A logistics route optimization method and system based on adaptive NSGAII

Abstract

A logistics route optimization method based on adaptive NSGAII includes: S1. Analyze and establish a basic mathematical model; S2. Design the chromosome coding of the individual population; S3. Use a combination of random and greedy methods to generate an initial population of size N, and perform Pareto non-dominated sorting on the population; S4. Obtain the next generation population through three evolutionary operations of selection, crossover, and mutation; S5. Merge the parent population and the child population to form a new population, and select better individuals to form a new parent population; S6. Generate a new child population through population evolution operator operation, and repeat step S4; S7. If the population iteration termination condition is not met, return to steps S4 and S5 until the termination condition is met; S8. Output the optimal solution set by the adaptive NSGAⅡ algorithm, thereby obtaining a distribution route for one vehicle. The present invention also includes a logistics route optimization system based on adaptive NSGAII. The present invention can solve the distribution route optimization problem.

Description

Technical Field

[0001] The present invention relates to the field of transportation technology, and in particular to a method and system for optimizing logistics routes. Background Art

[0002] In recent years, with the vigorous development of e-commerce, modern logistics has developed rapidly and vigorously, playing an important role in improving people's lives and promoting economic development. Among them, distribution is a link directly connected to consumers, and its service level is directly related to consumers' consumption experience, the core competitiveness and image of enterprises. Therefore, improving logistics distribution efficiency has become a key means for logistics companies to improve service levels and win the market.

[0003] In actual logistics distribution, there are many factors to consider, such as the number of vehicles, capacity, fuel consumption, maximum driving distance, possible traffic jams or other unexpected factors. At present, the overall operation level and service efficiency of the logistics industry need to be improved. Especially in some e-commerce promotion activities, the stimulation of consumption by preferential activities has led to a surge in logistics volume, and a large number of express deliveries in a short period of time has challenged the efficiency of logistics distribution. Therefore, it is necessary to study the optimization of distribution paths.

[0004] The present invention is mainly aimed at the logistics distribution problem of a single distribution center, two vehicle models, multiple distribution points and a limited vehicle collection. The specific definition is as follows: a logistics distribution center has a limited number of vehicles of two vehicle models, and goods are distributed from the distribution center to multiple distribution points. The distance between the distribution center and each distribution point, the demand of the distribution point, the maximum load of the vehicle, the fuel consumption cost, the unloading time, and the maximum mileage are known. It is required to reasonably arrange the vehicle distribution route and driving time, so as to improve the distribution efficiency and reduce the distribution cost.

[0005] The goals of urban logistics distribution problems mainly include minimizing monetary and time costs. For the dual-car distribution problem, the advantages of small cars are fast speed and low time cost, but their loading capacity is small, which requires an increase in the number of distribution cars and the number of distribution times, resulting in an increase in monetary costs; while the characteristics of large cars are the opposite of small cars. Summary of the invention

[0006] The present invention aims to overcome the above problems existing in the prior art and provides a logistics route optimization method and system based on adaptive NSGA-II.

[0007] The present invention adopts an improved NSGA-II delivery route optimization method and introduces an adaptive evolution strategy into the NSGA-II algorithm, which is conducive to forming a corresponding operator probability distribution according to the characteristics of the population itself during the population evolution process, better guiding the population evolution, thereby solving the delivery route optimization problem and outputting a Pareto optimal solution set that decision makers can choose from.

[0008] To achieve the above object, the technical solution adopted by the present invention is a logistics route optimization method based on adaptive NSGA-II, comprising the following steps:

[0009] S1: Analyze and establish basic mathematical models.

[0010] S11: In the actual delivery process, the vehicle will incur a certain amount of fuel consumption. Moreover, the weight of the vehicle, the length of the route, and different vehicle models all greatly affect the amount of fuel used. On a route, we assume that the same vehicle is used. For route i, the monetary cost C generated by logistics distribution i Including the delivery cost and fuel consumption of the vehicle, the calculation formula is as follows:

[0011] C i =V i +R i ×(O i +(W i ×α i )) (1)

[0012] Among them, V i is the dispatch cost of route i, R i is the total distance of branch route i, O i is the fuel consumption cost of route i, W i is the total load of branch route i, α i is the fuel consumption cost coefficient per ton per kilometer of route i.

