An elastic drop-off and pick-up transfer transportation system based on simulated annealing BATA algorithm

Abstract

The application discloses a kind of elastic sling based on simulated annealing BATA algorithm and hitch transport system, comprising: problem determining module, mixed integer programming model module, simulated annealing BATA algorithm module, verification module.Mixed integer programming model module includes nonlinear mixed integer programming model submodule and linear mixed integer programming model submodule;Simulated annealing BATA algorithm module includes initial solution construction submodule and operation operator submodule.The application greatly improves the utilization rate of tractor by considering the elastic working time of vehicle and solving the linear mixed integer programming model based on simulated annealing BATA algorithm, reduces the number of tractors and fixed costs, improves vehicle transportation efficiency, effectively reduces greenhouse gas emissions.The application can be widely used in long-distance freight transportation and other fields, opens a new chapter of low-carbon environmental protection and efficient transportation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of intelligent algorithm optimization and vehicle connection transportation, and particularly relates to an elastic drop-off and pick-up connection transportation system based on a simulated annealing BATA algorithm. BACKGROUND

[0002] Container connection transportation is an important transportation link before and after long-distance (such as railway and ocean) transportation, and drop-off and pick-up transportation is an important transportation mode in container connection transportation. A container connection transportation vehicle usually consists of a tractor and a trailer. The tractor has a power device and can move between different geographical locations by itself. The trailer has no power device, and thus its movement needs to be completed by being towed by the tractor. Compared with traditional container connection transportation, drop-off and pick-up transportation allows the tractor and the trailer to be separated from each other.

[0003] In many current studies on container connection transportation, the tractor usually leaves the yard to carry out related transportation activities after the transportation planning period starts. However, it takes a certain time for the customer to complete the loading and unloading of the container. Therefore, when the tractor arrives at the customer point, if the customer's loading and unloading of the container has not been completed, the tractor needs to wait. During the waiting period, the tractor is in an idle state, which will cause waste of tractor resources and thus reduce the utilization rate and transportation efficiency of the tractor. When the loading and unloading time of the container at the customer accounts for a large proportion in the entire transportation activity, the above negative problems will be more prominent. Therefore, although the existing traditional container connection transportation system has been widely used in long-distance transportation, it still has some deficiencies in the following aspects:

[0004] (1) The utilization rate of a single tractor is low, and the idle waiting time is long;

[0005] (2) The number of tractors used is too large, and the fixed cost is high;

[0006] (3) The system transportation efficiency is not high, and the greenhouse gas emission is large.

[0007] Therefore, the current common traditional container connection transportation system still has many problems in the above aspects, and therefore, it has become a hot problem in the connection transportation field to develop a drop-off and pick-up connection transportation system that can effectively improve the utilization rate of the tractor and the working efficiency of the transportation vehicle. SUMMARY

[0008] The main purpose of the present application is to improve the shortcomings of the existing traditional container intermodal transport system, and provide an elastic drop and hang intermodal transport system based on simulated annealing BATA algorithm. The system greatly improves the utilization rate of the tractor, reduces the number of tractors used and the fixed cost, improves the vehicle transportation efficiency, and effectively reduces the greenhouse gas emission by considering the elastic working time of the vehicle and solving the established linear mixed integer programming model based on the simulated annealing BATA algorithm.

[0009] The present application is implemented as follows: an elastic drop and hang intermodal transport system based on simulated annealing BATA algorithm comprises a problem determination module, a mixed integer programming model module, a simulated annealing BATA algorithm module and a verification module. The mixed integer programming model module comprises a nonlinear mixed integer programming model submodule and a linear mixed integer programming model submodule; the simulated annealing BATA algorithm module comprises an initial solution construction submodule and an operation operator submodule. The connection relationship of the modules and submodules is as follows: the problem determination module is connected with the mixed integer programming model module, the mixed integer programming model module is connected with the simulated annealing BATA algorithm module, and the simulated annealing BATA algorithm module is connected with the verification module; the nonlinear mixed integer programming model submodule in the mixed integer programming model module is connected with the linear mixed integer programming model submodule, and the initial solution construction submodule in the simulated annealing BATA algorithm module is connected with the operation operator submodule. The problem module is used to determine the problem to be solved, that is, how to complete all transportation tasks with the minimum total working time of the tractor within the planning period; the mixed integer programming model module is used to establish a solution model of the drop and hang intermodal transport problem, which comprises a nonlinear mixed integer programming model submodule and a linear mixed integer programming model submodule; the nonlinear mixed integer programming model submodule is used to establish a nonlinear solution model of the drop and hang intermodal transport problem; the linear mixed integer programming model submodule is mainly used for linearizing the above nonlinear mixed integer programming model; the simulated annealing BATA algorithm module mainly uses the simulated annealing BATA algorithm to solve the above mixed integer programming model, which comprises an initial solution construction submodule and an operation operator submodule; the initial solution construction submodule is used to construct an initial solution; the operation operator submodule comprises three different operation operators, i.e., a 2-OPT operator, a destroy-rebuild operator and a node insertion operator, which are used to solve the initial solution and find the optimal solution S* of the model; and the verification module is responsible for simulation experiment, which effectively verifies the effectiveness and superiority of the proposed mixed integer programming model and simulated annealing BATA algorithm, and effectively solves the problem of how to complete all transportation tasks with the minimum total working time of the tractor within the planning period.

