Multi-type fresh food community group purchase path planning method considering heterogeneous customer demands

Abstract

The invention discloses a multi-type fresh food community group purchase path planning method considering heterogeneous customer demands. The method comprises the following steps: S1, modeling three punishment costs according to a fresh food freshness decline curve and customer satisfaction; s2, constructing a multi-type fresh food community group purchase distribution path optimization model considering heterogeneous customer demands; and S3, using a large-scale neighbor search algorithm to solve a multi-type fresh food community group purchase path planning scheme considering heterogeneous customer demands. According to the method, a multi-type fresh food distribution efficient optimization scheme can be systematically designed according to goods taking time window requirements of different types of customers, the freshness requirement of the distributed fresh food is ensured, the satisfaction degree of the customers on the freshness of the distributed fresh food is improved, and the goods return probability is reduced. Meanwhile, a large-scale neighbor search algorithm is used as a distribution path optimization algorithm, the algorithm is high in applicability, high in expansibility and insensitive to real-time parameters, and good solving quality can be obtained.

Description

Technical Field

[0001] The present invention relates to the field of logistics technology, and specifically to a community group purchase route planning method for multiple types of fresh products taking into account heterogeneous customer demands. Background Art

[0002] With the continuous development of community group buying, online community group buying and other online platforms have become one of the main ways for people to obtain daily necessities, and people's lifestyles and consumption habits have gradually changed. This major social change has led to a surge in demand for online shopping. As a form of online shopping, community group buying of fresh products has also experienced rapid development. Therefore, with the gradual increase in demand for online shopping, the challenge of vehicle routing with time windows in the context of the community group buying model has become a hot topic in academic research today. In fact, because the freshness of fresh products will drop rapidly, community residents are prone to receive fresh products with lower freshness, which leads to a decrease in customer satisfaction.

[0003] Traditional delivery route planning solutions are difficult to adapt to the challenges of route planning that take into account the different pickup needs of heterogeneous customers and the perishable characteristics of fresh products in cold chain logistics. Therefore, how to systematically design a delivery plan so that fresh products can be delivered within the specified time and the satisfaction of community residents can be maximized is particularly important in community group buying. Therefore, there is an urgent need for a community group buying route planning solution for multiple types of fresh products that can take into account the needs of heterogeneous customers. Summary of the invention

[0004] The purpose of the present invention is to provide a community group purchasing route planning method for multiple types of fresh products that takes into account heterogeneous customer needs, so as to solve the above defects.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for planning a community group purchase route for multiple types of fresh products considering heterogeneous customer demands includes the following steps:

[0007] S1. Model three penalty costs based on the freshness decline curve of fresh products and customer satisfaction;

[0008] S2. Construct a community group purchase distribution route optimization model for multiple types of fresh products that takes into account heterogeneous customer needs;

[0009] S3. Use a large-scale nearest neighbor search algorithm to solve the community group purchase path planning solution for multiple types of fresh products that takes into account heterogeneous customer needs.

[0010] Preferably, in step S1, three penalty costs are modeled according to the freshness decline curve of fresh products and customer satisfaction, and the specific steps are as follows:

[0011] S101. Construct the freshness reduction function of fresh products stored in the refrigerated truck and at the group leader respectively:

[0012]

[0013] In the formula, α p and β p They represent the corruption coefficients related to the transportation process and the storage process in the store, t represents the time, represents the function of freshness decline of fresh products during transportation, represents the function of freshness decline of fresh products when stored in shops;

[0014] S102. Model the cost of freshness reduction in the refrigerated truck based on the relationship between the freshness reduction function and time. Costs incurred by loss of freshness at the group leader and penalty costs for late delivery to customers They are:

[0015]

[0016]

[0017] In the formula, ω represents the coefficient between the decrease in freshness and the loss caused by it, Indicates the time of cargo delivery, i.e. the time when the refrigerated truck unloads the cargo. Indicates the earliest time for customers to pick up the goods. Indicates the latest pickup time for customers;

