An Optimization Method for the Multi-objective Evaluation System of Small-piece Sorting in a Logistics Transfer Yard

Abstract

The present invention discloses an optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard, comprising the following steps: designing and defining multi-dimensional indicators of a sorting plan #imgabs0# for mapping and allocating cell resources and label resources; performing normalization processing on the multi-dimensional indicators and establishing a multi-objective sorting performance evaluation system; realizing standardized acquisition and real-time transmission of actual sorting data through an industrial Internet platform and a digital twin system; optimizing the mapping between parcel flow directions and cells, wherein the optimization part is divided into two stages: the first stage is global quantity distribution optimization; the second stage is quantity distribution optimization within a region. Based on a hybrid algorithm framework of a genetic algorithm and a simulated annealing algorithm, the present invention ensures global optimization and improves local search accuracy on the basis of establishing a multi-objective evaluation system, thereby effectively solving complex resource allocation problems in logistics sorting.

Description

Technical Field

[0001] The present invention relates to the field of logistics sorting, and particularly to an optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard. Background Art

[0002] With the rapid development of e-commerce and express logistics in recent years, small-piece sorting operations in logistics transfer yards face a series of problems such as high sorting volume, sorting errors, uneven utilization of sorting resources, and return flow processing. In the prior art, the evaluation indicators for sorting systems are often single, unable to comprehensively reflect the overall operation status of the sorting system from multiple perspectives such as sorting productivity, sorting accuracy rate, return flow rate, and sorting grid utilization rate, and there is a lack of effective connection between the evaluation results and resource allocation. At the same time, the development of digital twin technology has made it possible to monitor and predict the logistics sorting process in real time. How to construct a multi-objective evaluation system and use meta-heuristic optimization algorithms (such as genetic algorithms and simulated annealing algorithms) to dynamically and accurately allocate sorting resources reasonably has become a technical problem to be solved urgently. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard. Based on a hybrid algorithm framework of a genetic algorithm and a simulated annealing algorithm on the basis of establishing a multi-objective evaluation system, it ensures global optimization and improves local search accuracy, so as to effectively solve the complex resource allocation problem in logistics sorting.

[0004] The purpose of the present invention is achieved through the following technical solutions: An optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard includes the following steps:

[0005] Design and define multi-dimensional indicators for the sorting plan of the mapping allocation of grid resources and label resources ;

[0006] Perform normalization processing on the multi-dimensional indicators and establish a multi-objective sorting performance evaluation system;

[0007] Through the industrial Internet platform and the digital twin system, realize the standardized collection and real-time transmission of actual sorting data;

[0008] Optimize the mapping between the parcel flow direction and the grids. Among them, the optimization part is divided into two stages: the first stage is the global quantity distribution optimization; the second stage is the quantity distribution optimization within the region.

[0009] The beneficial effects of the present invention are as follows: Aiming at the problem that the traditional single sorting system evaluation index is difficult to comprehensively reflect the problems in the sorting process, the present invention constructs a multi-objective sorting performance evaluation system. Through strict mathematical definition, normalization processing, and weight assignment of various indicators, the overall evaluation function can comprehensively and objectively reflect the operation effect of the sorting system. And by using digital twin technology to construct a virtual model of the logistics transfer yard and sorting process, real-time acquisition of key indicator data is carried out through Internet of Things sensors and data acquisition devices to ensure the accuracy and real-time nature of the data on which the evaluation system is based. Based on the established multi-objective evaluation system, a hybrid algorithm framework based on genetic algorithm and simulated annealing algorithm is proposed to ensure global optimization and improve local search accuracy, thereby effectively solving the complex resource allocation problem in logistics sorting. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] The technical solution of the present invention will be further described in detail below with reference to the drawings, but the protection scope of the present invention is not limited to the following.