[0010] The problem module is used to determine the problem to be solved, i.e., how to complete all transportation tasks with the minimum total tractor working time in a planning period. The problem is described in detail as follows: a region contains a yard and import first stage subtasks, import second stage subtasks, export first stage subtasks and export second stage subtasks, which are denoted as I1, I2, O1 and O2 respectively. The number of import first stage subtasks is equal to that of import second stage subtasks, and the number of export first stage subtasks is equal to that of export second stage subtasks. Any import / export first stage subtask corresponds to a unique export / import second stage subtask, and both correspond to the same geographical location. Let δ(j) denote the second stage subtask corresponding to the first stage subtask j, and any subtask j∈I1∪O1 must be completed before δ(j)∈I2∪O2. The problem is described as a graph G=(N, A), where N=I1∪I2∪O1∪O2∪{0} is the node set, and A={i, j)|i∈N, j∈N, i≠j}\{(δ(j), j)|j∈I1∪O1} is the arc set. There are sufficient tractors and trailers in the yard, and each trailer carries a container. A part of the trailers carry import goods for importers, which are referred to as import heavy trailers, and the other part of the trailers carry empty containers, which are referred to as empty trailers. The tractor needs to send an import heavy trailer to each import first stage subtask node, and the import heavy trailer becomes an empty trailer after the import goods are unloaded. The tractor needs to send the corresponding empty trailer from the corresponding import second stage subtask node to the yard or to the export first stage subtask node. The tractor needs to send an empty trailer to each export first stage subtask node for loading export goods. The trailer full of export goods is referred to as an export heavy trailer, and the tractor needs to drag the export heavy trailer from the export second stage subtask node to the yard. The loading and unloading time of the container required by the first stage subtask node j∈I1∪O1 is p j , and the first and corresponding second stage subtask nodes can be accessed by different tractors. A tractor can simultaneously drag K(≥2) trailers, which can be of the same type or different types. It is assumed that the speed of the tractor is constant when it is dragging a trailer or not dragging a trailer. The travel time of the tractor between any two nodes i and j is t(i, j)(≥0). The travel time between any three nodes i, j and k satisfies t(i, j)+t(j, k)>t(i, k).

[0011] The above problem aims to complete all transportation tasks with the minimum total tractor working time in the planning period [0, H]. The tractor can freely choose the time to leave the yard for work, but must return to the yard before the end of the planning period. The tractor working time refers to the time from when the tractor first leaves the yard to when it finally returns to the yard, which can reflect the size of the transportation cost to a certain extent.

[0012] The mixed integer programming model module comprises a nonlinear mixed integer programming model submodule and a linear mixed integer programming model submodule, and is used for establishing a solution model of the drop and pull connection transportation problem.

[0013] The nonlinear mixed integer programming model in the nonlinear mixed integer programming model submodule is a Mixed-Integer Non-Linear Programming Model Considering Flexible Work Time, abbreviated as MINLP-FT model.

[0014] The linear mixed integer programming model submodule is mainly used for linearizing the above nonlinear mixed integer programming model (MINLP-FT model). Since the objective function of the MINLP-FT model is a nonlinear function, the solving efficiency of the model will be reduced to some extent, so the linear mixed integer programming model submodule linearizes the objective function into a linear mixed integer programming model considering flexible work time (Mixed-Integer Linear Programming Model Considering Flexible Work Time, abbreviated as MILP-FT model).

[0015] The simulated annealing BATA algorithm module comprises an initial solution construction submodule and an operation operator submodule, and is mainly used for solving the MILP-FT model by using the simulated annealing BATA algorithm.