[0018] S103. Based on the relationship between the delivery time and the pickup time windows of different customers, the total penalty cost function is calculated as:

[0019]

[0020] Preferably, in step S2, a community group purchase distribution route optimization model for multiple types of fresh products considering heterogeneous customer demands is constructed, and the specific steps are as follows:

[0021] S201. Comprehensively utilize the pickup time window requirements and product requirements of different customers at each community node to establish a community group purchase distribution route optimization model for multiple types of fresh products;

[0022] S202, set the objective function to:

[0023]

[0024] In the formula, c ij x ijk represents the basic transportation cost, where c ijrepresents the cost from the i-th to the j-th community, x ijk Indicates whether car k departs from location i and arrives at location j. represents the penalty cost, represents the total penalty, q ilp represents the total demand of the lth customer for the pth commodity in the i-th community;

[0025] The first target function c ij x ijk Represents the minimization of all basic costs in the distribution plan, the second term It means minimizing the penalty costs caused by the reduction of product freshness and late delivery;

[0026] Set the flow balance constraints as follows:

[0027]

[0028] In the formula, x 0jk indicates whether the kth car starts from the warehouse and arrives at the jth community. Then constraint (1) indicates that the kth car starts from the warehouse; x j，|N|+1，k Indicates whether the kth vehicle returns to the warehouse from the jth location. Then constraint (2) indicates that the kth vehicle will eventually return to the warehouse. Constraint (3) indicates that each vehicle will leave after reaching a certain community and serving the community, i.e., the flow balance constraint. Constraint (4) indicates that at least one vehicle is required to deliver at each community node. L ipk Indicates whether vehicle k provides the pth type of goods to the ith community, and constraint (5) indicates that the vehicle serves the community only when it passes through the community. The initial and final positions of each vehicle are the same station.

[0029] Set the load capacity constraint as follows:

[0030]

[0031]

[0032] In the formula, ρ k represents the maximum number of cargo types that the k-th vehicle can carry. Constraint (6) indicates that the actual cargo types that the vehicle can carry cannot exceed the upper limit. Constraint (7) indicates that the p-th commodity in the ith community must be delivered by at least one vehicle. Constraint (8) indicates that the k-th vehicle can deliver to the community only when it passes through the ith community. Constraint (9) indicates that the total load remains unchanged. Constraint (10) indicates that each community is visited in chronological order. Constraint (11) indicates that the time when leaving the warehouse is 0.

[0033] The following time window constraints are set to indicate that the delivery time of each vehicle meets the time window requirements of the community leader:

[0034]

[0035] In the formula, s i represents the service time of the ith community, e i and l i Indicates the earliest and latest time a community can receive goods.

[0036] Preferably, in step S3, a large-scale nearest neighbor search algorithm is used to solve a community group purchase path planning scheme for multiple types of fresh products that takes into account heterogeneous customer needs. The specific steps are as follows:

[0037] S301, using a heuristic greedy algorithm and random generation to initialize the population;

[0038] S302, using the Split algorithm to assign a delivery plan with a separator to each fixed-order node sequence;

[0039] S303, using elite selection strategy to select excellent individuals in the population, and performing crossover to obtain new sequences;

[0040] S304, using the Split algorithm again to assign a delivery plan with a separator to each fixed-order node sequence;

[0041] S305. Optimize the delivery path plan using a large neighborhood search algorithm including deletion, insertion, and local neighborhood search.

