CN114118926B - Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder - Google Patents
Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder Download PDFInfo
- Publication number
- CN114118926B CN114118926B CN202111451538.0A CN202111451538A CN114118926B CN 114118926 B CN114118926 B CN 114118926B CN 202111451538 A CN202111451538 A CN 202111451538A CN 114118926 B CN114118926 B CN 114118926B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- generation
- auxiliary
- task
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000013507 mapping Methods 0.000 claims abstract description 22
- 230000001174 ascending effect Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 11
- 210000000349 chromosome Anatomy 0.000 claims description 6
- 230000002068 genetic effect Effects 0.000 claims description 6
- 108091026890 Coding region Proteins 0.000 claims description 4
- 230000035772 mutation Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 238000002716 delivery method Methods 0.000 claims 2
- 239000011159 matrix material Substances 0.000 description 9
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
- G06Q10/0835—Relationships between shipper or supplier and carriers
- G06Q10/08355—Routing methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder, which is applied to a warehouse point and N distribution task pointsIn the distribution environment, the method comprises the following steps: 1. building a vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help The method comprises the steps of carrying out a first treatment on the surface of the 2. Solving the task model f by using the main population and the auxiliary population respectively origin And an auxiliary model f help Establishing a mapping relation between the auxiliary population and the main population by using the denoising self-encoder model; 3. on the one hand, during evolution, converting an optimal distribution scheme of the auxiliary population through a trained learning model, and putting the converted distribution scheme into the main population; on the other hand, the optimal distribution scheme of the main population is directly put into the auxiliary population, and the above processes are repeated until the algorithm is terminated; 4. and outputting the optimal distribution scheme under the given task. The invention can optimize the vehicle distribution path and the vehicle load variance at the same time, thereby improving the distribution efficiency.
Description
Technical Field
The invention belongs to the field of vehicle path optimization, and particularly relates to a vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder.
Background
Along with the rapid development of Internet economy, electronic commerce gradually goes into life of people, and online shopping becomes an important component of life of people. The last kilometer of the express service industry is not only related to the online shopping satisfaction degree of consumers, but also occupies more than 50% of the total cost of logistics; in addition, with the gradual perfection of labor methods and the pursuit of people for good life, the additional extension of working time is excluded by people. Therefore, how to optimize the logistics transportation path to minimize the path and make the labor intensity of the drivers for transporting goods consistent is a problem to be solved by each big logistics company.
In the context of vehicle path problems, it is assumed that a warehouse stores a collection of vehicles of equal transport capacity, size, model, scale, etc., that serve all distribution mission points within an area; each distribution task point can only be served by one vehicle and is not repeatable, but the vehicle can access a plurality of distribution task points on the premise of meeting the capacity; not only is a multi-objective optimal vehicle access path given that all delivery mission point requirements are met, but the capacity between these vehicles is balanced.
When solving the vehicle path problem with the conventional method, only the goal of the shortest total vehicle path is considered, and the more realistic problem of unbalanced workload is ignored, which helps the vehicle driver to better service the customer during subsequent work. In the prior art, only a single target is optimized, or two targets are converted into a single target in a weighted mode, so that the calculation load is increased, and the targets with the shortest path distance and balanced vehicle loading capacity cannot be obtained at last, thereby reducing the distribution efficiency.
Disclosure of Invention
The invention aims to solve the defects of the prior art, and provides a vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder, so that a vehicle transportation distribution route and variances of the vehicle loads can be optimized at the same time, the requirements of shortest driving distance and balanced vehicle loads are met, and distribution efficiency can be improved.