[0016] The simulated annealing BATA algorithm is a variant of the simulated annealing (Simulated Annealing, abbreviated as SA) algorithm: the simulated annealing algorithm with backtracking mechanism (Backtracking Mechanism, abbreviated as BATA). The backtracking mechanism is introduced into the SA algorithm, and the threshold-based SA algorithm is further improved, that is, the simulated annealing BATA algorithm is obtained. In the simulated annealing BATA algorithm, the threshold gradually decreases with the search process, and when the threshold is less than or equal to 0, a new threshold is generated, and the threshold backtracking is realized.

[0017] The initial solution construction submodule is used for constructing an initial solution, and is based on the Clarke-Wright (abbreviated as C-W) algorithm. The node sequence (r1, r2, …, r n )(r i ∈N\{0},i=1,2,…,n) is a path R of the drop and pull container connection transportation problem considering vehicle flexible working time. The path R represents that the tractor visits each node in turn from the yard, and finally returns to the yard.

[0018] The operation operator submodule includes three different operation operators, namely, a 2-OPT operator, a destroy-rebuild operator and a node insertion operator, which are used to find the optimal solution of the model.

[0019] The 2-OPT operator exchanges two nodes on the same path, and the operator tends to move the import subtask node and the export first-stage subtask node to the front end of the path. Among all the exchangeable nodes, the operator selects two nodes that can produce the minimum objective value for exchange. If the exchangeable nodes do not exist, the current solution remains unchanged.

[0020] The destroy-rebuild operator mainly operates on the current solution. For the current solution, the destroy-rebuild operator first removes a plurality of second-stage subtask nodes according to the similarity, that is, the solution is first destroyed. Then, the removed nodes are tried to be fused with the destroyed path again, that is, the solution is reconstructed. After the destruction-reconstruction, if a new neighborhood solution is produced, it is judged to be accepted or rejected. If the neighborhood solution cannot be successfully produced, the current solution remains unchanged.

[0021] The node insertion operator is responsible for taking the second-stage subtask node from the current solution and then reinserting it into the path. The operator selects the optimal node and inserts it into the corresponding optimal position to produce the solution with the minimum objective value. In addition, the removed node is not allowed to be inserted into the original path.

[0022] The verification module is responsible for performing simulation experiments to verify the effectiveness of the MILP-FT mathematical model and the simulated annealing BATA algorithm. All experiments are performed on a computer configured as Intel(R) Core(TM) i7-2600, 3.40GHz, 8GB RAM. The program uses C++ programming, and IBM ILOG CPLEX (12.6.1, 32-bit) is called by Visual Studio 2010 (32-bit) to solve the MILP-FT model. The simulation experiment results show that the simulated annealing BATA algorithm can obtain a better solution than the MILP-FT model in a shorter time.

[0023] Compared with the existing traditional container transfer transportation system, the elastic drop and hang transfer transportation system based on the simulated annealing BATA algorithm has the following advantages:

[0024] (1) effectively improve the utilization rate of the tractor

[0025] Through flexible drop and hang operation, the waiting time of the tractor can be effectively reduced. In this way, the work load of a single tractor is increased, more transportation operations can be carried out in a unit of time, and the utilization rate of the tractor is improved.

[0026] (2) Reduce the number of tractors used and reduce fixed costs

[0027] Taking into account the flexible working hours of the drop-and-hook transport, the utilization rate of tractors is improved, and the number of tractors required for transportation operations is reduced, thereby reducing the costs of purchasing tractors, hiring drivers, and daily vehicle maintenance and management.

[0028] (3) Improve transportation efficiency and reduce greenhouse gas emissions

[0029] Improve tractor transport efficiency by more than 30%, reduce transport costs by about 30%, and save fuel by 20%-30%. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Schematic diagram of the system structure of an embodiment of the present invention.

[0031] Figure 2 1 is a schematic diagram of the module flow of the simulated annealing BATA algorithm module 3 according to an embodiment of the present invention.

[0032] Figure 3 This is a schematic diagram of the sub-module flow of the initial decomposition and construction sub-module 3-1 according to an embodiment of the present invention.

[0033] Markings in the figure: 1. Problem determination module; 2. Mixed integer programming model module; 2-1. Nonlinear mixed integer programming model sub-module; 2-2. Linear mixed integer programming model sub-module; 3. Simulated annealing BATA algorithm module; 3-1. Initial solution construction sub-module; 3-2. Operation operator sub-module; 4. Verification module. DETAILED DESCRIPTION