[0012] In recent years, with the rapid development of e-commerce and express logistics, the volume of parcels in the logistics network has continued to increase rapidly. As the hub and centralized processing site in the logistics network, the logistics transfer yard plays a crucial role. It is not only the core node for the centralized collection, distribution, and re-scheduling of parcels, but also undertakes the important tasks of improving the overall logistics efficiency, reducing transportation costs, and ensuring the delivery timeliness. In the transfer yard, a large number of parcels are converged, diverted, and preliminarily sorted according to the destination in the shortest time, providing strong support for subsequent rapid delivery. Logistics sorting, as an important link in the transfer yard operation, plays a key role connecting the front and the back in the entire logistics process. By adopting advanced cross-belt sorting machines and intelligent systems, the accurate division and efficient diversion of parcels can be realized, which can not only significantly improve the sorting speed and accuracy, but also effectively reduce the backflow and resource waste problems caused by mis-sorting or repeated sorting. Specifically:

[0013] As Figure 1 shown, a method for optimizing a multi-objective evaluation system for small-piece sorting in a logistics transfer yard includes the following steps:

[0014] First, for the core key indicators of logistics sorting: sorting capacity indicator, sorting accuracy indicator, sorting return rate indicator, and sorting grid utilization rate indicator, a sorting plan for mapping and allocating the current grid resources and label resources is designed and defined with multi-dimensional indicators:

[0015] 1. Sorting capacity: The sorting capacity indicators include the immediate sorting capacity, peak sorting capacity, and steady-state sorting capacity. Among them, the immediate sorting capacity is the number of packages sorted per minute under a sliding window, defined as:

[0016]

[0017] Among them, T is the minute length of the sliding window, is the number of packages sorted within the time minutes; represents the end time of the sorting shift.

[0018] The peak sorting capacity is defined as the highest capacity value that occurs during the entire sorting shift, defined as

[0019]

[0020] The steady-state capacity is defined as the average capacity during the steady-state period of the sorting system. Among them, the system steady-state period is defined as the time period when the immediate capacity is 80%-100% of the peak capacity. The specific definition is as follows:

[0021]

[0022] 2. Sorting accuracy rate: The sorting accuracy rate is defined as the ratio of the number of packages correctly sorted into the specified compartments during the sorting shift to the total number of sorted packages. Defined as

[0023] 100%

[0024] Among them, is the total number of sorted packages during the sorting shift, is the number of packages correctly sorted into the specified compartments during the sorting shift.

[0025] 3. Sorting return rate: The sorting return rate is defined as the packages whose scanning times on the conveyor belt exceed the upper limit during the sorting process. Returns usually result from uneven distribution of packages and compartment resources, so this indicator requires minimizing as much as possible. Defined as:

[0026] 100%

[0027] Among them, is the total number of returned packages during the shift.

[0028] 4. Sorting compartment utilization rate: Considering that the personnel layout in the site is evenly distributed among different sorting areas, the workload of personnel in each area needs to be as even as possible, that is, the number of times the compartments in each area are locked should be as equal as possible during the current shift. For the current logistics transfer yard, there areJ sorting areas, each area j has chutes. For area j the i th chute's locking times during the current shift is

[0029] = 1, 2, ..., .

[0030] On this basis, for area j the average locking times of all chutes is

[0031] .

[0032] Therefore, in the case of the most balanced utilization rate of chutes, the ideal locking times of all chutes in area j should all be . To describe the dispersion of actual sorting in area j , we use the standard deviation of the locking times in area j to model:

[0033]

[0034] where, when the locking times of all chutes are more balanced, is closer to 0. Therefore, we use the idea of coefficient of variation to construct the regional uniformity index , defined as:

[0035]

[0036] On this basis, considering the uniformity of the utilization rate of sorting chutes in each area of the entire logistics transfer yard, we use the average uniformity index within each area to describe the overall uniformity of the utilization rate of sorting chutes, defined as:

[0037]

[0038] Normalize the multi-dimensional index and establish a multi-objective sorting performance evaluation system;

[0039] From the four aspects of the efficiency, quality, stability, and resource utilization of sorting in the logistics transfer yard through these four indicators, comprehensively evaluate the sorting performance of the logistics sorting system. Based on the comprehensive definitions of the above four indicators, we normalize each of the above indicators to form an evaluation vector:

[0040]