[0042] Preferably, in step S301, the population is initialized using a heuristic greedy algorithm and random generation, and the specific steps are as follows:

[0043] In each generation, the population is divided into two subpopulations according to whether the solution is feasible or not. The number of solutions in each subpopulation satisfies the lower limit μ and the upper limit μ+λ, where μ=25 and λ=40. In the population initialization, 4μ individuals are generated. In order to generate each individual, 100 sets of random sequences need to be generated. These sequences are divided into initial solutions through the Split algorithm, the fitness of each initial solution is calculated, and the best one is put into the population as the initial individual. All individuals are trained, and if an infeasible individual is generated, a correction operation is performed. At the end of initialization, the individuals in the population are divided into two subpopulations based on feasibility, and then the individuals of these two subpopulations are selected, and the two selected subpopulations are used as the first generation of individuals.

[0044] Preferably, in step S302, a Split algorithm is used to assign a delivery plan with a separator to each fixed-order node sequence, and the specific steps are as follows:

[0045] The Split algorithm requires the delivery order of all nodes and the cost between any arc segments arc(i, j), i, j∈V, where each node represents the demand for a certain fresh product in the community;

[0046] The cost of arc(i, j), i, j∈V is expressed as:

[0047]

[0048] In the formula, f ij represents the total cost of a delivery visit from community i to community j. Each arc arc(i, j), i, j∈V ensures that a vehicle can only deliver to a community once; c 0，i+1 represents the basic delivery cost from the warehouse to the first community i; c j，0 represents the cost from the last community j to the warehouse; c h，h+1 represents the basic cost from community h to community h+1. If h and h+1 represent different demands for different types of fresh products in the same community, then c h，h+1 =0; represents the total cost of vehicle delivery between communities, arc(i, j), i, j∈V also includes the penalty cost of different customers in different communities; let τ hk represents the arrival time of vehicle k, q hlp represents the demand of customer l for fresh food p in community h, then H lp (τ hk )q hlp represents the penalty cost of customer l for fresh product p in community h, so the total penalty cost is expressed as M is a sufficiently large positive integer.

[0049] After calculating all costs of arcs arc(i, j), i, j∈V, we can obtain the optimal solution in a fixed node sequence.

[0050] Preferably, in step S303, an elite selection strategy is adopted to select excellent individuals in the population and perform crossover to obtain a new sequence. The specific steps are as follows:

[0051] First, randomly select two parents from all individuals; then, randomly select two cut points i and j in Parent1, and copy the substrings P1(i)...P1(j) to the corresponding positions in Child1; finally, starting from the j+1 node in Parent2, if the node is not found in C1, insert the node into C1 in turn, and another child C2 can be obtained by exchanging P1 and P2 in the same way.

[0052] Preferably, in step S305, a large neighborhood search algorithm including deletion, insertion, and local neighborhood search is used to optimize the delivery path plan, and the specific steps are as follows:

[0053] Each operation traverses all possible pairs (u, v) in the route. The elements in each pair can contain one or more nodes. All nodes in a node pair may come from the same route or from different routes. Each node contains information about the community order number and product type. x and y represent the next nodes of u and v on their paths, respectively. T(u) represents the route of node u.

[0054] M1. If u is a customer node, remove u and insert it after v;

[0055] M2. If u and x are customers, remove them and then insert (u, x) after v.

[0056] M3. If u and x are customers, remove them and then insert (x,u) after v.

[0057] M4. If u and x are customers, exchange u and x;

[0058] M5. If u, x and v are customers, exchange (u, x) and v;

[0059] M6. If (u,x) and (v,y) are customers, exchange (u,x) and (v,y);

[0060] M7. If T(u) = T(v), replace (u,x) and (v,y) with (u,v) and (x,y);

[0061] M8. If T(u)≠T(v), replace (u,x) and (v,y) with (u,v) and (x,y);

[0062] M9. If T(u)≠T(v), replace (u,y) and (x,v) with (u,v) and (x,y).