[0034] Example: Figure 1 As shown, an embodiment of the present invention provides a flexible drop-and-hook shuttle transport system based on a simulated annealing BATA algorithm, comprising four main modules and four submodules, namely: a problem determination module 1, a mixed integer programming model module 2, a simulated annealing BATA algorithm module 3, and a verification module 4. The mixed integer programming model module 2 comprises a nonlinear mixed integer programming model submodule 2-1 and a linear mixed integer programming model submodule 2-2; the simulated annealing BATA algorithm module 3 comprises an initial solution construction submodule 3-1 and an operator submodule 3-2. The problem determination module 1 is connected to the mixed integer programming model module 2, the mixed integer programming model module 2 is connected to the simulated annealing BATA algorithm module 3, and the simulated annealing BATA algorithm module 3 is connected to the verification module 4; the nonlinear mixed integer programming model submodule 2-1 and the linear mixed integer programming model submodule 2-2 in the mixed integer programming model module 2 are connected, and the initial solution construction submodule 3-1 and the operator submodule 3-2 in the simulated annealing BATA algorithm module 3 are connected.

[0035] The problem module 1 is used to determine the problem to be solved, i.e. how to complete all the transportation tasks in the planning period with the minimum total tractor working time, which is described in detail as follows: a region contains a yard and import first stage subtasks, import second stage subtasks, export first stage subtasks and export second stage subtasks, which are denoted as I1, I2, O1 and O2 respectively. The number of import first stage subtasks is equal to that of import second stage subtasks, and the number of export first stage subtasks is equal to that of export second stage subtasks. Any import / export first stage subtask corresponds to a unique export / import second stage subtask, and both correspond to the same geographical location. Let δ(j) denote the second stage subtask corresponding to the first stage subtask j, and any subtask j∈I1∪O1 must be completed before δ(j)∈I2∪O2; the problem is described as a graph G=(N, A), where N=I1∪I2∪O1∪O2∪{0} is the node set, and A={(i, j) | i∈N, j∈N, i≠j}\{(δ(j), j) | j∈I1∪O1} is the arc set; the yard is parked with sufficient tractors and trailers, and each trailer is loaded with a container; a part of the trailers are loaded with import goods for importers, referred to as import heavy trailers, and the other part of the trailers are loaded with empty containers, referred to as empty trailers. The tractor needs to send an import heavy trailer for each import first stage subtask node, and the import heavy trailer becomes an empty trailer after the import goods in the container are unloaded. The tractor needs to send the corresponding empty trailer from the corresponding import second stage subtask node to the yard or to the export first stage subtask node. The tractor needs to send an empty trailer for each export first stage subtask node to load export goods. The trailer loaded with the export goods is referred to as an export heavy trailer, and the tractor needs to drag the export heavy trailer from the export second stage subtask node to the yard. The loading and unloading time of the container required by the first stage subtask node j∈I1∪O1 is p j , and the first and corresponding second stage subtask nodes can be accessed by different tractors. A tractor can simultaneously drag K(≥2) trailers, which can be of the same type or different types. It is assumed that the speed of the tractor is constant when it is dragging a trailer or not dragging a trailer. The travel time of the tractor between any two nodes i and j is t(i, j)(≥0). For any three nodes i, j and k, t(i, j)+t(j, k)>t(i, k).

[0036] The goal of the problem is to complete all the transport tasks with the minimum total tractor working time within the planning horizon [0, H]. The tractor can freely choose the time to leave the yard for work, but must return to the yard before the end of the planning horizon. The tractor working time refers to the time from the tractor initially leaving the yard to finally returning to the yard, which can to some extent reflect the size of the transport cost.

[0037] The mixed integer programming model module 2 includes a nonlinear mixed integer programming model submodule 2-1 and a linear mixed integer programming model submodule 2-2 for establishing a solution model of the drop-and-pull connection transportation problem.

[0038] The nonlinear mixed integer programming model (MINLP-FT model) in the nonlinear mixed integer programming model submodule 2-1 is described as the following formula:

[0039]

[0040]