In order to achieve the aim of the invention, the invention adopts the following technical scheme:
the invention relates to a vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder, which is characterized by being applied to a distribution environment consisting of a warehouse point and N distribution task points and being carried out according to the following steps:
s1, building a vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help ；
S1.1, acquiring relevant data of a vehicle distribution task, wherein the relevant data comprises the following steps: distribution warehouse points and distribution task points, task demand and maximum vehicle capacity;
s1.2, respectively constructing a vehicle distribution task model f by using the formula (1) and the formula (2) origin Is a function of the objective function of:
in the formula (1), F 1 C is the length of the total travel path ij Representing the distance between the ith delivery task point and the jth delivery task point, i, j E [0, N]The method comprises the steps of carrying out a first treatment on the surface of the When i=0, c 0j Representing the distance from the warehouse point to the jth delivery task point, c when j=0 i0 Representing the distance, x, from the ith delivery task point to the warehouse point ij Indicating whether a path exists between the ith delivery task point and the jth delivery task point, if x ij If =1, it indicates that there is a path, if x ij =0, then no path exists;
in the formula (2), F 2 For the variance of the vehicle load, M is the total number of vehicles currently in use, d m Indicating the total amount of loading of the mth vehicle,representing the average load of the vehicle;
s1.3, constructing a vehicle distribution task model f by using the formulas (3) - (9) origin Is a constraint on (c):
equation (3) indicates that each dispensing task point can only be driven in by one vehicle;
equation (4) indicates that each dispensing task point can only be driven out by one vehicle;
equation (5) represents that the number of vehicles driving into the warehouse is at most K; k represents the maximum number of vehicles allowed to be used;
equation (6) represents that the number of vehicles that leave the warehouse is at most K;
equation (7) represents a loop path cancellation constraint equation; wherein u is i Representing the load of the vehicle after the ith delivery task point of the vehicle service; u (u) j Representing the load capacity of the vehicle after the j-th delivery task point is serviced; d represents a unified maximum load of the vehicle; q i Representing the demand of the ith delivery task point; q j Representing the demand of the j-th delivery task point;
equation (8) indicates that the total capacity of the vehicle after servicing the ith delivery mission point should be less than the maximum load D;
formula (9) represents a variable x ij Only the value 1 or 0 can be taken.
S1.4, constructing a vehicle distribution task auxiliary model f taking the formula (1) as an objective function and taking the formulas (3), (4) and (7) as constraint conditions help ；
S2, respectively according to the vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help Constructing a main population allocation scheme and an auxiliary population allocation scheme:
enabling the chromosome coding sequence of each individual in the main population and the auxiliary population to represent the vehicle access sequence of N distribution task points, wherein 0 represents a warehouse point and the population scale is P;
randomly initializing each individual in the main population and the auxiliary population, and correspondingly serving as an initial main population allocation scheme and an initial auxiliary population allocation scheme;
s3, setting the number of neurons in the denoising self-encoder model DAE as N; taking an initial auxiliary population allocation scheme as input of the denoising self-encoder model DAE, taking an initial main population allocation scheme as a label of the denoising self-encoder model DAE, and training the denoising self-encoder model DAE to obtain an initial mapping relation model of an auxiliary population and a main population;
s4, defining the current iteration times as t, initializing t=1, and taking an initial auxiliary population distribution scheme and an initial main population distribution scheme as a t generation auxiliary vehicle distribution scheme and a t generation main vehicle distribution scheme respectively; taking the initial mapping relation model as a t generation mapping relation model;
s5, solving a t generation auxiliary vehicle distribution scheme by adopting a genetic algorithm based on the t generation main vehicle distribution scheme to obtain a t+1 generation auxiliary vehicle distribution scheme, solving the t generation main vehicle distribution scheme by adopting a model and a genetic algorithm to obtain a t+1 generation main vehicle distribution scheme, taking the t+1 generation auxiliary vehicle distribution scheme as the input of the denoising self-encoder model DAE, taking the t+1 generation main vehicle distribution scheme as the label of the denoising self-encoder model DAE, training the denoising self-encoder model DAE, and taking the trained mapping relation model as a t+1 generation mapping relation model;
s6, after the value of t+1 is assigned to T, judging whether the current iteration number T reaches the maximum iteration number T max If yes, output the T max Replacing a vehicle distribution scheme and taking the vehicle distribution scheme as an optimal vehicle distribution scheme; otherwise, step S5 is performed.
The vehicle task distribution method of the invention is also characterized in that the t+1st generation auxiliary vehicle distribution scheme in the step S5 is obtained according to the following process:
s5.1a, selecting two individuals from the t-th generation auxiliary vehicle distribution scheme as t-th generation auxiliary parents using the tournament; then, the t-generation auxiliary parent is crossed by adopting OX to obtain a t-generation auxiliary child, and then the t-generation auxiliary child is improved by utilizing a local search operator, so that the improved t-generation auxiliary child is stored in a t+1th-generation temporary auxiliary vehicle distribution scheme;
s5.2a, slave-to-vehicle delivery task model f origin Each individual in the t-th generation main vehicle distribution scheme is sequenced in an ascending order according to the selected objective function value, and the individual corresponding to the selected minimum objective function value is used as the t-th generation optimal main vehicle distribution scheme and stored in the t+1th generation temporary auxiliary vehicle distribution scheme;
s5.3a, assisting the model f in terms of vehicle delivery tasks for each individual in the t+1st generation temporary assistance vehicle delivery regime help The objective function values of the vehicles are sorted in ascending order, and the first P delivery schemes with smaller objective function values are selected to form a t+1st generation auxiliary vehicle delivery scheme.