[0013] Let k(x) be the number of routes in the logistics distribution plan x. Then the total monetary cost C(x) generated by plan x is:

[0014]

[0015] S12: On each route, we consider the time cost caused by the unloading time required at each delivery point and the vehicle travel time. Let P i is the total number of delivery points for route i, R i is the total distance of branch route i, S i is the average speed of vehicles traveling on the road section, U i The unloading time for each distribution point. For route i, the time cost of logistics distribution is T i The calculation formula is as follows:

[0016]

[0017] Then the time cost T(x) generated by the delivery plan x is:

[0018]

[0019] S13: Evaluate the superiority of this system by minimizing both monetary cost and time cost, which is defined as follows:

[0020] minF(x)=(C(x),T(x)) T (5)

[0021] Assume that the dispatch cost of large trucks is vb, and the dispatch cost of small trucks is vl; the maximum load weight of large trucks is wb, and the maximum load weight of small trucks is wl; the fuel consumption cost coefficient per kilometer and per ton of large trucks is αb, and the fuel consumption cost coefficient per kilometer and per ton of small trucks is αl; the unloading time of large trucks at each distribution point is ub, and the unloading time of small trucks at each distribution point is ul; the average speed of large trucks is sb, and the average speed of small trucks is sl; the fuel consumption cost per kilometer of large trucks is ob, and the fuel consumption cost per kilometer of small trucks is ol.

[0022] Therefore, the model is established as follows:

[0023] minF(x)=(C(x),T(x)) T (6)

[0024]

[0025] Where X is the individual decision space, k(x) is the number of sub-routes in option x, and R i is the total distance of branch route i, P i is the total number of delivery points for route i, W i is the total cargo volume of route i. ③ indicates that for route i, no matter what type of vehicle, the load cannot be greater than the maximum load of its corresponding vehicle. ④~⑧ indicate that on route i, the corresponding vehicle model information is assigned to the vehicles used. ⑨ indicates the maximum driving distance of the vehicle corresponding to route i.

[0026] S2: Design the chromosome coding of the individual population. The present invention adopts a fixed-length non-repeating integer coding method. Each chromosome gene represents a corresponding location to visit a customer point, and the chromosome length M is the total number of customers to be visited.

[0027] S3: Generate an initial population of size N by combining random and greedy methods, and perform Pareto non-dominated sorting on the population. The steps of non-dominated sorting are S31 to S35.

[0028] S31: Calculate the dominance count n of all individuals i (the number of all individuals that dominate individual i) and calculate the dominating solution set S i (The set of individuals dominated by individual i).

[0029] S32: Find the non-dominated individuals in the population, that is, n i= 0, put the non-dominated individuals into the set F1.

[0030] S33: For each individual in F1, find the set of individuals S dominated by each individual in the set i , for S i Individual l in n l Subtract 1 from the original value, that is, n l =n l -1, store individual l in set H.

[0031] S34: Define the set F1 as the first-level non-dominated set, and label each individual in F1 with the same non-dominated sequence i rank .

[0032] S35: For the individuals in set H, follow the above steps S32, S33 and S34 until all individuals are stratified.

[0033] S4: The next generation population is obtained through the three evolutionary operations of selection, crossover and mutation. The specific steps of the evolutionary operations of selection, crossover and mutation are shown in S41 to S43.

[0034] S41: In the selection operation, a bidding competition is used to select excellent population individuals with low Pareto levels (i.e., better Pareto dominance) and small crowding (i.e., large crowding distance), and save them to an elite archive set of fixed size.

[0035] S42: In the crossover operation, the present invention introduces an adaptive strategy, with a crossover probability that changes adaptively with the number of iterations, selects a crossover operation from a crossover operator pool composed of multiple crossover operators according to the operator selection probability distribution, and performs feedback adjustment on the operator selection weight according to the result of this crossover operation. The crossover operator description is shown in S421 to S425.