[0041] The objective function (1) is responsible for minimizing the total tractor working time, where ∑ i∈N\{0} x i0 (s i +t(i, 0)) is the total time of the tractor finally returning to the yard, ∑ i∈N\{0} x 0i (s i-t(0, i)) is the total time when the tractor initially leaves the yard. Constraint (2) ensures that each node is visited only once. Constraint (3) ensures the balance of out-degree and in-degree. Constraints (4) and (5) limit the time when the tractor leaves the node. Among them, constraint (4) ensures the continuity of the tractor's departure time for two nodes visited successively by the same tractor, and M is a positive number greater than 0. Constraint (5) establishes a sequential service relationship between the first and corresponding second-stage subtasks. Constraints (6)-(13) establish a trailer quantity constraint. Among them, constraint (6) limits the total number of trailers that the tractor can tow. If the tractor is traveling on (i, j)∈A, constraint (6) ensures that the number of trailers towed by the tractor does not exceed the maximum number limit. Otherwise, the tractor does not tow any trailers. Constraints (7)-(13) describe the change in the number of trailers. Among them, constraint (7) states that when the tractor leaves node j∈I1, the number of imported heavy trailers towed is reduced by 1. Constraint (8) ensures that the number of imported trailers remains unchanged when the tractor leaves the non-import first-stage subtask node. Constraint (9) states that when the tractor leaves node j∈I2, the number of empty trailers it tows increases by 1. Constraint (10) states that when the tractor leaves node j∈O1, the number of empty trailers it tows decreases by 1. Constraint (11) ensures that the number of empty trailers it tows remains unchanged when the tractor leaves the corresponding node for all cases except the two cases described in constraints (9) and (10). Constraint (12) states that when the tractor leaves node j∈O2, the number of exported trailers increases by 1. Constraint (13) ensures that when the tractor leaves node j∈N\O2\{0}, the number of exported trailers remains unchanged. Constraints (14)-(18) define decision variables, where constraint (14) specifies the earliest and latest times that the tractor leaves the node, ensuring that all tasks are completed within the planning period.

[0042] The linear mixed integer programming model submodule 2-2 is mainly used to linearize the above nonlinear mixed integer programming model. Since the objective function (1) of the MINLP-FT model built by the nonlinear mixed integer programming model submodule 2-1 is a nonlinear function, which will reduce the solution efficiency of the model to a certain extent, the linear mixed integer programming model submodule 2-2 linearizes the objective function into formula (19):

[0043]

[0044]

[0045] in, and is the decision variable. i0 = 1, constraint (20) becomes Since the objective function (19) is to minimize the objective, Can only be equal to s i+ t(i, 0). When x i0 = 0, constraint (20) becomes According to the right half of constraint (14), we have -H + s i + t(i, 0) < 0. Thus, constraint (20) is relaxed, and constraint (23) and objective function (19) force When x 0i = 1, constraint (21) becomes and is relaxed, and constraint (22) becomes Combining the minimization objective of objective function (19), can only be equal to s i - t(0, i); when x 0i = 0, constraint (22) becomes According to the left half of constraint (14), we have s i - t(0, i) > 0. Thus, constraint (22) is relaxed, and constraint (21) becomes Combining constraint (24), we have

[0046] From the above, we can obtain a linear mixed integer programming model (MILP-FT model) considering flexible working hours, which is composed of objective function (19), constraints (2)-(18) and constraints (20)-(24).

[0047] The simulated annealing BATA algorithm module 3 includes an initial solution construction submodule 3-1 and an operation operator submodule 3-2, and mainly uses the simulated annealing BATA algorithm to solve the MILP-FT model.

[0048] The module flowchart of the simulated annealing BATA algorithm module 3 is as shown in Figure 2 First, an initial solution S (the initial solution is simultaneously regarded as the current solution S CUR and the current optimal solution S * ) is constructed, and the parameters are initialized: let I NO = 0, I ITER = 0, T MAX = f(S), T H = T MAX , ΔT H = wT MAX , I ITER = I ITER + 1. Then, the current solution is sent into an iterative search process. In each iteration, one of the three operation operators, 2-OPT operator, destroy-reconstruct operator and node insertion operator, is used on the current solution, and a new neighborhood solution S NIf the neighborhood solution is better than the current optimal solution, it will replace the current optimal solution and the current solution. On the contrary, if the neighborhood solution is worse than the current optimal solution, but the gap between it and the current solution is within a given threshold T H If the solution is within the range, the solution is accepted and replaced by the current solution. In each iteration, the threshold is increased with a constant step size ΔT H If the threshold drops to 0 or below, a new threshold greater than 0 is randomly generated. NO-MAX After iterations, if the current optimal solution has not been updated, the current optimal solution will be used to replace the current solution. ITER-MAX When , the simulated annealing BATA algorithm terminates.

[0049] The initial solution construction submodule 3-1 is used to construct the initial solution based on the Clarke-Wright (CW) algorithm. n )(r i ∈N\{0}, i=1,2,…,n) is a path R for the drop-and-hook container shuttle transport problem with flexible vehicle operating hours. Path R represents a tractor starting from the yard, visiting each node in sequence, and finally returning to the yard. The length of path R is defined as:

[0050]

[0051] The flowchart of the initial solution construction submodule 3-1 is as follows Figure 3 As shown: First, initialize |N|-1 paths (S), each path contains only one node i∈N\{0}. Calculate the path length savings for each pair of paths and sort them from large to small. Assume<R1,R2> The path pair with the greatest path length savings is R2. If appending R2 to R1 yields a better solution than the current one, the resulting solution replaces the current one. The path length savings are recalculated for the current solution. If appending R2 to R1 yields a worse solution than the current one, the append operation is attempted for the next pair of paths. This process repeats until no further path fusion is achieved, and the current optimal solution S* is output.

