost:

[0077]

[0078] Among them, fuel sk represents the fuel consumption per unit mileage when vehicle k of model s is running; a sk is the fuel consumption per unit mileage of vehicle k of model s under no-load condition; m s is the basic urea consumption cost per unit mileage of vehicle k of model s; br is the relationship proportional coefficient between fuel consumption and urea consumption, and br > 0.

[0079] During the transportation of the vehicle, weather conditions, road conditions, traffic situations, and load will affect the vehicle driving speed. In the data preprocessing link of the system, the road conditions on the path will be analyzed and converted into weights and added to the comprehensive road condition influence factor δ ij Therefore, it will not increase the complexity of problem-solving. And the temperature will determine whether the vehicle running on this section turns on the air conditioner.

[0080] Classify the weather conditions according to wind, rain, snow, and fog, classify the road conditions according to the bumpiness of the road, etc., classify the traffic conditions according to the congestion degree under the road width, classify the temperature according to the temperature, and the influence of the load affects the average driving speed of the vehicle according to the ratio of the total weight of the goods loaded in the carriage to the rated load of the vehicle. Establish a classification table to classify weather (Weather), road (Bump), and traffic (Congestion) into four levels: A, B, C, and D, and each level corresponds to a value between 0 and 1, as shown in Table 6. After synthesis, the road condition influence factor on the path from node i to node j is δ ij .

[0081] In some embodiments, the calculation method of δ ij is as follows:

[0082] δ ij = α1·Weather i，j + α2·Bump i，j + α3·Congestion i，j ;

[0083] α1, α2, and α3 are weights respectively. Weather i，j is the weather influence factor, Bump i，j is the road influence factor, Congestion i，j is the traffic influence factor. When δ ij is greater than the set value, the planned path from node i to node j is not passable.

[0084] In some embodiments, the calculation method of the vehicle fixed cost F s is as follows:

[0085]

[0086] Among them, F1 represents the handling and transportation cost, represents the daily vehicle insurance cost, represents the vehicle annual inspection cost, represents the company affiliation cost, represents the driver cost.

[0087] Driver cost includes the driver's salary and the driver's fatigue cost. The fatigue level of the driver during driving will affect the driving condition and lead to traffic accidents. Therefore, the fatigue level of the driver should be related to the vehicle repair cost and the driver's medical cost after a traffic accident. At this time, the fatigue level of the driver is equivalent to the probability of a traffic accident occurring. Therefore, the driver's fatigue cost is obtained as follows:

[0088]

[0089] wage s is the driver's salary corresponding to the s model; ε is the driver's fatigue coefficient; accident is the relevant cost after an accident occurs.

[0090] The vehicle loading plan model is mainly modeled based on the results obtained from the path selection and vehicle type selection models, and designs the dynamic placement plan of the goods in this path. When no placement plan can be satisfied, the path and vehicle type will be reselected, so no objective function is set.

[0091] In order to better depict the position of the goods in the carriage, a Cartesian coordinate system is established now to accurately locate the position of each good in the carriage, so as to optimize the loading strategy and space utilization. It helps to effectively manage and schedule the goods in three-dimensional space.

[0092] Taking the vertex at the lower left rear corner of the carriage as the coordinate origin (0, 0, 0) (the front of the vehicle is considered the rear of the carriage), the x-axis is established parallel to the width W of the carriage, the y-axis is established parallel to the length L of the carriage, and the z-axis is established parallel to the height H of the carriage. As Figure 4 shown.

[0093] In some embodiments, determining the loading plan of each vehicle includes:

[0094] Taking the vertex at the lower left rear corner of the carriage as the coordinate origin, the x-axis is established parallel to the width W of the carriage, the y-axis is established parallel to the length L of the carriage, and the z-axis is established parallel to the height H of the carriage to establish a coordinate system;

[0095] Set three-dimensional loading constraints:

[0096] This constraint means that the position coordinates of any bin in the carriage cannot exceed the carriage range of the kth vehicle of the s model, L s 、W s, H s are the length, width and height of the carriage of Model S

[0097] This constraint means that the bin must be placed parallel to the carriage, that is, the length and width directions of the goods should be parallel to the length and width directions of the carriage of the k-th vehicle of Model S.