The t+1st generation host vehicle delivery schedule in step S5 is obtained as follows:
s5.1b, selecting two individuals from the t-th generation host vehicle distribution scheme as t-th generation parents using the tournament;
s5.2b, generating the t-th random number rand t If random number rand t Less than threshold delta 1 The father generation of the t generation is improved by adopting RBX crossing and mutation operators, and the child generation of the t generation is obtained and then stored in the temporary host vehicle distribution scheme of the t+1th generation; otherwise, directly storing the t generation father generation into the t+1 generation temporary host vehicle distribution scheme;
each individual in the S5.3b, t+1st generation auxiliary vehicle delivery scheme delivers the task auxiliary model f according to the vehicle help The objective function values of the (1) are sequenced in an ascending order, and an individual with the smallest objective function value is input into a t generation mapping relation model, so that t+1 generation offspring are obtained and stored in a t+1 generation temporary main vehicle distribution scheme;
s5.4b, slave vehicle delivery task model f origin An objective function is selected, each individual in the t+1st temporary host vehicle distribution scheme is sequenced in an ascending order according to the selected objective function value, and a first sequenced individual is obtained;
distributing task model f from vehicle origin And carrying out ascending sort on each individual in the t+1st temporary host vehicle distribution scheme according to the selected objective function value to obtain a second sorted individual;
and selecting individuals with the first P serial numbers from the first ordered individuals and the second ordered individuals, sequentially adding the individuals with the first P serial numbers into a t+1th generation host vehicle distribution scheme, and if the individuals with the first P serial numbers are less than P, sequentially selecting the individuals with the first serial numbers from the first ordered individuals, and respectively supplementing the rest individuals in the t+1th generation host vehicle distribution scheme.
Compared with the prior art, the invention has the beneficial effects that:
1. according to the invention, by constructing the model which is more in line with the actual vehicle path problem, introducing an auxiliary population, constructing the corresponding relation between the auxiliary population and the main population by adopting a method of training a denoising self-encoder model, fitting a historical evolution track, and the optimal solution distribution scheme predicted by the model can accelerate population convergence, so that the distribution efficiency is improved;
2. the method is beneficial to balancing convergence and diversity from a global angle, helping a main population jump out of local optimum, increasing population diversity, guiding the main population to optimize towards the optimum direction, avoiding the problem of wasting evaluation times caused by slow algorithm iteration, helping enterprises save time cost and making reasonable decisions as soon as possible;
3. according to the invention, by considering the two targets at the same time, a group of optimal allocation schemes in all aspects are obtained, different choice conditions among the targets are reflected, and the logistics enterprises can select a proper implementation scheme according to actual conditions, so that the aims of saving transportation cost and balancing labor time are realized, and the enterprises can formulate longer-term development planning.
Drawings
FIG. 1 is a schematic diagram of the method for converting population solution sequence information into a distance matrix;
FIG. 2 is a schematic diagram of the invention for establishing the correspondence between the auxiliary population and the main population based on the denoising self-encoder model;
FIG. 3 is a schematic diagram of converting a model predicted distance matrix into a master population sequence information;
FIG. 4 is a method framework of the method of the present invention.