[0052] The operation operator submodule 3-2 includes three different operation operators: 2-OPT operator, destruction-reconstruction operator and node insertion operator, which are used to find the optimal solution of the model.

[0053] The 2-OPT operator exchanges two nodes on the same path, and the operator tends to move the import subtask node and the export first-stage subtask node to the front end of the path. Among all the exchangeable nodes, the operator selects two nodes that can produce the minimum objective value for exchange. If exchangeable nodes do not exist, the current solution remains unchanged. The specific operation process of the 2-OPT operator is as follows: first, given the current solution S CUR , let r i and r j be the first pair of exchangeable nodes in S CUR , let S CUR = S; second, exchange nodes r i and r j , and if the generated solution is better than S, replace S with the solution; finally, if r i and r j are not the last pair of exchangeable nodes in S CUR , let them be the next pair of exchangeable nodes and go to the second step; otherwise, let S CUR = S, and end the operation.

[0054] The destroy-reconstruct operator mainly operates on the current solution. For the current solution, the destroy-reconstruct operator first removes a number of second-stage subtask nodes according to the similarity, that is, the solution is first destroyed. Then, the operator attempts to fuse the removed nodes with the destroyed path again, that is, the solution is reconstructed. After the destruction and reconstruction, if a new neighborhood solution is generated, it is judged to be accepted or rejected. If a neighborhood solution cannot be successfully generated, the current solution remains unchanged.

[0055] The node insertion operator is responsible for taking the second-stage subtask nodes from the current solution and then inserting them into the path. The operator selects the optimal node and inserts it into the corresponding optimal position to generate a solution with the minimum objective value. In addition, the removed node is not allowed to be inserted into the original path.

[0056] The verification module 4 is responsible for performing simulation experiments to verify the effectiveness of the MILP-FT mathematical model and the simulated annealing BATA algorithm proposed above. All experiments are performed on a computer configured as Intel(R) Core(TM) i7-2600, 3.40GHz, 8GB RAM. The program uses C++ programming, and IBM ILOG CPLEX (12.6.1, 32-bit) is called to solve the MILP-FT model by Visual Studio 2010 (32-bit). The simulation results of the simulated annealing BATA algorithm and the C-W algorithm on the MILP-FT model are compared based on the examples R1-F to R28-F, C17-F to C29-F (where R represents a random example, 1 represents the example number, F represents flexible working hours, and C represents an example with clustering characteristics of customers). The simulation experiment results show that both the C-W algorithm and the simulated annealing BATA algorithm can obtain better solutions than the model MILP-FT in a very short time: for all C#-F type examples, the C-W algorithm produces a target value better than the model MILP-FT, with the maximum and minimum improvement of the model MILP-FT target value reaching 57.63% (C29-F) and 13.96% (C17-F), respectively, and the average improvement reaching 38.12%, and the average solving time is only 9.08 seconds, which is much smaller than the 7200 seconds solving time of the model MILP-FT; the maximum and minimum improvement of the simulated annealing BATA algorithm to the model MILP-FT target value are 59.73% (C29-F) and 17.21% (C17-F), respectively, the average target value improvement is 40.66%, the average solving time of the simulated annealing BATA algorithm is 1671.40 seconds, and the maximum solving time is 6235.20 seconds (example C29-F), which are all less than the 7200 seconds solving time of the model MILP-FT. Finally, the experimental results of the simulated annealing BATA algorithm and the C-W algorithm are compared, and it is found that the target value of the simulated annealing BATA algorithm is improved by an average of 4.25% compared with the target value of the C-W algorithm.

[0057] In summary, the flexible drop and pull connection transportation system based on the simulated annealing BATA algorithm can effectively improve the utilization rate of the tractor, reduce the number of tractors used and fixed costs, improve the efficiency of vehicle transportation, and reduce greenhouse gas emissions. It can be said that the development and use of drop and pull transportation is the transformation of traditional container connection transportation mode to modern container connection transportation mode, which has good social and economic benefits. At the same time, it also lays the foundation for establishing a resource-saving, low-carbon and environmentally friendly, and service comprehensive road transportation, so the flexible drop and pull connection transportation system can be widely used in long-distance cargo transportation and other fields, opening a new chapter of low-carbon and environmentally friendly and efficient transportation.