[0036] S421: Partial Mapped Crossover (PMC), first randomly select the start and end positions of the exchange fragment, exchange the corresponding start and end position fragments of the parent chromosome, and replace the conflicting gene points with each other.

[0037] S422: Order Crossover (OC), first randomly select the starting and ending positions of the continuous fragments, copy the corresponding fragments of parent 1 to the corresponding positions of offspring 1, then select the genes missing from offspring 1 from parent 2 and add them to offspring 1 according to the order in parent 1, and offspring 2 is generated in a similar way.

[0038] S423: Position-Based Crossover (PBC), first randomly select multiple discontinuous gene point positions, copy the corresponding gene points of parent 1 to the corresponding positions of child individual 1, then select the genes missing from child 1 from parent 2 and add them to child 1 in the order of parent 1, and the generation method of child 2 is similar.

[0039] S424: Cycle Crossover (CC), first randomly select a gene from parent 1, then find the gene at the corresponding position in parent 2, then return to parent 1 to find the gene at the corresponding gene position in parent 2, and repeat this operation until a closed loop is formed. The genes contained in the closed loop are copied to the same position in offspring 1, and the remaining gene points are selected sequentially from parent 2 to form offspring 2 in the same way.

[0040] S425: Subtour Exchange Crossover (SEC), first randomly select a group of consecutive genes from parent 1 and replace the corresponding gene groups in parent 2 in the original order.

[0041] S43: In the mutation operation, the same method as the adaptive crossover strategy is adopted, with the mutation probability adaptively changing with the number of iterations, the mutation operation is selected from the operator pool composed of multiple mutation operators according to the operator selection probability distribution, and the operator selection weight is feedback-adjusted according to the result of this mutation operation. The description of the mutation operator is shown in S431 to S433.

[0042] S431: Exchange Mutation (EM), randomly selects two gene point positions on the chromosome and exchanges their values.

[0043] S432: Inversion Mutation (IM), randomly selects a group of consecutive gene segments of the chromosome and reverses their order.

[0044] S433: Random Mutation (RM), randomly selects a group of consecutive gene fragments of the chromosome and disrupts their order.

[0045] S44: Operator adaptive feedback mechanism, for each new generation of individuals generated by crossover and mutation operations, the weight ratio of the selected operator is adjusted by observing its individual quality: if the offspring is better than two parents and finally enters the elite archive set, the weight is increased by 4 points; if the offspring is better than one parent and enters the elite archive set, the weight is increased by 3 points; if the offspring is better than two parents but does not enter the elite archive set, the weight is increased by 2 points; if the offspring is better than one parent but does not enter the elite archive set, the weight is increased by 1 point; the weights of other cases remain unchanged. Finally, all operator weights are normalized.

[0046] S5: Merge the parent population and the child population to form a new population, and then perform fast non-dominated sorting on the new population, and calculate the crowding degree of the individuals in each non-dominated layer. According to the non-dominated relationship and the crowding degree of the individuals, select the better individuals to form a new parent population;

[0047] The congestion degree is calculated as shown in steps S51 to S54.

[0048] S51: Initialization, n d =0,n∈1,…N

[0049] S52: For each objective function f m , sort the individuals of this level according to the objective function, record is the individual objective function value f m The maximum value of is the individual objective function value f m The minimum value of .

[0050] S53: The congestion degree of the two boundaries after sorting is 1 d and N d Set to ∞.

[0051] S54: Calculate the congestion degree:

[0052]

[0053] S6: Generate a new offspring population through population evolution operator operation, and repeat steps S41 to S44.

[0054] S7: If the population iteration termination condition is not met, return to S4 and S5 until the termination condition is met.

[0055] S8: The adaptive NSGAⅡ algorithm outputs the optimal solution set. Each solution in the solution set represents a customer point visit sequence. An auxiliary array is used to store the starting delivery point of each delivery vehicle, thereby obtaining the delivery route for a vehicle.