[0098] This constraint means that the goods cannot be inverted or placed obliquely and must be placed directly upwards to ensure that the height direction of the bin is parallel to the height direction of the carriage of the k-th vehicle of Model S.

[0099] This constraint means that the placement of the bin needs to follow the principle of first in, last out.

[0100] When then x1 > x2, y1 > y2, This constraint means that it is necessary to comprehensively evaluate whether the bin follows the principle of "first in, last out" according to its volume, weight and fragility.

[0101] Among them, let

[0102] denotes the set of the remaining items placed in the same k-th vehicle of Model S as item ;

[0103] denotes the set of items placed in the same k-th vehicle of Model S as item whose top surface is at the same height as the bottom surface of the p-th item of customer i;

[0104] denotes the coordinates of the projection of the item to be taken at node i on the bottom surface of the carriage at the lower left rear corner, denotes the coordinates of the projection of the item to be taken at node i on the bottom surface of the carriage at the upper right front corner, denotes the coordinates of the projection of the item to be taken at node j on the bottom surface of the carriage at the lower left rear corner, denotes the coordinates of the projection of the item to be taken at node j on the bottom surface of the carriage at the lower right front corner, denotes the projection coordinate value of the item to be taken at node i in the vertical direction (z-axis) of the carriage, denotes the projection coordinate value of the top of the item to be taken at node i in the vertical direction (z-axis) of the carriage, denotes the item The projected coordinate value of the bottom in the vertical direction (z-axis) of the carriage Indicates the item to be taken at node i The width of Indicates the item to be taken at node i The length of Indicates the item to be taken at node i The width of Indicates the item to be taken at node j The projected coordinate value in the vertical direction (z-axis) of the carriage Indicates the item to be taken at node j The projected coordinate value of the top in the vertical direction (z-axis) of the carriage Indicates the item to be taken at node i The projected coordinate value of the top in the direction of the carriage bottom surface (y-axis) Indicates the item to be taken at node j The projected coordinate value of the bottom in the direction of the carriage bottom surface (y-axis) Indicates the item to be taken at node j The projected coordinate value of the top in the direction of the carriage bottom surface (y-axis) Indicates the comprehensive value of the p-th bin taken by the k-th vehicle of s model at node i Indicates the comprehensive value of the r-th bin taken by the k-th vehicle of s model at node j

[0105]

[0106] From the above figure, the conclusion can be drawn: If the items and the item have overlapping parts on the carriage bottom surface, then x1 > x2 and y1 > y2. If either x1 ≤ x2 or y1 ≤ y2 holds, then the items and the item must have no overlapping parts in the projection on the carriage bottom surface. Similarly, let If the items and have overlapping parts in the projection on the back surface of the carriage, then x1 > x2 and z1 > z2. If either x1 ≤ x2 or z1 ≤ z2 holds, then the items and the item must have no overlapping parts in the projection on the back surface of the carriage. If the items and have overlapping parts in the projection on the left side surface of the carriage, then y1 > y2 and z1 > z2. If either y1 ≤ y2 or z1 ≤ z2 holds, then the items and There must be no overlapping parts in the left - hand side projection of the carriage.

[0107] The comprehensive value of the p - th bin taken by the k - th vehicle of model s at node i It is necessary to calculate the weighted average of the three values of volume, weight, and fragility value. The larger the comprehensive value, the higher the bin (box) should be placed. Therefore, for the comprehensive value, volume and weight are extremely small - type indicators, and the fragility value is an extremely large - type indicator. First, the extremely small - type indicators are normalized; secondly, since the units of volume, weight, and fragility value are different, it is necessary to eliminate the dimensions, then use the analytic hierarchy process to find the weights of the three variables respectively, and finally calculate the weighted average to obtain the comprehensive value.

[0108] In some embodiments, the comprehensive value is calculated as follows:

[0109]

[0110] Among them, v represents the volume of the bin, g represents the weight of the bin and the internal goods, fr represents the fragility value of the internal goods of the bin, and ω1, ω2, ω3 are the weight values respectively.