[0063] The beneficial effects of the present invention are:

[0064] The present invention considers the needs of heterogeneous customers for a community group purchase of multiple types of fresh products. While considering the needs of heterogeneous customers, it comprehensively considers the pickup time window requirements and satisfaction with the freshness of fresh products of different types of customers in the community, and comprehensively uses the different freshness reduction degrees of multiple types of fresh products to construct a multiple type of fresh product distribution model that considers the pickup needs of heterogeneous customers. The system design relies on a path planning solution based on a large neighbor search algorithm, which can realize the design of a multiple type of fresh product distribution efficient optimization solution for the pickup time window requirements of different types of customers, ensures the freshness requirements of the delivered fresh products, improves the customer's satisfaction with the freshness of the delivered fresh products, and reduces the probability of returns. At the same time, a large-scale neighbor search algorithm is used as a distribution path optimization algorithm. The algorithm has strong applicability, high scalability, is insensitive to real-time parameters, and can achieve good solution quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 : Schematic diagram of the process of the present invention;

[0066] Figure 2 : Flowchart of the large-scale nearest neighbor search algorithm in the method of the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described below in conjunction with the embodiments. It should be noted that these are merely examples and illustrations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they should be deemed to fall within the protection scope of the present invention.

[0068] Embodiment 1:

[0069] Figure 1 It is a schematic diagram of the process of the present invention, such as Figure 1 As shown in FIG. 1 , a community group purchase route planning method for multiple types of fresh products considering heterogeneous customer demands includes the following steps:

[0070] S1. Model three penalty costs based on the freshness decline curve of fresh products and customer satisfaction. The specific steps are as follows:

[0071] S101. Construct the freshness reduction function of fresh products stored in the refrigerated truck and at the group leader respectively:

[0072]

[0073] In the formula, α p and β p They represent the corruption coefficients related to the transportation process and the storage process in the store, t represents the time, represents the function of freshness decline of fresh products during transportation, Represents the freshness reduction function of fresh products when stored in stores.

[0074] S102. Model the cost of freshness reduction in the refrigerated truck based on the relationship between the freshness reduction function and time. Costs incurred by loss of freshness at the group leader and penalty costs for late delivery to customers They are:

[0075]

[0076] In the formula, ω represents the coefficient between the decrease in freshness and the loss caused by it, Indicates the time of cargo delivery, i.e. the time when the refrigerated truck unloads the cargo. Indicates the earliest time for customers to pick up the goods. Indicates the latest pickup time for customers.

[0077] S103. Based on the relationship between the delivery time and the pickup time windows of different customers, the total penalty cost function is calculated as:

[0078]

[0079] S2. Construct a community group purchase distribution path optimization model for multiple types of fresh products that takes into account heterogeneous customer needs. The specific steps are as follows:

[0080] S201. Comprehensively utilize the pickup time window requirements and product requirements of different customers at each community node to establish a community group purchase distribution route optimization model for multiple types of fresh products;

[0081] S202, set the objective function to:

[0082]

[0083] In the formula, c ij x ijk represents the basic transportation cost, where c ij represents the cost from the i-th to the j-th community, x ijk Indicates whether car k departs from place i and arrives at place j; represents the penalty cost, represents the total penalty, q ilp represents the total demand of the lth customer for the pth commodity in the i-th community;

[0084] The first target function c ij x ijk Represents the minimization of all basic costs in the distribution plan, the second term It means minimizing the penalty costs caused by the reduction of product freshness and late delivery.

[0085] S203: Construct a distribution model that includes flow balance constraints, load constraints, and time window constraints.

[0086] Set the flow balance constraints as follows:

[0087]

[0088] In the formula, x 0jk indicates whether the kth car starts from the warehouse and arrives at the jth community. Then constraint (1) indicates that the kth car starts from the warehouse; x j，|N|+1，k Indicates whether the kth vehicle returns to the warehouse from the jth location. Then constraint (2) indicates that the kth vehicle will eventually return to the warehouse. Constraint (3) indicates that each vehicle will leave after reaching a certain community and serving the community, i.e., the flow balance constraint. Constraint (4) indicates that at least one vehicle is required to deliver at each community node. L ipk Indicates whether vehicle k provides the pth type of goods to the ith community, and constraint (5) indicates that the vehicle serves the community only when it passes through the community. The initial and final positions of each vehicle are the same station.