Detailed Description
In this embodiment, a vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder is applied to a distribution environment composed of a warehouse point and N distribution task points, as shown in fig. 4, and is performed according to the following steps:
s1, building a vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help ；
S1.1, acquiring relevant data of a vehicle distribution task, wherein the relevant data comprises the following steps: distribution warehouse points and distribution task points, task demand and maximum vehicle capacity;
s1.2, respectively constructing a vehicle distribution task model f by using the formula (1) and the formula (2) origin Is a function of the objective function of:
in the formula (1), F 1 C is the length of the total travel path ij Representing the distance between the ith delivery task point and the jth delivery task point, i, j E [0, N]The method comprises the steps of carrying out a first treatment on the surface of the When i=0, c 0j Representing the distance from the warehouse point to the jth delivery task point, c when j=0 i0 Representing the distance, x, from the ith delivery task point to the warehouse point ij Indicating whether a path exists between the ith delivery task point and the jth delivery task point, if x ij If =1, it indicates that there is a path, if x ij =0, then no path exists;
in the formula (2), F 2 For the variance of the vehicle load, M is the total number of vehicles currently in use, d m Indicating the total amount of loading of the mth vehicle,representing the average load of the vehicle;
s1.3, constructing a vehicle distribution task model f by using the formulas (3) - (9) origin Is a constraint on (c):
equation (3) indicates that each dispensing task point can only be driven in by one vehicle;
equation (4) indicates that each dispensing task point can only be driven out by one vehicle;
equation (5) represents that the number of vehicles driving into the warehouse is at most K; k represents the maximum number of vehicles allowed to be used;
equation (6) represents that the number of vehicles that leave the warehouse is at most K;
equation (7) represents a loop path cancellation constraint equation; wherein u is i Representing the load of the vehicle after the ith delivery task point of the vehicle service; u (u) j Representing the load capacity of the vehicle after the j-th delivery task point is serviced; d represents a unified maximum load of the vehicle; q i Representing the demand of the ith delivery task point; q j Representing the demand of the j-th delivery task point;
equation (8) indicates that the total capacity of the vehicle after servicing the ith delivery mission point should be less than the maximum load D;
formula (9) represents a variable x ij Only the value 1 or 0 can be taken.
S1.4, constructing a vehicle distribution task auxiliary model f taking the formula (1) as an objective function and taking the formulas (3), (4) and (7) as constraint conditions help ；
S2, respectively according to the vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help Constructing a main population allocation scheme and an auxiliary population allocation scheme:
s2.1, enabling a chromosome coding sequence of each individual in the main population and the auxiliary population to show a vehicle access sequence of N distribution task points, wherein 0 represents a warehouse point, and the population scale is P;
s2.2, randomly initializing each individual in the main population and the auxiliary population, and correspondingly serving as an initial main population allocation scheme and an initial auxiliary population allocation scheme:
s2.2.1 randomly initializing each individual in the auxiliary population and the main population, scanning the chromosomes of each individual in the main population from beginning to end by using a split method, and traversing each distribution task in sequence, so that the corresponding distribution task is accessed by the same vehicle, and finally obtaining the shortest access path;
s2.2.2 calculating objective function values of all allocation schemes in the main population and the auxiliary population, wherein the objective function value of the ith main population individual isThe objective function value of the ith auxiliary population individual is F i ；
S2.2.3, the auxiliary population and the main population after the objective function is calculated are respectively used as an initial auxiliary population allocation scheme and an initial main population allocation scheme.
S3, setting the number of neurons in the denoising self-encoder model DAE as N; taking an initial auxiliary population allocation scheme as the input of the denoising self-encoder model DAE, taking an initial main population allocation scheme as the label of the denoising self-encoder model DAE, and training the denoising self-encoder model DAE to obtain an initial mapping relation model of the auxiliary population and the main population;
s4, defining the current iteration times as t, initializing t=1, and taking an initial auxiliary population distribution scheme and an initial main population distribution scheme as a t generation auxiliary vehicle distribution scheme and a t generation main vehicle distribution scheme respectively; taking the initial mapping relation model as a t generation mapping relation model;
s5, solving a t generation auxiliary vehicle distribution scheme by adopting a genetic algorithm based on the t generation main vehicle distribution scheme to obtain a t+1 generation auxiliary vehicle distribution scheme, solving the t generation main vehicle distribution scheme by adopting a model and a genetic algorithm to obtain a t+1 generation main vehicle distribution scheme, taking the t+1 generation auxiliary vehicle distribution scheme as the input of a denoising self-encoder model DAE, taking the t+1 generation main vehicle distribution scheme as the label of the denoising self-encoder model DAE, training the denoising self-encoder model DAE, and taking the trained mapping relation model as a t+1 generation mapping relation model:
the t+1st generation auxiliary vehicle distribution scheme in step S5 is obtained as follows:
s5.1a, selecting two individuals from the t-th generation auxiliary vehicle distribution scheme as t-th generation auxiliary parents using the tournament; then, the t-generation auxiliary parent is crossed by adopting OX to obtain a t-generation auxiliary child, and then the t-generation auxiliary child is improved by utilizing a local search operator, so that the improved t-generation auxiliary child is stored in a t+1th-generation temporary auxiliary vehicle distribution scheme;
s5.2a, slave-to-vehicle delivery task model f origin And ascending order each individual in the t-th generation main vehicle distribution scheme according to the selected objective function value, and storing the individual corresponding to the selected minimum objective function value as the t-th generation optimal main vehicle distribution schemeT+1st generation temporary auxiliary vehicle delivery protocol;
s5.3a, assisting the model f in terms of vehicle delivery tasks for each individual in the t+1st generation temporary assistance vehicle delivery regime help The objective function values of the vehicles are sorted in ascending order, and the first P delivery schemes with smaller objective function values are selected to form a t+1st generation auxiliary vehicle delivery scheme.