[0111] This stowage method combines the vehicle's rated volume, load constraint, and physical properties of goods (volume, weight, fragility) to achieve a fast solution for the global optimal stowage plan. The algorithm is deeply coupled with the demand prediction model and the path planning model, supporting intelligent decision - making throughout the supply chain process. According to the constraint conditions and the target cost function in the model, the input process of the algorithm is set.

[0112] When the predicted order demand (i.e., the output of the LSTM model), the vehicle parameter set, and the goods attribute matrix are input, the result is initialized, a random stowage plan is generated, and the adaptive adjustment neighborhood search range is determined. During the destruction and repair process, by removing some goods and re - inserting them, the space utilization rate is iteratively optimized and increased. The acceptance criterion for returns adopts accepting inferior solutions with a probability to avoid local optimality. Finally, the inertia weight of the PSO is adjusted according to historical performance, and the optimal stowage plan with the coordinates and stacking order of the goods is output.

[0113] During the algorithm training process, the method of dynamic space merging is adopted to support the intelligent merging of the remaining space to reduce fragmented space. Through fragility scoring and weight distribution optimization, finally, in the case study, the cargo damage rate is reduced by more than 15%, and the hybrid optimization strategy reduces the calculation time by 60%, which can adapt to large - scale real - time scheduling scenarios.

[0114] In some embodiments, in the demand prediction step, the recurrent neural network order prediction model is trained by the backpropagation algorithm, and during the training process, some neural network units are randomly discarded.

[0115] In some embodiments, the objective function of the total cost minimization model is solved by an adaptive large neighborhood search algorithm.

[0116] In neighborhood search, the Metropolis criterion of simulated annealing is used for probabilistic selection, increasing the implicit neighborhood range, avoiding being trapped in local optimal solutions, and improving the global search ability of the algorithm.

[0117] This solution uses a combination of an adaptive large neighborhood search algorithm and a particle swarm optimization algorithm to achieve vehicle route planning. Specifically, the initial solution is obtained by solving with the particle swarm optimization algorithm first, and then the optimal solution is obtained by using the adaptive large neighborhood search algorithm. Input the daily demand of each product in each location, the scheduled time, and the inventory of each warehouse, initialize the parameters, and iteratively output the optimal route planning, including the quantity of each type of vehicle required, the route planning of each vehicle, the expected arrival time at each node, and the loading and unloading volume at each node.

[0118] Exemplarily, 20 known warehouses and supply nodes are given, and their demands, inventories, and scheduled times are randomly generated. Figure 3 It can be verified that the algorithm has convergence and converges to the optimal solution. And after multiple verifications, as Figure 5 shown, it can be proved that under the same given conditions, the optimal solution converges to the same result and has good robustness.

[0119] In some embodiments, during the calculation process of the objective function of the total cost minimization model, obtaining the initial solution by an adaptive particle swarm optimization algorithm is also included.

[0120] An adaptability mechanism is introduced into the particle swarm optimization algorithm and the large neighborhood search algorithm, and the operator weights are updated according to the quality of the solution, avoiding being trapped in local optimal solutions and improving the quality of the optimal solution.

[0121] The particle swarm optimization algorithm has strong global search ability, and the large neighborhood search algorithm has strong local solution exploration ability. The combination of the two can avoid local search and improve the algorithm efficiency.

[0122] Certainly, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples either. Changes, modifications, additions, or substitutions made by those of ordinary skill in the art within the essence scope of the present invention should also belong to the protection scope of the present invention.

Claims

1. A material transportation method integrating trunk line pick-up and delivery, characterized in that, Including: A demand forecasting step of training a recurrent neural network order prediction model to predict order demand; An intelligent routing and string-point planning step of determining a cargo collection path according to the order demand and the corresponding supplier distribution; A vehicle type selection step of setting a loading parameter β according to the number of cargo collection warehouses and the supply-demand relationship on each cargo collection path, calculating the cargo collection demand of each cargo collection warehouse, and determining the vehicle type with the lowest cost as the goal according to the cargo collection demand; A stowage planning step of determining the stowage plan of the goods in the cargo collection path according to the planned cargo collection path and the vehicle type. When the stowage plan cannot be satisfied, return to the intelligent routing and string-point planning step to re-plan the cargo collection path and / or re-determine the vehicle type.