[0056] The system for implementing the logistics route optimization method based on adaptive NSGAII of the present invention comprises: a basic mathematical model analysis and establishment module, a population individual chromosome coding design module, an initial population generation and Pareto non-dominated sorting module, a next generation population generation module, a new parent population generation module, a new child population generation module, a population iteration cycle module, and a path generation module, which are connected in sequence, wherein:

[0057] Basic mathematical model analysis and establishment module, analyze and establish basic mathematical models;

[0058] The population individual chromosome coding design module designs the population individual chromosome coding, adopts a fixed-length non-repeating integer coding method, each chromosome gene represents the corresponding position of the customer visit point, and the chromosome length M is the total number of customers to be visited;

[0059] The initial population generation and Pareto non-dominated sorting module uses a combination of random and greedy methods to generate an initial population of size N, and performs Pareto non-dominated sorting on the population;

[0060] The next generation population generation module obtains the next generation population through three evolutionary operations: selection, crossover, and mutation;

[0061] The new parent population generation module merges the parent population with the child population to form a new population, and then performs fast non-dominated sorting on the new population, while calculating the crowding degree of the individuals in each non-dominated layer; based on the non-dominated relationship and the crowding degree of the individuals, the better individuals are selected to form a new parent population;

[0062] The new offspring population generation module generates a new offspring population through population evolution operator operation, and repeats the operation process of the next generation population generation module;

[0063] The population iteration loop module, if the population iteration termination condition is not met, returns to the next generation population generation module and the new parent generation population generation module until the termination condition is met;

[0064] The path generation module generates an optimized logistics route based on the calculation results of the population iteration loop module. The adaptive NSGAⅡ algorithm outputs the optimal solution set. Each solution in the solution set represents a customer point visit sequence. An auxiliary array is used to store the starting delivery point of each delivery vehicle, thereby obtaining the delivery route for a vehicle.

[0065] Compared with the prior art, the present invention has the following advantages:

[0066] (1) Optimize the monetary cost and time cost in parallel, use the adaptive NSGAⅡ multi-objective optimization algorithm to solve the distribution route optimization problem, drive the algorithm design and solution based on actual needs, and provide decision makers with a variety of high-quality solutions.

[0067] (2) In view of the shortcomings of the NSGAⅡ multi-objective optimization algorithm, the present invention proposes an adaptive operator selection strategy to improve the single fixed evolutionary operator of NSGAⅡ, so that different populations can adaptively select better evolutionary operators according to their own characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of the adaptive NSGA-II algorithm of the method of the present invention.

[0069] Figure 2is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0070] In order to better understand the present invention, the present invention is described in detail below with reference to the accompanying drawings and examples.

[0071] Examples:

[0072] A logistics center has 5 small delivery vehicles and 3 large delivery vehicles. The maximum load of small vehicles is 4T, the fixed dispatch cost is 100 yuan / vehicle, the average fuel consumption cost per kilometer is 0.56 yuan, the average unloading time is 10 minutes / vehicle, the average driving speed is 50 kilometers / hour, the load fuel consumption increase coefficient is 0.1 yuan / ton*kilometer, and the maximum single delivery distance is 50 kilometers; the maximum load of large vehicles is 8T, the fixed dispatch cost is 150 yuan / vehicle, the average fuel consumption cost per kilometer is 0.89 yuan, the average unloading time is 18 minutes / vehicle, the average driving speed is 30 kilometers / hour, the load fuel consumption increase coefficient is 0.06 yuan / ton*kilometer, and the maximum single delivery distance is 80 kilometers. It is necessary to deliver goods to 20 customers. Among them, the coordinates of the logistics center are (14.2km, 13.1km), and it is required to arrange the delivery routes of the vehicles reasonably to minimize the money cost and time cost.

[0073]

[0074]

[0075] S1: Analyze and build models based on examples.

[0076] S2: Design the chromosome coding of the individual population. The present invention adopts a fixed-length non-repeating integer coding method. Each chromosome gene represents a corresponding location visit customer point, and the chromosome length M=20.