[0041] Among them, for the sorting accuracy rate, sorting return rate, and grid utilization rate indicators, they are normalized indicators themselves. Therefore, only the production capacity indicator needs to be normalized:

[0042]

[0043] Among them, and are the theoretical peak production capacity and the upper bound of the steady-state production capacity obtained by setting the locking time to 0 by means of the established digital twin system through inputting the same real parcel information (such as the specific arrival of parcels, the distribution of parcel flow directions, etc.) under this sorting shift. On this basis, we set the weight vector as , then the multi-objective sorting performance evaluation system can be expressed as:

[0044]

[0045] Or it can be directly expressed using a function as:

[0046]

[0047] Among them, respectively represent when k = 1, 2, 3, 4

[0048] Through the industrial Internet platform and the digital twin system, realize the standardized collection and real-time transmission of actual sorting data;

[0049] Based on the established multi-objective sorting performance evaluation system, utilize the established industrial Internet platform and digital twin system to realize the standardized collection and real-time transmission of actual sorting data. Specifically, through the sensors in the cross-belt sorter, the handheld RFID devices in the logistics transfer yard, and the Internet of Things devices such as the cameras in the yard, real-time collection of sorting data is carried out, and the number of sorted parcels, the accurate number of sorted parcels, the return number, and the number of parcels falling into each grid in the sorting area are dynamically obtained, so as to provide real-time and accurate data support for the multi-objective evaluation system.

[0050] On this basis, we propose a hybrid algorithm framework based on the genetic algorithm and the simulated annealing algorithm to optimize the mapping between the parcel flow direction and the grids. Among them, the optimization part is divided into two stages: 1. The overall quantity optimization between the overall parcel flow direction and the sorting grids (global quantity distribution optimization) 2. Under the condition that the global quantity is determined, the quantity optimization between the parcel flow direction and the sorting grids in each sorting area (internal quantity distribution optimization in the area). The overall algorithm idea is as follows:

[0051] The overall quantity optimization between the overall parcel flow direction and the sorting grids

[0052] 1. Considering the stability of the parcel flow between sorting shifts, we take the overall parcel flow within the current nearest sorting shift and the mapping of the overall quantity between sorting bays as the initial population, and use the initial population as the candidate solution. x , the candidate solution x is a vector representing the number of bays assigned to each parcel flow, and satisfies

[0053]

[0054] where is the total number of sorting bays.

[0055] 2. For the candidate solution x , the sorting plan under this candidate solution (the mapping relationship between parcel flow and bays) is evaluated by a multi-objective evaluation system through the digital twin system, and then the overall comprehensive index

[0056] Then the fitness in the genetic algorithm is defined as F ( x ) = Q ( x ).

[0057] 3. Global search of genetic algorithm: The candidate solutions in the population are selected, crossed, and mutated according to the fitness F ( x ) to generate a new generation of population. Among them, to ensure the constraint (the total number of parcels remains unchanged), the crossover and mutation operations need to be designed or coordinated with repair operations.

[0058] 4. Local search of simulated annealing: Local search is performed within the neighborhood of some elite candidate solutions. The specific operations are as follows:

[0059] a. Select the parcel flows in the current candidate solution that contribute excessively and insufficiently to the objective function;

[0060] b. Generate a candidate solution ;

[0061] c. Calculate the difference If , then accept; if , then accept with a probability of exp( );

[0062] d. Lower the temperature and repeat the above process until the temperature is lower than the set lower limit.

[0063] 5. Repeat the genetic operation and local annealing optimization, continuously update the population until the convergence condition is met (such as the maximum number of generations or insufficient fitness improvement). At this time, the candidate solution is the global solution 。

[0064] Optimization of the quantity distribution inside the area

[0065] 1. Among the obtained global solutions , divide each sorting grid according to the sorting area it belongs to. For each sorting area j , the candidate solutions within it satisfy that the total number of occupied grids is fixed. Considering that the sorting production capacity index cannot be quantitatively estimated inside the area, the multi-objective evaluation system in each area is adjusted to

[0066]

[0067] 2. Inside each sorting area j , generate a local population (derive the initial perturbation solution from the global solution), use the genetic algorithm (selection, crossover, mutation) to generate multiple candidate allocation schemes within the area, and calculate the corresponding 。