2. The material transportation method for integrated trunk line picking and delivery according to claim 1, characterized in that, In the intelligent routing and string-point planning step, a total cost minimum model is established, and the objective function of the total cost minimum model is: Among them, S represents the set of all vehicle types, K represents the set of all vehicle quantities of model s, V represents the set of all cargo compartments, F s represents the total fixed cost of vehicles of type s, means the vehicle travels from node i to node j, otherwise C s represents the cost per unit mileage of the vehicle in use, d ij represents the distance from node i to node j, C6 represents the time penalty cost per unit volume of goods, represents the unloading volume of the vehicle at node i, represents the time taken for the vehicle to reach node j, t1 represents the soft time window requirement, σ represents the carbon emission penalty coefficient, e s represents the amount of carbon dioxide emitted per kilometer traveled by a vehicle of model s; The constraint function of the total cost minimum model is: Among them, indicates that the vehicle delivers goods at node i, otherwise indicates that the vehicle delivers goods at node j, otherwise f i sk represents the pick-up volume of the vehicle at node i, represents the pick-up volume of the vehicle at node j, represents the unloading volume of the vehicle at node j, represents the pick-up weight of the vehicle at node i, represents the unloading weight of the vehicle at node i, represents the pick-up weight of the vehicle at node j, represents the unloading weight of the vehicle at node j, represents the set of node numbers that the vehicle has passed through when it arrives at point i. β represents the maximum ratio of the goods in the vehicle to the rated volume of the vehicle, N s represents the rated volume of a vehicle of model s, G s represents the rated load of a vehicle of model s, D i represents the total volume of goods sent out from the pick-up warehouse node i, E i represents the total volume of goods required by the pick-up warehouse node i, represents the time elapsed when the vehicle arrives at node i, represents the time required for the vehicle to travel from node a to node j, represents that the vehicle travels from node a to node j, otherwise represents the operation time of the vehicle at node i. t2 represents the hard time window requirement, represents the length of the goods taken away at node i, represents the width of the goods taken away at node i, represents the height of the goods taken away at node i, represents the length of the goods unloaded at node i, represents the width of the goods unloaded at node i, represents the height of the goods unloaded at node i, represents the number of m parts taken away by the k-th vehicle of model s at node i, represents the number of m parts unloaded by the k-th vehicle of model s at node i. represents the unit weight of m parts taken away by the k-th vehicle of model s at node i, represents the unit weight of m parts unloaded by the k-th vehicle of model s at node i, is the set of the number of parts taken away by the k-th vehicle of model s at node i.

3. The material transportation method for integrated trunk line picking and delivery according to claim 2, wherein Vehicle-enabled cost per unit mileage C s is calculated as follows: Among them, C1 represents the high-speed cost per kilometer of the vehicle, represents the fuel cost consumed per kilometer by the s-model vehicle; represents the average vehicle maintenance cost per kilometer traveled by the s-model vehicle, represents the urea cost consumed per kilometer by the s-model vehicle, represents the air-conditioning fuel consumption cost per kilometer traveled by the s-model vehicle; o s represents the basic fuel consumption cost per unit mileage of the vehicle of model s, represents the influence rate of the vehicle of model s on the basic fuel consumption when the vehicle is unloaded, represents the influence rate of the vehicle of model s on the fuel consumption under full load, represents the ratio of the weight of the goods in the vehicle to the rated load of the vehicle when the vehicle of model s travels from node i to node j; Among them, u ij represents the basic time spent from node i to node j without interference; It represents the impact rate of vehicle type s on driving time when it is empty. Indicates the impact rate of vehicle type s on driving time under full load, ε ij The fuel consumption when the air conditioner is turned on on the road section from node i to node j is ε. The air conditioner is not turned on by default. ij =0,δ ij represents the road condition impact factor on the path from node i to node j; Among them, fuel sk represents the fuel consumption per unit mileage when vehicle k of model s is running; a sk is the fuel consumption per unit mileage of vehicle k of model s under the condition of no load; m s is the basic urea consumption cost per unit mileage of vehicle of model s; br is the relationship proportionality coefficient between fuel consumption and urea consumption, and br > 0.