[0089] Set the load capacity constraint as follows:

[0090]

[0091] In the formula, ρ k represents the maximum number of cargo types that the k-th vehicle can carry. Constraint (6) indicates that the actual cargo types of the vehicle cannot exceed the upper limit. Constraint (7) indicates that the p-th commodity in the i-th community is delivered by at least one vehicle. Constraint (8) indicates that the k-th vehicle can deliver to the community only when it passes through the i-th community. Constraint (9) indicates that the total load remains unchanged; Constraint (10) indicates that each community is visited in chronological order; Constraint (11) indicates that the time when leaving the warehouse is 0.

[0092] The following time window constraints are set to indicate that the delivery time of each vehicle meets the time window requirements of the community leader:

[0093]

[0094] In the formula, s i represents the service time of the ith community, e i and l i Indicates the earliest and latest time a community can receive goods.

[0095] S3, Figure 2 is a flow chart of the large-scale nearest neighbor search algorithm in the method of the present invention, such as Figure 2As shown in the figure, a large-scale nearest neighbor search algorithm is used to solve the community group purchase path planning scheme for multiple types of fresh products considering heterogeneous customer needs. The specific steps are as follows:

[0096] S301, using a heuristic greedy algorithm and random generation to initialize the population.

[0097] In each generation, the population is divided into two subpopulations according to whether the solution is feasible. The number of solutions in each subpopulation satisfies the lower limit μ and the upper limit μ+λ, where μ=25 and λ=40. In the population initialization, 4μ individuals are generated. In order to generate each individual, 100 sets of random sequences need to be generated. Then Figure 2 The "sub-process" in the above is as follows: these sequences are divided into initial solutions through the Split algorithm, the fitness of each initial solution is calculated, and the best one is put into the population as the initial individual; all individuals are trained, and if an infeasible individual is generated, a correction operation is performed. At the end of initialization, the individuals in the population are divided into two sub-populations classified based on feasibility, and then the individuals of these two sub-populations are selected, and the two selected sub-populations are used as the first generation of individuals.

[0098] S302, using the Split algorithm to assign a delivery plan with separators to each fixed-order node sequence.

[0099] The Split algorithm requires the delivery order of all nodes and the cost between any arc segments arc(i, j), i, j∈V, where each node represents the demand for a certain fresh product in the community;

[0100] The cost of arc(i, j), i, j∈V is expressed as:

[0101]

[0102] In the formula, f ， represents the total cost of a delivery visit from community i to community j. Each arc arc(i, j), i, j∈V ensures that a vehicle can only deliver to a community once; c 0，i+1 represents the basic delivery cost from the warehouse to the first community i; c j，0 represents the cost from the last community j to the warehouse; c h，h+1 represents the basic cost from community h to community h+1. If h and h+1 represent different demands for different types of fresh products in the same community, then c h，h+1 =0; represents the total cost of vehicle delivery between communities, arc(i, j), i, j∈V also includes the penalty cost of different customers in different communities; let τ hk represents the arrival time of vehicle k, q hlp represents the demand of customer l for fresh food p in community h, then Hlp (τ hk )q hlp represents the penalty cost of customer l for fresh product p in community h, so the total penalty cost is expressed as M represents a sufficiently large positive integer.

[0103] After calculating all costs of arcs arc(i, j), i, j∈V, we can obtain the optimal solution in a fixed node sequence.

[0104] S303. Based on the fitness function (which is related to the freshness of fresh products, customer satisfaction and basic cost in practical problems), an elite selection strategy is used to select excellent individuals in the population, and crossover is performed to obtain a new sequence. The crossover steps are as follows:

[0105] First, randomly select two parents from all individuals; then, randomly select two cut points i and j in Parent1(P1), and copy the substrings P1(i)...P1(j) to the corresponding positions in Child1(C1); finally, starting from j+1 in Parent2(P2), if the node is not found in C1, insert the node into C1 in turn. Another child C2 can be obtained in the same way by exchanging P1 and P2.