[0058] The above merely describes preferred embodiments of the present application. The above embodiments are merely used to illustrate the technical solutions of the present application, and not intended to limit the present application. Any variations or replacements of the present application, which are apparent to those skilled in the art and are within the technical scope of the present application, should be encompassed by the protection of the present application.

Claims

1. A flexible drop-and-hook transport system based on simulated annealing BATA algorithm, characterized by include: Problem determination module, mixed integer programming model module, simulated annealing BATA algorithm module, verification module; wherein the mixed integer programming model module includes: nonlinear mixed integer programming model submodule, linear mixed integer programming model submodule; simulated annealing BATA algorithm module includes: initial solution construction submodule, operation operator submodule; the connection relationship between modules and submodules is: the problem determination module is connected to the mixed integer programming model module, the mixed integer programming model module is connected to the simulated annealing BATA algorithm module, and the simulated annealing BATA algorithm module is connected to the verification module; the nonlinear mixed integer programming model submodule and the linear mixed integer programming model submodule in the mixed integer programming model module are connected, and the initial solution construction submodule and the operation operator submodule in the simulated annealing BATA algorithm module are connected; the problem module is used to determine how to solve the drop-and-hook shuttle transportation problem to be solved, which is how to complete all transportation tasks with the minimum total working time of the tractor within the planning period; the mixed integer programming model module is used to establish a solution model for the drop-and-hook shuttle transportation problem, including the nonlinear mixed integer programming model submodule and the linear mixed integer programming model submodule in the mixed integer programming model module. The nonlinear mixed integer programming model submodule and the linear mixed integer programming model submodule are used. The nonlinear mixed integer programming model of the nonlinear mixed integer programming model submodule is used to establish a nonlinear solution model for the drop-and-hook shuttle transportation problem. The linear mixed integer programming model submodule is mainly used to linearize the above nonlinear mixed integer programming model; the simulated annealing BATA algorithm module mainly uses the simulated annealing BATA algorithm to solve the above mixed integer programming model, including an initial solution construction submodule and an operation operator submodule. The initial solution construction submodule is used to construct an initial solution; the operation operator submodule includes three different operation operators, namely 2-OPT operator, destruction-reconstruction operator and node insertion operator, which are used to solve the above initial solution and find the optimal solution S* of the model; the verification module is responsible for conducting simulation experiments, which effectively verifies the effectiveness and superiority of the mixed integer programming model and simulated annealing BATA algorithm proposed above, and effectively solves the drop-and-hook shuttle transportation problem of how to complete all transportation tasks with the minimum total working time of the tractor within the planning period.

2. The flexible drop-and-hook transport system based on simulated annealing BATA algorithm according to claim 1 is characterized in that: The problem module is used to determine the drop-and-hook transport problem to be solved, that is, how to complete all transport tasks with the minimum total working time of the tractor within the planning period. The problem is described in detail as follows: a certain area contains a yard and import first-stage subtasks, import second-stage subtasks, export first-stage subtasks and export second-stage subtasks, and the subtask sets are I1, I2, O1 and O2 respectively; among them, the number of import first-stage subtasks and import second-stage subtasks is the same, and the number of export first-stage subtasks and export second-stage subtasks is the same; any import / export first-stage subtask corresponds to only one export / import second-stage subtask, and the two correspond to the same geographical location; δ(j) represents the second-stage subtask corresponding to the first-stage subtask j, and any subtask j∈I1∪O1 must be completed before δ(j)∈I2∪O2; the problem is described as a graph G=(N,A), where N=I1∪I2∪O1∪O2∪{0} is a node set, and A= {(i,j)|i∈N,j∈N,i≠j}\{(δ(j),j)|j∈I1∪O1} is an arc set; there are sufficient tractors and trailers parked on the yard, and each trailer is equipped with a container; the containers on some trailers are loaded with imported goods required by importers, referred to as import heavy trailers, and the containers on other trailers are loaded with empty containers, referred to as empty trailers; the tractor needs to send an import heavy trailer to each import first-stage subtask node. When the imported goods in the container are unloaded, the import heavy trailer immediately becomes an empty trailer; the tractor needs to send the corresponding empty trailer back to the yard from the corresponding import second-stage subtask node, or to the export first-stage subtask node; the tractor needs to send an empty trailer to each export first-stage subtask node for loading export goods; the trailer full of export goods is called an export heavy trailer, and the tractor needs to tow the export heavy trailer from the export second-stage subtask node back to the yard; the container loading and unloading time required for the first-stage subtask node j∈I1∪O1 is p j , the first and corresponding second-stage subtask nodes can be visited by different tractors; a tractor can tow K (≥2) trailers at the same time, and these trailers can be of the same type or different types; it is assumed that the speed of the tractor is constant when towing or not towing a trailer; the time it takes for the tractor to travel between any two nodes i and j is t(i, j) (≥0); between any three nodes i, j and k, t(i, j) + t(j, k) > t(i, k); The goal of the problem is to complete all transportation tasks within the planning period [0, H] with the minimum total tractor working time. The tractor can freely choose the time to leave the yard to work, but must return to the yard before the end of the planning period. The tractor working time refers to the time from the time the tractor initially leaves the yard to the time it finally returns to the yard. This time can reflect the size of the transportation cost to a certain extent.