The t+1st generation host vehicle distribution scheme in step S5 is obtained as follows:
s5.1b, selecting two individuals from the t-th generation host vehicle distribution scheme as t-th generation parents using the tournament;
s5.2b, generating the t-th random number rand t If random number rand t Less than threshold delta 1 The father generation of the t generation is improved by adopting RBX crossing and mutation operators, and the child generation of the t generation is obtained and then stored in the temporary host vehicle distribution scheme of the t+1th generation; otherwise, directly storing the t generation father generation into the t+1 generation temporary host vehicle distribution scheme;
each individual in the S5.3b, t+1st generation auxiliary vehicle delivery scheme delivers the task auxiliary model f according to the vehicle help The objective function values of the (2) are sequenced in an ascending order, and the individual with the smallest objective function value is input into a t generation mapping relation model, so that t+1th generation offspring are obtained and stored into a t+1th generation temporary host vehicle distribution scheme:
specifically, a t+1st generation host vehicle distribution scheme is solved by adopting a t generation mapping relation model, as shown in fig. 3, chromosome paths in an auxiliary population t+1st generation optimal distribution scheme are converted into a distance matrix A, the distance matrix A is input into a denoising self-encoder DAE, k-means clustering, paired distance sorting and split method segmentation operation are carried out on the output of the denoising self-encoder DAE, and finally the model is put into a t+1st generation temporary host vehicle distribution scheme;
s5.4b, slave vehicle delivery task model f origin An objective function is selected, each individual in the t+1st temporary host vehicle distribution scheme is sequenced in an ascending order according to the selected objective function value, and a first sequenced individual is obtained;
distributing task model f from vehicle origin Is selected to be anotherThe objective functions are sequenced in an ascending order according to the selected objective function value for each individual in the t+1st temporary host vehicle distribution scheme, and a second sequenced individual is obtained;
and selecting individuals with the first P serial numbers from the first ordered individuals and the second ordered individuals, sequentially adding the individuals with the first P serial numbers into the t+1th generation host vehicle distribution scheme, and if the individuals with the first P serial numbers are less than P, sequentially selecting the first individuals in the first ordered individuals to respectively complement the rest individuals in the t+1th generation host vehicle distribution scheme.
The t+1st generation mapping relation model in the step S5 is obtained according to the following process:
s5.1c, taking a t+1th generation distribution scheme of the auxiliary population as the input of the denoising self-encoder DAE, taking a t+1th generation historical distribution scheme of the main population as the output of the denoising self-encoder DAE, and training the denoising self-encoder DAE as shown in FIG. 2, so as to construct a mapping relation model of the auxiliary population and the main population;
s5.1.1c, when the main population distribution scheme is used as a de-noising self-encoder DAE training label, the chromosome coding sequence information needs to be converted into distance matrix information B (N x N), as shown in FIG. 1, any element B of the distance matrix information ij (1.ltoreq.i, j.ltoreq.N) represents the distance between the ith delivery node and the jth delivery node, wherein α and β each represent a fixed smaller, larger value; the distances between the delivery nodes in the same path are integer multiples of the smaller value alpha, and the distances between the delivery nodes in different paths are larger value beta, so that the denoising self-encoder DAE can ensure that the delivery nodes are learned to be served by the same vehicle, wherein the distances between the delivery nodes in the same path are linearly increased along with the number of the accessed delivery nodes. For example, the distribution nodes 1,2,3 are a distribution node sequence currently visited by the same vehicle, the distances from the distribution node 1 to the distribution nodes 2,3 are α,2×α, respectively, and similarly, the distances from the distribution node 3 to the distribution nodes 1,2 are 2×α, α, respectively.