[0077] S3: 50% random and 50% greedy methods are used to generate an initial population of size N=300, and a non-dominated sort is performed on the initial population. The steps of the non-dominated sort are S31 to S35.

[0078] S31: Calculate the dominance count n of all individuals i (the number of all individuals that dominate individual i) and calculate the dominating solution set S i (The set of individuals dominated by individual i).

[0079] S32: Find the non-dominated individuals in the population, that is, n i = 0, put the non-dominated individuals into the set F1.

[0080] S33: For each individual in F1, find the set of individuals S dominated by each individual in the set i, for S i Individual l in n l Subtract 1 from the original value, that is, n l =n l -1, store individual l in set H.

[0081] S34: Define the set F1 as the first-level non-dominated set, and label each individual in F1 with the same non-dominated sequence i rank .

[0082] S35: For the individuals in set H, follow the above steps S32, S33 and S34 until all individuals are stratified.

[0083] S4: The next generation population is obtained through the three evolutionary operations of selection, crossover and mutation. The specific steps of the evolutionary operations of selection, crossover and mutation are shown in S41 to S43.

[0084] S41: In the selection operation, a bidding competition is used to select excellent population individuals with low Pareto level (i.e., better Pareto dominance) and small crowding degree (i.e., large crowding distance), and save them into an elite archive set of size 20.

[0085] S42: In the crossover operation, a crossover operation is selected from a crossover operator pool composed of multiple crossover operators according to the operator selection probability distribution with a crossover probability that changes adaptively with the number of iterations, and the operator selection weight is feedback-adjusted according to the result of this crossover operation. The crossover operator description is shown in S421 to S425.

[0086] S421: Partial Mapped Crossover (PMC), first randomly select the start and end positions of the exchange fragment, exchange the corresponding start and end position fragments of the parent chromosome, and replace the conflicting gene points with each other.

[0087] S422: Order Crossover (OC), first randomly select the starting and ending positions of the continuous fragments, copy the corresponding fragments of parent 1 to the corresponding positions of offspring 1, then select the genes missing from offspring 1 from parent 2 and add them to offspring 1 according to the order in parent 1, and offspring 2 is generated in a similar way.

[0088] S423: Position-Based Crossover (PBC), first randomly select multiple discontinuous gene point positions, copy the corresponding gene points of parent 1 to the corresponding positions of child individual 1, then select the genes missing from child 1 from parent 2 and add them to child 1 in the order of parent 1, and the generation method of child 2 is similar.

[0089] S424: Cycle Crossover (CC), first randomly select a gene from parent 1, then find the gene at the corresponding position in parent 2, then return to parent 1 to find the gene at the corresponding gene position in parent 2, and repeat this operation until a closed loop is formed. The genes contained in the closed loop are copied to the same position in offspring 1, and the remaining gene points are selected sequentially from parent 2 to form offspring 2 in the same way.

[0090] S425: Subtour Exchange Crossover (SEC), first randomly select a group of consecutive genes from parent 1 and replace the corresponding gene groups in parent 2 in the original order.

[0091] S43: In the mutation operation, the same method as the adaptive crossover strategy is adopted, with the mutation probability adaptively changing with the number of iterations, the mutation operation is selected from the operator pool composed of multiple mutation operators according to the operator selection probability distribution, and the operator selection weight is feedback-adjusted according to the result of this mutation operation. The description of the mutation operator is shown in S431 to S433.

[0092] S431: Exchange Mutation (EM), randomly selects two gene point positions on the chromosome and exchanges their values.

[0093] S432: Inversion Mutation (IM), randomly selects a group of consecutive gene segments of the chromosome and reverses their order.

[0094] S433: Random Mutation (RM), randomly selects a group of consecutive gene fragments of the chromosome and disrupts their order.