3. The flexible drop-and-hook transport system based on simulated annealing BATA algorithm according to claim 1 is characterized in that: The mixed integer programming model module includes a nonlinear mixed integer programming model submodule and a linear mixed integer programming model submodule, which is used to establish a solution model for the drop-and-hook transport problem; The nonlinear mixed integer programming model in the nonlinear mixed integer programming model submodule is the Mixed-Integer Non-Linear Programming Model Considering Flexible Work Time, referred to as the MINLP-FT model; The linear mixed integer programming model submodule is mainly used to linearize the above-mentioned nonlinear mixed integer programming model (MINLP-FT model). Since the objective function of the MINLP-FT model is a nonlinear function, this will reduce the solution efficiency of the model to a certain extent. Therefore, the linear mixed integer programming model submodule linearizes the objective function into a linear mixed integer programming model considering flexible work time (i.e., Mixed-Integer Linear Programming Model Considering Flexible Work Time, referred to as MILP-FT model).

4. The flexible drop-and-hook transport system based on simulated annealing BATA algorithm according to claim 1 is characterized in that: The simulated annealing BATA algorithm module includes an initial solution construction submodule and an operation operator submodule, and mainly uses the simulated annealing BATA algorithm to solve the MILP-FT model; The simulated annealing BATA algorithm is a variant of the simulated annealing (SA) algorithm: a simulated annealing algorithm with a backtracking mechanism (BATA). The backtracking mechanism is introduced into the SA algorithm to further improve the threshold-based SA algorithm, thereby obtaining the simulated annealing BATA algorithm. In the simulated annealing BATA algorithm, the threshold value gradually decreases as the search process progresses. When the threshold value is less than or equal to 0, a new threshold value is generated to achieve threshold backtracking. The initial solution construction submodule is used to construct the initial solution based on the Clarke-Wright (CW) algorithm; the node sequence (r1, r2, ..., r n )(r i ∈N\{0}, i=1,2,…,n) is a path R for the drop-and-hook container shuttle transportation problem considering flexible vehicle working hours. Path R indicates that the tractor starts from the yard, visits each node in sequence, and finally returns to the yard. The operator submodule includes three different operators: 2-OPT operator, destruction-reconstruction operator and node insertion operator, which are used to find the optimal solution of the model; The 2-OPT operator swaps two nodes on the same path. This operator tends to move the import subtask node and the export first-stage subtask node toward the front of the path. Among all the swappable nodes, this operator selects the two nodes that can produce the minimum target value for swapping. If no exchangeable nodes exist, the current solution remains unchanged; The destruction-reconstruction operator mainly operates on the current solution. For the current solution, the destruction-reconstruction operator first removes several second-stage subtask nodes according to similarity, that is, it destroys the solution first; then it attempts to merge these removed nodes with the destroyed path, that is, reconstruct the solution; after the destruction-reconstruction process, if a new neighborhood solution is generated, it is judged and accepted or rejected; if a neighborhood solution is not successfully generated, the current solution remains unchanged; The node insertion operator is responsible for removing the subtask nodes of the second stage from the current solution and then reinserting them into the path; the operator selects the optimal node and inserts it into the corresponding optimal position to produce a solution with the minimum target value; in addition, the removed node is not allowed to be inserted into the path where it originally existed.

5. The flexible drop-and-hook transport system based on simulated annealing BATA algorithm according to claim 1 is characterized in that: The verification module is responsible for conducting simulation experiments to verify the effectiveness of the proposed MILP-FT mathematical model and simulated annealing BATA algorithm. All experiments were conducted on a computer configured with an Intel(R) Core(TM) i7-2600, 3.40GHz, and 8GB of RAM. The program was written in C++ and IBM ILOG CPLEX (12.6.1, 32-bit) was called using Visual Studio 2010 (32-bit) to solve the MILP-FT model. The simulation experimental results show that the simulated annealing BATA algorithm can obtain a better solution than the MILP-FT model in a shorter time.

Images