[0068] 3. Similar to the global level, use the simulated annealing algorithm to perform local search and adjustment on the elite candidate solutions within the area

[0069] 4. Take the best solution finally obtained within each sorting area j , and update it to the global solution to obtain a global and detailed optimization scheme that specifically meets the multi-objective evaluation system 。

[0070] In summary, the present invention constructs an intelligent logistics optimization architecture based on a multi-objective sorting performance evaluation system and a digital twin platform, proposes a hybrid genetic and simulated annealing meta-heuristic algorithm framework, realizes the dynamic collaborative optimization among the key performance indicators in the sorting system, and provides a new theoretical basis and technical method for the comprehensive scheduling and intelligent decision-making of the logistics sorting system

Claims

1. An optimization method for the multi-objective evaluation system of small-piece sorting in a logistics transfer yard, characterized in that: Including the following steps: Sorting plan for mapping and allocating compartment resources and package label resources Design and define multi-dimensional indicators; The multi-dimensional indicators adopt the core key indicators for logistics sorting, including sorting production capacity indicators, sorting accuracy rate indicators, sorting return rate indicators, and sorting grid utilization rate indicators; The sorting plan for mapping and allocating grid resources and label resources The design and definition of multi-dimensional indicators include: (1) Sorting production capacity indicators, including instant sorting production capacity, peak sorting production capacity, and steady-state sorting production capacity; among them, the instant sorting production capacity is the number of packages sorted per minute under the sliding window, defined as: Among them, T is the minute length of the sliding window, is the number of packages sorted within the time minutes, represents the end time of the sorting shift; The peak sorting production capacity is defined as the highest production capacity value that appears during the entire sorting shift, defined as: The steady-state sorting capacity is defined as the average capacity during the steady state period of the sorting system , where the steady state period of the system is defined as the time period when the instant capacity is between 80% and 100% of the peak capacity; (2) The sorting accuracy rate indicator is defined as the ratio of the number of packages correctly sorted into the specified grid to the total number of sorted packages in this sorting shift: Among them, is the total number of sorted packages in this sorting shift, is the number of packages correctly sorted into the specified bays in this sorting shift; (3) The sorting return rate indicator is defined as the packages whose scanning times on the conveyor belt exceed the upper limit during the sorting process. The return is usually caused by uneven distribution of packages and grid resources, so this indicator requires as low as possible, defined as: Among them, is the total number of packages returned in this shift; (4)Sorting grid utilization rate: Assume that there are currently J sorting areas in the current logistics transfer yard, and each area j has grids; for the j th grid in area i , the number of times the grid is locked in the current shift is: For the area j within, the average number of times of locking the cells for all the cells is: To characterize the dispersion of the actual sorting within the region j , the standard deviation of the number of locked cells within the region j is used to model as follows: Among them, when the locking times of all the cell locks are more balanced, it is closer to 0; using the idea of coefficient variation, construct the regional uniformity index , which is defined as: Use the average uniformity indicator within each region to describe the overall uniformity of the sorting grid utilization rate, defined as: Perform normalization processing on the multi-dimensional indicators and establish a multi-objective sorting performance evaluation system; Through the industrial Internet platform and the digital twin system, realize the standardized collection and real-time transmission of actual sorting data; Optimize the mapping between the package flow direction and the grids. Among them, the optimization part is divided into two stages: the first stage is the global quantity distribution optimization; the second stage is the quantity distribution optimization within the region.

2. The optimization method of the multi-objective evaluation system for small-piece sorting in a logistics transfer yard according to claim 1, wherein: The multi-objective sorting performance evaluation system established by performing normalization processing on the multi-dimensional indicators includes: After normalizing the above various indicators, they form an evaluation vector: Among them, for the sorting accuracy rate, sorting return rate, and grid utilization rate indicators, they are already normalized indicators, and only the production capacity indicators need to be normalized: Among them, and are the theoretical peak production capacity and the upper bound of the steady-state production capacity obtained by setting the cell locking time to 0 by inputting the same package information under the actual sorting shift with the help of the established digital twin system; Set the weight vector as , then the multi-object sorting performance evaluation system is expressed as: Or directly expressed using a function as: Among them, respectively represent when k = 1, 2, 3, 4 .