S5.1.2c, auxiliary population distribution scheme distance matrix a (N) was constructed using the method of step s4.1.1.
S5.2c, as shown in FIG. 2, the moments are respectivelyArrays a and B are used as denoising self-encoder DAE model inputs and labels, training is performed in the column input model (a= { a 1 ,a 2 ,a 3 ,...,a N },B＝{b 1 ,b 2 ,b 3 ,...,b N -x) wherein the loss function L sq As shown in formula (10):
sig(x)＝2/(1+exp((-2)*x))-1, (11)
f(x)＝x, (12)
wherein A represents a distance matrix corresponding to an auxiliary population of the input denoising self-encoder DAE model, B represents a distance matrix label of a main population of the denoising self-encoder DAE model, functions sig (x) and f (x) are respectively expressed as a formula (11) and a formula (12), x represents an independent variable, and a loss function L is minimized sq Until model training is completed.
S6, after the value of t+1 is assigned to T, judging whether the current iteration number T reaches the maximum iteration number T max If yes, output the T max Replacing a vehicle distribution scheme and taking the vehicle distribution scheme as an optimal vehicle distribution scheme; otherwise, step S5 is performed.
Claims (3)
1. A vehicle task distribution method based on a co-evolution algorithm of a denoising self-encoder is characterized by being applied to a distribution environment consisting of a warehouse point and N distribution task points, and comprises the following steps of:
s1, building a vehicle distribution task model f origin And a vehicle distribution task auxiliary model f help ；
S1.1, acquiring relevant data of a vehicle distribution task, wherein the relevant data comprises the following steps: distribution warehouse points and distribution task points, task demand and maximum vehicle capacity;
s1.2, respectively constructing a vehicle distribution task model f by using the formula (1) and the formula (2) origin Is a function of the objective function of:
in the formula (1), F 1 C is the length of the total travel path ij Representing the distance between the ith delivery task point and the jth delivery task point, i, j E [0, N]The method comprises the steps of carrying out a first treatment on the surface of the When i=0, c 0j Representing the distance from the warehouse point to the jth delivery task point, c when j=0 i0 Representing the distance, x, from the ith delivery task point to the warehouse point ij Indicating whether a path exists between the ith delivery task point and the jth delivery task point, if x ij If =1, it indicates that there is a path, if x ij =0, then no path exists;
in the formula (2), F 2 For the variance of the vehicle load, M is the total number of vehicles currently in use, d m Indicating the total amount of loading of the mth vehicle,representing the average load of the vehicle;
s1.3, constructing a vehicle distribution task model f by using the formulas (3) - (9) origin Is a constraint on (c):
equation (3) indicates that each dispensing task point can only be driven in by one vehicle;
equation (4) indicates that each dispensing task point can only be driven out by one vehicle;
equation (5) represents that the number of vehicles driving into the warehouse is at most K; k represents the maximum number of vehicles allowed to be used;
equation (6) represents that the number of vehicles that leave the warehouse is at most K;
equation (7) represents a loop path cancellation constraint equation; wherein u is i Representing the load of the vehicle after the ith delivery task point of the vehicle service; u (u) j Representing the load capacity of the vehicle after the j-th delivery task point is serviced; d represents a unified maximum load of the vehicle; q i Representing the demand of the ith delivery task point; q j Representing the demand of the j-th delivery task point;
equation (8) indicates that the total capacity of the vehicle after servicing the ith delivery mission point should be less than the maximum load D;
formula (9) represents a variable x ij Only the value 1 or 0 can be taken;
s1.4, constructing a vehicle distribution task auxiliary model f taking the formula (1) as an objective function and taking the formulas (3), (4) and (7) as constraint conditions help ；
S2, respectively according toVehicle distribution task model f origin And a vehicle distribution task auxiliary model f help Constructing a main population allocation scheme and an auxiliary population allocation scheme:
enabling the chromosome coding sequence of each individual in the main population and the auxiliary population to represent the vehicle access sequence of N distribution task points, wherein 0 represents a warehouse point and the population scale is P;
randomly initializing each individual in the main population and the auxiliary population, and correspondingly serving as an initial main population allocation scheme and an initial auxiliary population allocation scheme;
s3, setting the number of neurons in the denoising self-encoder model DAE as N; taking an initial auxiliary population allocation scheme as input of the denoising self-encoder model DAE, taking an initial main population allocation scheme as a label of the denoising self-encoder model DAE, and training the denoising self-encoder model DAE to obtain an initial mapping relation model of an auxiliary population and a main population;
s4, defining the current iteration times as t, initializing t=1, and taking an initial auxiliary population distribution scheme and an initial main population distribution scheme as a t generation auxiliary vehicle distribution scheme and a t generation main vehicle distribution scheme respectively; taking the initial mapping relation model as a t generation mapping relation model;
s5, solving a t generation auxiliary vehicle distribution scheme by adopting a genetic algorithm based on the t generation main vehicle distribution scheme to obtain a t+1 generation auxiliary vehicle distribution scheme, solving the t generation main vehicle distribution scheme by adopting a model and a genetic algorithm to obtain a t+1 generation main vehicle distribution scheme, taking the t+1 generation auxiliary vehicle distribution scheme as the input of the denoising self-encoder model DAE, taking the t+1 generation main vehicle distribution scheme as the label of the denoising self-encoder model DAE, training the denoising self-encoder model DAE, and taking the trained mapping relation model as a t+1 generation mapping relation model;
s6, after the value of t+1 is assigned to T, judging whether the current iteration number T reaches the maximum iteration number T max If yes, output the T max Replacing a vehicle distribution scheme and taking the vehicle distribution scheme as an optimal vehicle distribution scheme; otherwise, step S5 is performed.