[0095] S44: Operator adaptive feedback mechanism, for each new generation of individuals generated by crossover and mutation operations, the weight ratio of the selected operator is adjusted by observing its individual quality: if the offspring is better than two parents and finally enters the elite archive set, the weight is increased by 4 points; if the offspring is better than one parent and enters the elite archive set, the weight is increased by 3 points; if the offspring is better than two parents but does not enter the elite archive set, the weight is increased by 2 points; if the offspring is better than one parent but does not enter the elite archive set, the weight is increased by 1 point; the weights of other cases remain unchanged. Finally, all operator weights are normalized.

[0096] S5: Merge the parent population and the child population to form a new population, and then perform fast non-dominated sorting on the new population, and calculate the crowding degree of the individuals in each non-dominated layer. According to the non-dominated relationship and the crowding degree of the individuals, select the better individuals to form a new parent population;

[0097] The congestion degree is calculated as shown in steps S51 to S54.

[0098] S51: Initialization, n d =0, n∈1,…N

[0099] S52: For each objective function f m , sort the individuals of this level according to the objective function, record is the individual objective function value f m The maximum value of is the individual objective function value f m The minimum value of .

[0100] S53: The congestion degree of the two boundaries after sorting is 1 d and N d Set to ∞.

[0101] S54: Calculate the congestion degree:

[0102]

[0103] S6: Generate a new offspring population through population evolution operator operation, and repeat steps S41 to S44.

[0104] S7: If the population iteration termination condition is not met, return to S4 and S5 until the termination condition is met.

[0105] S8: The adaptive NSGAⅡ algorithm outputs the optimal solution set. Each solution in the solution set represents a customer point visit sequence. An auxiliary array is used to store the starting delivery point of each delivery vehicle, thereby obtaining the delivery route for a vehicle.

[0106] The present invention adopts an improved NSGA-II delivery route optimization method and introduces an adaptive evolution strategy into the NSGA-II algorithm, which is conducive to forming a corresponding operator probability distribution according to the characteristics of the population itself during the population evolution process, better guiding the population evolution, and outputting a Pareto optimal solution set that decision makers can choose from, so as to better solve the delivery route optimization problem.

[0107] The system for implementing the logistics route optimization method based on adaptive NSGAII of the present invention comprises: a basic mathematical model analysis and establishment module, a population individual chromosome coding design module, an initial population generation and Pareto non-dominated sorting module, a next generation population generation module, a new parent population generation module, a new child population generation module, a population iteration cycle module, and a path generation module, which are sequentially connected. The above modules correspond to the contents of steps S1 to S8 of the method of the present invention in sequence.

[0108] Although the specific implementation process of the present invention is described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific implementation schemes. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A logistics route optimization method based on adaptive NSGAII, characterized by: The following steps are involved: S1: Analyze and establish basic mathematical models; S11: On a route, assuming the same vehicle is used, for route i, the monetary cost C of logistics distribution i Including the vehicle's delivery cost and fuel consumption, the calculation formula is as follows: C i =V i +R i ×(O i +(W i ×α i )) Among them, V i is the dispatch cost of route i, R i is the total distance of branch route i, O i is the fuel consumption cost of route i, W i is the total load of branch route i, α i is the fuel consumption cost coefficient per kilometer and per ton of route i; Let k(x) be the number of routes in the logistics distribution plan x; then the total monetary cost C(x) generated by plan x is: S12: On each route, consider the time cost caused by the unloading time required for each delivery point and the vehicle travel time; let P i is the total number of delivery points for route i, R i is the total distance of branch route i, S i is the average speed of vehicles traveling on the route, U i is the unloading time for each distribution point; for route i, the time cost of logistics distribution is T i The calculation formula is as follows: Then the time cost T(x) generated by the delivery plan x is: S13: Evaluate the superiority of this system by minimizing both monetary cost and time cost, which is defined as follows: minF(x)=(C(x),T(x)) T Assume that the dispatching cost of a large truck is vb, and the dispatching cost of a small truck is vl; the maximum load of a large truck is wb, and the maximum load of a small truck is wl; the fuel consumption cost coefficient per kilometer and per ton of a large truck is αb, and the fuel consumption cost coefficient per kilometer and per ton of a small truck is αl; the unloading time of a large truck at each distribution point is ub, and the unloading time of a small truck at each distribution point is ul; the average speed of a large truck is sb, and the average speed of a small truck is sl; the fuel consumption cost per kilometer of a large truck is ob, and the fuel consumption cost per kilometer of a small truck is ol; Therefore, the model is established as follows: minF(x)=(C(x),T(x)) T Where X is the individual decision space, k(x) is the number of sub-routes in option x, and R i is the total distance of branch route i, P i is the total number of delivery points for route i, W i is the total cargo volume of route i; Formula ③ indicates that for route i, no matter what type of vehicle, the load capacity cannot be greater than the maximum load capacity of the corresponding vehicle; Formulas ④ to ⑧ indicate that the corresponding vehicle type information is assigned to the vehicles on route i; Formula 9 indicates the maximum driving distance of the vehicle corresponding to route i; S2: Design the chromosome coding of the population individuals, using a fixed-length non-repeating integer coding method. Each chromosome gene represents the corresponding location of the customer visit point, and the chromosome length M is the total number of customers to be visited; S3: Generate an initial population of size N by combining random and greedy methods, and perform Pareto non-dominated sorting on the population; wherein the steps of non-dominated sorting are S31 to S35; S31: Calculate the dominance count n of all individuals i , that is, the number of all individuals that dominate individual i, and calculate the dominating solution set S i , that is, the set of individuals dominated by individual i; S32: Find the non-dominated individuals in the population, that is, n i = 0, put the non-dominated individuals into the set F1; S33: For each individual in F1, find the set of individuals S dominated by each individual in the set i , for S i Individual l in n l Subtract 1 from the original value, that is, n l =n l -1, store individual l in set H; S34: Define the set F1 as the first-level non-dominated set, and label each individual in F1 with the same non-dominated sequence i rank ; S35: for the individuals in the set H, follow the above steps S32, S33 and S34 until all individuals are stratified; S4: The next generation population is obtained through the three evolutionary operations of selection, crossover and mutation; the specific steps of the evolutionary operations of selection, crossover and mutation are shown in S41 to S43; S41: In the selection operation, a competitive bidding competition is used to select excellent individuals of the population with low Pareto level, i.e., better Pareto dominance, and low crowding degree, i.e., large crowding distance, and save them to an elite archive set of fixed size; S42: In the crossover operation, an adaptive strategy is introduced to select a crossover operation from a crossover operator pool consisting of multiple crossover operators according to the operator selection probability distribution with a crossover probability that changes adaptively with the number of iterations, and feedback adjustment is performed on the operator selection weight according to the result of this crossover operation; wherein the crossover operator description is shown in S421 to S425; S421: Partial matching crossover, first randomly select the start and end positions of the exchange fragment, exchange the fragments of the corresponding start and end positions of the parent chromosome, and replace the conflicting gene points with each other; S422: Sequential crossover, first randomly select the start and end positions of the continuous segments, copy the corresponding segments of parent 1 to the corresponding positions of offspring 1, then select the genes missing from offspring 1 from parent 2 and add them to offspring 1 in the order of parent 1, and generate offspring 2 in the same way; S423: Based on positional crossover, first randomly select multiple discontinuous gene point positions, copy the corresponding gene points of parent 1 to the corresponding positions of child individual 1, then select the genes missing from child 1 from parent 2 and add them to child 1 in the order of parent 1, and generate child 2 in the same way; S424: Circular crossover, first randomly select a gene from parent 1, then find the gene at the corresponding position in parent 2, then return to parent 1 to find the gene at the corresponding gene position in parent 2, repeat this operation until a closed loop is formed, the genes included in the closed loop are copied to the same position in offspring 1, and the remaining gene points are selected sequentially from parent 2, and offspring 2 is formed in the same way; S425: Subpath exchange crossover, first randomly select a set of consecutive genes from parent 1 and replace the corresponding gene set in parent 2 in the original order; S43: In the mutation operation, the same method as the adaptive crossover strategy is adopted, with the mutation probability adaptively changing with the number of iterations, the mutation operation is selected from the operator pool composed of multiple mutation operators according to the operator selection probability distribution, and the operator selection weight is feedback-adjusted according to the result of this mutation operation; wherein the mutation operator description is shown in S431 to S433; S431: Exchange mutation, randomly select two gene point positions on the chromosome and exchange their values; S432: Inversion mutation, a set of consecutive gene segments of a chromosome are randomly selected and their order is reversed; S433: scramble mutation, randomly selects a group of consecutive gene segments of the chromosome and disrupts their order; S44: Operator adaptive feedback mechanism, for each new generation of individuals generated by crossover and mutation operations, the weight ratio of the selected operator is adjusted by observing its individual quality: if the offspring is better than two parents and finally enters the elite archive set, the weight is increased by 4 points; if the offspring is better than one parent and enters the elite archive set, the weight is increased by 3 points; if the offspring is better than two parents but does not enter the elite archive set, the weight is increased by 2 points; if the offspring is better than one parent but does not enter the elite archive set, the weight is increased by 1 point; the weights of other cases remain unchanged; finally, the weights of all operators are normalized; S5: The parent population and the child population are merged to form a new population, and then the new population is quickly non-dominated sorted, and the crowding degree of the individuals in each non-dominated layer is calculated; according to the non-dominated relationship and the crowding degree of the individuals, the better individuals are selected to form a new parent population; The congestion degree is calculated as shown in steps S51 to S54: S51: Initialization, n d =0,n∈1,…N S52: For each objective function f m , sort the individuals of the population according to the objective function, record is the individual objective function value f m The maximum value of is the individual objective function value f m The minimum value of S53: The congestion degree of the two boundaries after sorting is 1 d and N d Set to ∞; S54: Calculate the congestion degree: S6: Generate a new offspring population through population evolution operator operation, and repeat steps S41 to S44; S7: If the population iteration termination condition is not met, return to steps S4 and S5 until the termination condition is met; S8: The adaptive NSGAⅡ algorithm outputs the optimal solution set. Each solution in the solution set represents a customer point visit sequence. An auxiliary array is used to store the starting delivery point of each delivery vehicle, thereby obtaining the delivery route of a vehicle.

2. A system for implementing the logistics route optimization method based on adaptive NSGAII as described in claim 1, characterized in that: It includes a basic mathematical model analysis and establishment module, a population individual chromosome coding design module, an initial population generation and Pareto non-dominated sorting module, a next generation population generation module, a new parent population generation module, a new child population generation module, a population iteration cycle module, and a path generation module, wherein: Basic mathematical model analysis and establishment module, analyze and establish basic mathematical models; The population individual chromosome coding design module designs the population individual chromosome coding, adopts a fixed-length non-repeating integer coding method, each chromosome gene represents the corresponding position of the customer visit point, and the chromosome length M is the total number of customers to be visited; The initial population generation and Pareto non-dominated sorting module uses a combination of random and greedy methods to generate an initial population of size N, and performs Pareto non-dominated sorting on the population; The next generation population generation module obtains the next generation population through three evolutionary operations: selection, crossover, and mutation; The new parent population generation module merges the parent population with the child population to form a new population, and then performs fast non-dominated sorting on the new population, while calculating the crowding degree of the individuals in each non-dominated layer; based on the non-dominated relationship and the crowding degree of the individuals, the better individuals are selected to form a new parent population; The new offspring population generation module generates a new offspring population through population evolution operator operation, and repeats the operation process of the next generation population generation module; The population iteration loop module, if the population iteration termination condition is not met, returns to the next generation population generation module and the new parent generation population generation module until the termination condition is met; The path generation module generates an optimized logistics route based on the calculation results of the population iteration loop module. The adaptive NSGAⅡ algorithm outputs the optimal solution set. Each solution in the solution set represents a customer point visit sequence. An auxiliary array is used to store the starting delivery point of each delivery vehicle, thereby obtaining the delivery route of a vehicle.

Images