4. The material transportation method of trunk line picking and delivering integration according to claim 3, characterized in that δ ij The calculation method is as follows: δ ij = α1·Weather ij + α2·Bump i，j + 03·Congestion i，j ; α1, α2, and α3 are weights respectively, Weather i，j is the weather impact factor, Bump i，j is the road impact factor, Congestion i，j is the traffic impact factor. When δ ij is greater than the set value, the planned path from node i to node j is not passable.

5. The material transportation method for integrated trunk line picking and delivery according to claim 2, wherein Vehicle fixed cost F s The calculation method is as follows: Among them, F1 represents the handling and transportation cost, represents the daily vehicle insurance cost, represents the vehicle annual inspection cost, represents the company affiliation cost, represents the driver cost; wage s It is the driver's wage corresponding to Model S; ε is the driver fatigue coefficient; accident is the related cost after an accident occurs.

6. The material transportation method for integrated trunk line picking and delivery according to any one of claims 2-5, characterized in that, Determining the stowage plan of each vehicle includes: Taking the vertex at the lower left rear corner of the carriage as the coordinate origin, establishing an x-axis parallel to the carriage width W, a y-axis parallel to the carriage length L, and a z-axis parallel to the carriage height H to establish a coordinate system; Setting three-dimensional stowage constraints: Among them, let Denote the set of the remaining items placed in the k-th vehicle of the same s model as the item ​ Denote the articles placed in the k-th vehicle of the s model vehicle, and the set of articles whose top surfaces are at the same height as the bottom surface of the p-th article of customer i The item to be picked up at node i The coordinates of the projection at the lower left rear corner of the carriage bottom surface, The item to be picked up at node i The coordinates of the projection at the upper right front corner of the carriage bottom surface, The item to be picked up at node j The coordinates of the projection at the lower left rear corner of the carriage bottom surface, The item to be picked up at node j The coordinates of the projection at the lower right front corner of the carriage bottom surface. The item to be picked up at node i The projection coordinate value in the vertical direction (z-axis) of the carriage, The item to be picked up at node i The projection coordinate value of the top in the vertical direction (z-axis) of the carriage, The item to be picked up at node i The projection coordinate value of the bottom in the vertical direction (z-axis) of the carriage, The item to be picked up at node i The width, The item to be picked up at node i The length, The item to be picked up at node i The width, The item to be picked up at node j The projection coordinate value in the vertical direction (z-axis) of the carriage, The item to be picked up at node j The projection coordinate value of the top in the vertical direction (z-axis) of the carriage, Represents the comprehensive value of the p-th bin picked up by the k-th vehicle of model s at node i The projection coordinate value of the top in the direction of the carriage bottom surface (y-axis), Represents the comprehensive value of the item picked up by the k-th vehicle of model s at node j The projection coordinate value of the bottom in the direction of the carriage bottom surface (y-axis), Represents the comprehensive value of the item picked up by the k-th vehicle of model s at node j The projection coordinate value of the top in the direction of the carriage bottom surface (y-axis), Represents the comprehensive value of the p-th bin picked up by the k-th vehicle of model s at node i, Represents the comprehensive value of the r-th bin picked up by the k-th vehicle of model s at node j.

7. The material transportation method for integrated trunk line picking and delivery according to claim 6, characterized in that, Comprehensive value The calculation method is as follows: where v represents the volume of the bin, g represents the weight of the bin and the internal goods, fr represents the fragility value of the internal goods of the bin, and ω1, ω2, and ω3 are weight values respectively.

8. The material transportation method for integrated trunk line picking and delivering according to any one of claims 1-5, characterized in that In the demand forecasting step, the recurrent neural network order prediction model is trained by the backpropagation algorithm, and during the training process, some neural network units are randomly discarded.

9. The material transportation method for integrated trunk line picking and delivering according to any one of claims 2-5, characterized in that, The objective function of the total cost minimum model is solved by the adaptive large neighborhood search algorithm.

10. The material transportation method for integrated trunk line picking and delivery according to claim 9, characterized in that, During the calculation process of the objective function of the total cost minimum model, an initial solution is also obtained by the adaptive particle swarm algorithm.

Images