2. The vehicle task delivery method according to claim 1, wherein the t+1st generation auxiliary vehicle delivery plan in step S5 is obtained as follows:
s5.1a, selecting two individuals from the t-th generation auxiliary vehicle distribution scheme as t-th generation auxiliary parents using the tournament; then, the t-generation auxiliary parent is crossed by adopting OX to obtain a t-generation auxiliary child, and then the t-generation auxiliary child is improved by utilizing a local search operator, so that the improved t-generation auxiliary child is stored in a t+1th-generation temporary auxiliary vehicle distribution scheme;
s5.2a, slave-to-vehicle delivery task model f origin Each individual in the t-th generation main vehicle distribution scheme is sequenced in an ascending order according to the selected objective function value, and the individual corresponding to the selected minimum objective function value is used as the t-th generation optimal main vehicle distribution scheme and stored in the t+1th generation temporary auxiliary vehicle distribution scheme;
s5.3a, assisting the model f in terms of vehicle delivery tasks for each individual in the t+1st generation temporary assistance vehicle delivery regime help The objective function values of the vehicles are sorted in ascending order, and the first P delivery schemes with smaller objective function values are selected to form a t+1st generation auxiliary vehicle delivery scheme.
3. The vehicle task delivery method according to claim 1, wherein the t+1st generation host vehicle delivery plan in step S5 is obtained as follows:
s5.1b, selecting two individuals from the t-th generation host vehicle distribution scheme as t-th generation parents using the tournament;
s5.2b, generating the t-th random number rand t If random number rand t Less than threshold delta 1 The father generation of the t generation is improved by adopting RBX crossing and mutation operators, and the child generation of the t generation is obtained and then stored in the temporary host vehicle distribution scheme of the t+1th generation; otherwise, directly storing the t generation father generation into the t+1 generation temporary host vehicle distribution scheme;
each individual in the s5.3b, t+1st generation auxiliary vehicle distribution scheme is on a vehicle-by-vehicle basisDistribution task auxiliary model f help The objective function values of the (1) are sequenced in an ascending order, and an individual with the smallest objective function value is input into a t generation mapping relation model, so that t+1 generation offspring are obtained and stored in a t+1 generation temporary main vehicle distribution scheme;
s5.4b, slave vehicle delivery task model f origin An objective function is selected, each individual in the t+1st temporary host vehicle distribution scheme is sequenced in an ascending order according to the selected objective function value, and a first sequenced individual is obtained;
distributing task model f from vehicle origin And carrying out ascending sort on each individual in the t+1st temporary host vehicle distribution scheme according to the selected objective function value to obtain a second sorted individual;
and selecting individuals with the first P serial numbers from the first ordered individuals and the second ordered individuals, sequentially adding the individuals with the first P serial numbers into a t+1th generation host vehicle distribution scheme, and if the individuals with the first P serial numbers are less than P, sequentially selecting the individuals with the first serial numbers from the first ordered individuals, and respectively supplementing the rest individuals in the t+1th generation host vehicle distribution scheme.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111451538.0A CN114118926B (en) | 2021-12-01 | 2021-12-01 | Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111451538.0A CN114118926B (en) | 2021-12-01 | 2021-12-01 | Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114118926A CN114118926A (en) | 2022-03-01 |
| CN114118926B true CN114118926B (en) | 2024-02-13 |
Family
ID=80369163
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111451538.0A Active CN114118926B (en) | 2021-12-01 | 2021-12-01 | Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114118926B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647821A (en) * | 2018-05-09 | 2018-10-12 | 浙江工业大学 | A kind of differential evolution logistics distribution method for optimizing route based on Parameter Self-learning |