[0106] S304, proceed again Figure 2 "Sub-process": Use the Split algorithm to assign a delivery plan with a separator to each fixed-order node sequence, see S301 for details.

[0107] S305. Use a large neighborhood search algorithm including deletion, insertion, and local neighborhood search to optimize the delivery path plan. The local search operator is M1-M9. After traversing M1-M9 in sequence, perform a "sub-process" to assign a delivery plan with a separator, and complete the local search step. If the optimal solution is improved, return to the cross-loop above steps until the maximum number of iterations or the maximum running time is reached; if there is no improvement, perform a population diversification operation, that is, add the diversity ranking based on the Hamming distance to the objective function with a certain weighted coefficient, select more diverse individuals, increase the possibility of variation, and then return to the cross-loop above steps.

[0108] Each operation traverses all possible pairs (u, v) in the route. The elements in each pair can contain one or more nodes. All nodes in a node pair may come from the same route or from different routes. Each node contains information about the community order number and product type. x and y represent the next nodes of u and v on their paths, respectively. T(u) represents the route of node u.

[0109] M1. If u is a customer node, remove u and insert it after v;

[0110] M2. If u and x are customers, remove them and then insert (u, x) after v.

[0111] M3. If u and x are customers, remove them and then insert (x,u) after v.

[0112] M4. If u and x are customers, exchange u and x;

[0113] M5. If u, x and v are customers, exchange (u, x) and v;

[0114] M6. If (u,x) and (v,y) are customers, exchange (u,x) and (v,y);

[0115] M7. If T(u) = T(v), replace (u,x) and (v,y) with (u,v) and (x,y);

[0116] M8. If T(u)≠T(v), replace (u,x) and (v,y) with (u,v) and (x,y);

[0117] M9. If T(u)≠T(v), replace (u,y) and (x,v) with (u,v) and (x,y).

[0118] The present invention considers the needs of heterogeneous customers for a community group purchase of multiple types of fresh products. While considering the needs of heterogeneous customers, it comprehensively considers the pickup time window requirements and satisfaction with the freshness of fresh products of different types of customers in the community, and comprehensively uses the different freshness reduction degrees of multiple types of fresh products to construct a multiple type of fresh product distribution model that considers the pickup needs of heterogeneous customers. The system design relies on a path planning solution based on a large neighbor search algorithm, which can realize the design of a multiple type of fresh product distribution efficient optimization solution for the pickup time window requirements of different types of customers, ensures the freshness requirements of the delivered fresh products, improves the customer's satisfaction with the freshness of the delivered fresh products, and reduces the probability of returns. At the same time, a large-scale neighbor search algorithm is used as a distribution path optimization algorithm. The algorithm has strong applicability, high scalability, is insensitive to real-time parameters, and can achieve good solution quality.

[0119] The above is an exemplary description of the invention. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as such non-substantial improvements are made by adopting the method concept and technical solution of the present invention, or the concept and technical solution of the invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A community group purchase route planning method for multiple types of fresh products considering heterogeneous customer needs, characterized in that: The following steps are involved: S1. Model three types of penalty costs based on the freshness decline curve of fresh products and customer satisfaction; S2. Construct a community group purchase distribution route optimization model for multiple types of fresh products that takes into account heterogeneous customer needs; S3. Use a large-scale nearest neighbor search algorithm to solve the community group purchase path planning solution for multiple types of fresh products that takes into account heterogeneous customer needs.

2. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 1 is characterized in that: In step S1, three penalty costs are modeled according to the freshness decline curve of fresh products and customer satisfaction. The specific steps are as follows: S101. Construct the freshness reduction function of fresh products stored in the refrigerated truck and at the group leader respectively: In the formula, α p and β p They represent the corruption coefficients related to the transportation process and the storage process in the store, t represents the time, represents the function of freshness decline of fresh products during transportation, represents the function of freshness decline of fresh products when stored in shops; S102. Model the cost of freshness reduction in the refrigerated truck based on the relationship between the freshness reduction function and time. Costs incurred by loss of freshness at the group leader and penalty costs for late delivery to customers They are: In the formula, ω represents the coefficient between the decrease in freshness and the loss caused by it, Indicates the time of cargo delivery, i.e. the time when the refrigerated truck unloads the cargo. Indicates the earliest time for customers to pick up the goods. Indicates the latest pickup time for customers; S103. Based on the relationship between the delivery time and the pickup time windows of different customers, the total penalty cost function is calculated as:

3. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 1 is characterized in that: In step S2, a community group purchase distribution path optimization model for multiple types of fresh products considering heterogeneous customer needs is constructed. The specific steps are as follows: S201. Comprehensively utilize the pickup time window requirements and product requirements of different customers at each community node to establish a community group purchase distribution route optimization model for multiple types of fresh products; S202, set the objective function to: In the formula, c ij x ijk represents the basic transportation cost, where c ij represents the cost from the i-th to the j-th community, x ijk Indicates whether car k departs from place i and arrives at place j. represents the penalty cost, represents the total penalty, q ilp represents the total demand of the lth customer for the pth commodity in the i-th community; The first target function c ij x ijk Represents the minimization of all basic costs in the distribution plan, the second term It means minimizing the penalty costs caused by the reduction of product freshness and late delivery; S203, constructing a distribution model including flow balance constraints, load constraints and time window constraints; First, the model variables are described: Set the flow balance constraints as follows: In the formula, x 0jk indicates whether the kth car starts from the warehouse and arrives at the jth community. Then constraint (1) indicates that the kth car starts from the warehouse; x j，|N|+1，k indicates whether the kth vehicle returns to the warehouse from the jth location. Then constraint (2) indicates that the kth vehicle eventually returns to the warehouse. Constraint (3) indicates that each vehicle will leave after completing the service at a certain community, i.e., the flow balance constraint. Constraint (4) indicates that at least one vehicle is required to deliver at each community node. ipk Indicates whether vehicle k provides the pth type of goods to the i-th community. Constraint (5) indicates that the vehicle serves the community only when it passes through the community. The initial and final positions of each vehicle are the same station. Set the load capacity constraint as follows: In the formula, ρ k represents the maximum number of cargo types that the kth vehicle can carry. Then constraint (6) indicates that the actual cargo types that the vehicle can carry cannot exceed the upper limit. Constraint (7) indicates that the pth commodity in the ith community is delivered by at least one vehicle; Constraint (8) indicates that the kth vehicle can deliver to the community only when it passes through the ith community; Constraint (9) indicates that the total load remains unchanged; Constraint (10) indicates that each community is visited in chronological order; Constraint (11) indicates that the time when leaving the warehouse is 0; Set the following time window constraints to ensure that each vehicle's delivery time meets the time window requirements of the community leader: In the formula, s i represents the service time of the ith community, e i and l i Indicates the earliest and latest time a community can receive goods.

4. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 1 is characterized in that: In step S3, a large-scale nearest neighbor search algorithm is used to solve a community group purchase path planning scheme for multiple types of fresh products that takes into account heterogeneous customer needs. The specific steps are as follows: S301, using a heuristic greedy algorithm and random generation to initialize the population; S302, using the Split algorithm to assign a delivery plan with a separator to each fixed-order node sequence; S303, using elite selection strategy to select excellent individuals in the population, and performing crossover to obtain new sequences; S304, using the Split algorithm again to assign a delivery plan with a separator to each fixed-order node sequence; S305. Optimize the delivery path plan using a large neighborhood search algorithm including deletion, insertion, and local neighborhood search.

5. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 4 is characterized in that: In step S301, the population is initialized using a heuristic greedy algorithm and random generation. The specific steps are as follows: In each generation, the population is divided into two subpopulations according to whether the solution is feasible or not. The number of solutions in each subpopulation satisfies the lower limit μ and the upper limit μ+λ, where μ=25 and λ=40. In the population initialization, 4μ individuals are generated. In order to generate each individual, 100 sets of random sequences need to be generated. These sequences are divided into initial solutions through the Split algorithm, the fitness of each initial solution is calculated, and the best one is put into the population as the initial individual. All individuals are trained, and if an infeasible individual is generated, a correction operation is performed. At the end of initialization, the individuals in the population are divided into two subpopulations based on feasibility, and then the individuals of these two subpopulations are selected, and the two selected subpopulations are used as the first generation of individuals.

6. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 4 is characterized in that: In step S302, a Split algorithm is used to assign a delivery plan with a separator to each fixed-order node sequence. The specific steps are as follows: The Split algorithm requires the delivery order of all nodes and the cost between any arc segments arc(i, j), i, j∈V, where each node represents the demand for a certain fresh product in the community; The cost of arc(i, j), i, j∈V is expressed as: In the formula, f ij represents the total cost of a delivery visit from community i to community j. Each arc arc(i, j), i, j∈V ensures that a vehicle can only deliver to a community once; c 0，i+1 represents the basic delivery cost from the warehouse to the first community i; c j，0 represents the cost from the last community j to the warehouse; c h，h+1 represents the basic cost from community h to community μ+1. If μ and μ+1 represent different demands for various types of fresh products in the same community, then c h，h+1 =0; represents the total cost of vehicle delivery between communities, arc(i, j), i, j∈V also includes the penalty cost of different customers in different communities; let τ hk represents the arrival time of vehicle k, q hlp represents the demand of customer l for fresh food p in community h, then H lp (τ hk )q hlp represents the penalty cost of customer l for fresh product p in community h, so the total penalty cost is expressed as M is a sufficiently large positive integer; After computing all costs of arcs arc(i,j),i,j∈V, we can obtain the optimal solution in a fixed node sequence.

7. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 4 is characterized in that: In step S303, an elite selection strategy is used to select excellent individuals in the population and perform crossover to obtain a new sequence. The specific steps are as follows: First, randomly select two parents from all individuals; then, randomly select two cut points i and j in Parent1, and copy the substrings P1(i)...P1(j) to the corresponding positions in Child1; finally, starting from the j+1 node in Parent2, if the node is not found in C1, insert the node into C1 in turn, and another child C2 can be obtained by exchanging P1 and P2 in the same way.

8. The method for planning routes for community group purchase of multiple types of fresh products considering heterogeneous customer needs according to claim 4 is characterized in that: In step S305, a large neighborhood search algorithm including deletion, insertion, and local neighborhood search is used to optimize the delivery path plan. The specific steps are as follows: Each operation traverses all possible pairs (u, v) in the route. The elements in each pair can contain one or more nodes. All nodes in a node pair may come from the same route or from different routes. Each node contains information about the community order number and product type. x and y represent the next nodes of u and v on their paths, respectively. T(u) represents the route of node u. Traversing M1-M9 in sequence is a complete local search operation. M1. If u is a customer node, remove u and insert it after v; M2. If u and x are customers, remove them and then insert (u, x) after v. M3. If u and x are customers, remove them and then insert (x,u) after v. M4. If u and x are customers, exchange u and x; M5. If u, x and v are customers, exchange (u, x) and v; M6. If (u,x) and (v,y) are customers, exchange (u,x) and (v,y); M7. If T(u) = T(v), replace (u,x) and (v,y) with (u,v) and (x,y); M8. If T(u)≠T(v), replace (u,x) and (v,y) with (u,v) and (x,y); M9. If T(u)≠T(v), replace (u,y) and (x,v) with (u,v) and (x,y).

Images