3. An optimization method for a multi-objective evaluation system for small-item sorting in a logistics transfer yard according to claim 2, characterized in that: The realization of the standardized collection and real-time transmission of actual sorting data through the industrial Internet platform and the digital twin system includes: Real-time collect sorting data through sensors in the cross-belt sorter, handheld RFID devices in the logistics transfer yard, and cameras in the yard, and dynamically obtain the number of sorted packages per unit time, the accurate number of sorted packages, the return number, and the number of packages dropped into each grid in the sorting area.

4. An optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard according to claim 1, characterized in that: The global quantity distribution optimization, that is, the overall quantity optimization between the overall package flow direction and the sorting grids, includes: A1. Considering the stability of the parcel flow direction between sorting shifts, the overall parcel flow direction within the current nearest sorting shift and the mapping of the overall quantity between sorting compartments are used as the initial population, and the initial population is used as the candidate solution x , the candidate solution x is a vector representing the number of compartments assigned to each parcel flow direction and satisfies Among them, is the total number of sorting bays; among them, represents the number of bays allocated to the flow of the i-th parcel, represents the set of parcel flows; A2. For the candidate solution x , the sorting plan under this candidate solution is evaluated by the digital twin system using a multi-objective evaluation system, and then the overall comprehensive index is calculated Then the fitness in the genetic algorithm is defined as F ( x ) = Q ( x ); A3. Global search of genetic algorithm: Select, crossover, and mutate the candidate solutions in the population according to the fitness F ( x ) to generate a new generation of population; among them, to ensure the constraint, the total number of packages remains unchanged; A4. Simulated annealing local search: Perform local search within the neighborhood of some elite candidate solutions. The specific operation is: Select the package flow directions in the current candidate solutions that contribute too much and too little to the objective function; Among them, the improvement value of the overall comprehensive index brought by allocating one additional grid opening to the flow direction of each package in the candidate solution is defined as the contribution value. Let the candidate solution be , where represents the number of grid openings allocated to the flow direction of the i-th package, and G is the total number of package flow directions. First, the standard basis vector is defined as . Only the i-th component in this vector is 1. At this time, the contribution value of the candidate solution x on the i-th component is . Contributing too much to the objective function means that the contribution value of the package flow direction is higher than the preset contribution excess threshold; contributing too little to the objective function means that the contribution value of the package flow direction is lower than the preset contribution low threshold; b. Generate a candidate solution by diverting a parcel flow from an over-supplied bin to an under-supplied bin in the objective function. ; c. Calculate the difference , if , then accept; if , then accept with probability exp( ) d. Lower the temperature Repeat the above process until the temperature is lower than the set lower limit; A5. Repeat the genetic operation and local annealing optimization to continuously update the population until the convergence condition is met. At this time, the candidate solution is the global solution. .

5. An optimization method for a multi-objective evaluation system for small-piece sorting in a logistics transfer yard according to claim 4, characterized in that: The quantity distribution optimization within the region, that is, under the condition of determining the global quantity, the quantity optimization between the package flow direction and the sorting grids within each sorting region, includes: B1. Among the obtained global solutions , each sorting cell is divided according to the sorting area it belongs to; for each sorting area j , the candidate solutions therein satisfy that the total number of occupied cells is fixed; considering that the sorting production capacity index cannot be quantitatively estimated within the area, the multi-objective evaluation system in each area is adjusted to: B2. In each sorting area j generate a local population, that is, derive an initial perturbed solution from the global solution, use the genetic algorithm to generate multiple candidate allocation schemes within the area, and calculate the corresponding ; B3. Use the simulated annealing algorithm to perform local search and adjustment on the elite candidate solutions within the region; B4 Obtain the best solution finally obtained in each sorting area j and update it to the global solution so as to obtain a global and detailed optimization plan that specifically meets the multi-objective evaluation system. ​

Images