| CN110689165A (en) * | 2019-08-29 | 2020-01-14 | 浙江工业大学 | Vehicle path optimization method based on road network reachability |
| CN112132317A (en) * | 2020-08-21 | 2020-12-25 | 浙江工业大学 | Vehicle path optimization method based on multi-population coevolution |
| CN112417604A (en) * | 2020-12-04 | 2021-02-26 | 安徽大学 | Electric vehicle distribution task self-adaptive distribution method based on multi-objective optimization |
-
2021
- 2021-12-01 CN CN202111451538.0A patent/CN114118926B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108647821A (en) * | 2018-05-09 | 2018-10-12 | 浙江工业大学 | A kind of differential evolution logistics distribution method for optimizing route based on Parameter Self-learning |
| CN110689165A (en) * | 2019-08-29 | 2020-01-14 | 浙江工业大学 | Vehicle path optimization method based on road network reachability |
| CN112132317A (en) * | 2020-08-21 | 2020-12-25 | 浙江工业大学 | Vehicle path optimization method based on multi-population coevolution |
| CN112417604A (en) * | 2020-12-04 | 2021-02-26 | 安徽大学 | Electric vehicle distribution task self-adaptive distribution method based on multi-objective optimization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114118926A (en) | 2022-03-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860754A (en) | AGV scheduling method based on ant colony and genetic algorithm | |
| CN111311158B (en) | Electric logistics vehicle path planning method under limited charging facility condition | |
| CN101620694A (en) | Logistics transportation decision-making support system and method, and transportation management system | |
| CN114897217A (en) | Fresh food electricity merchant front-end bin site selection-path planning method | |
| Li et al. | Deploying autonomous mobile lockers in a two-echelon parcel operation | |
| CN110276488A (en) | A kind of vehicle routing optimization method based on matrix in block form and fuzzy haulage time | |
| CN112417604B (en) | Multi-objective optimization-based self-adaptive distribution method for electric vehicle distribution tasks | |
| CN113361073A (en) | Inventory path joint optimization method based on improved Lagrange relaxation algorithm | |
| CN113627712A (en) | Method for optimizing operation sequence of shuttle vehicle of storage system | |
| CN114358233A (en) | Multi-AGV path planning problem optimization method and system based on double-hybrid particle swarm | |
| Hani et al. | Simulation based optimization of a train maintenance facility | |
| CN109359771A (en) | A kind of line haul vehicle and goods matching algorithm based on big data | |
| CN114037335B (en) | Vehicle and cargo matching intelligent decision method and system for maximizing carrying capacity | |
| CN114118926B (en) | Vehicle task distribution method based on co-evolution algorithm of denoising self-encoder | |
| Fallah et al. | A green competitive vehicle routing problem under uncertainty solved by an improved differential evolution algorithm | |
| Zhou et al. | Optimization for service routes of pallet service center based on the pallet pool mode | |
| CN115358455A (en) | Electric vehicle path optimization method considering battery replacement and hybrid time window constraints | |
| CN113988570A (en) | Multi-objective evolutionary algorithm-based tourism bus scheduling optimization method | |
| CN111445100B (en) | Vehicle and goods matching method based on self-adaptive time window under limited transport capacity | |
| CN114707707A (en) | Method and system for scheduling AGV task based on improved genetic algorithm | |
| CN105654185B (en) | Boat/vehicle splicing transportation optimization method based on improved differential evolution algorithm | |
| CN112836846A (en) | Multi-depot and multi-direction combined transportation scheduling double-layer optimization algorithm for cigarette delivery | |
| Huang et al. | Split-order consolidation optimization for online supermarkets: Process analysis and optimization models | |
| CN111652550A (en) | Method, system and equipment for intelligently searching optimal loop set | |
| Li et al. | Arctangent Adaptive Inertia Weight Particle Swarm Optimization for Logistics Distribution Route Planning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |