Passenger and freight co-transport bus route planning method based on adaptive large neighborhood search algorithm

Abstract

The invention relates to the technical field of information and communication suitable for logistics management, and aims to provide a passenger and freight co-transport bus route planning method based on an adaptive large neighborhood search algorithm. Comprising the steps of receiving passenger and freight transportation information through a reservation mode, and determining a passenger and freight co-transportation strategy based on different demand characteristics; a path optimization model is established, and on the premise of meeting all requirements, factors of vehicle capacity and time window limitation are emphasized, and a path scheme with the lowest cost is sought; solving the path optimization model based on a self-adaptive large neighborhood search algorithm, breaking a current solution by using a damage operation from an initial solution, and exploring a new solution space by using a repair operation; and finding a path scheme meeting a cost minimization target through continuous iteration. According to the invention, supply and demand balance in a passenger and freight co-transport mode can be realized, preferential meeting of passenger demands is taken as a core, a freight transport service is taken into consideration, and a new view angle and a solution are provided for efficient operation of a reservation type public transportation system.

Description

Technical Field

[0001] The present invention relates to the field of information and communication technology applicable to logistics management purposes, and in particular to a method for planning a scheduled passenger-freight shared public transportation route based on an adaptive large neighborhood search algorithm. Background Art

[0002] During peak hours of urban traffic, the traffic volume increases significantly due to the increase in travel demand, and freight vehicles, due to their slow speed and large size, often become bottlenecks on the road, further exacerbating congestion and leading to a sharp increase in carbon emissions. On the contrary, during non-peak hours, the capacity of urban public transportation systems is often in excess, which provides a way to solve the problem of imbalance between peak and non-peak passenger and freight transportation demand - that is, during non-peak hours, make full use of the surplus capacity of the public transportation system to coordinate and carry part of the freight demand, and realize the effective integration of passenger and freight transportation.

[0003] Therefore, how to effectively analyze and integrate passenger and freight demands, and rationally plan the operation route of scheduled buses according to actual conditions, so that they can meet customer needs at the lowest cost, is a key issue that needs to be solved urgently. In order to meet the above challenges, we give priority to meeting passenger travel needs, while using the remaining capacity to transport goods. By constructing a mixed integer programming model for the vehicle routing problem in the passenger and freight co-transport scenario, and establishing a suitable algorithm to solve it.

[0004] Chinese invention patent CN111798067A proposes a method for autonomous vehicle delivery path planning based on an adaptive large neighborhood search algorithm, including: calculating the distance between any two points; generating a vehicle path for autonomous vehicle delivery using a car carrier alone as an initial solution; using an adaptive large neighborhood search algorithm to select a neighborhood operator to operate on a feasible solution to form a new feasible solution; evaluating the quality of the new feasible solution and then evaluating the neighborhood operator; using a simulated annealing algorithm to control whether the feasible solution of the adaptive large neighborhood search algorithm is accepted, until the algorithm reaches the termination condition and outputs the path planning scheme for the collaborative delivery of the autonomous vehicle and the car carrier. This method performs delivery path planning for special cargo delivery such as autonomous vehicles that can move their own positions, extracts a mathematical model, and adds a split operator and a fusion operator to solve the problem on the basis of the Removal operator and the Insertion operator. However, this technical solution is only applicable to the path planning of autonomous vehicles with a single delivery purpose (especially long-distance transportation between long distances), and cannot be used to achieve passenger and cargo co-transportation between freight vehicles with different delivery purposes and urban public transportation under urban traffic conditions.

[0005] In order to effectively solve the vehicle path optimization model in the passenger and freight transport scenario, the choice of optimization algorithm is crucial. Accurate algorithms such as branch and bound method and cutting plane method can ensure the optimal solution. Through systematic search and pruning strategies, they can theoretically exhaust the solution space to achieve the goal of accurate solution. However, due to the high computational complexity, such algorithms often take a lot of time, especially when the problem scale is large, the solution efficiency is significantly affected. Therefore, the present invention aims to propose an algorithm mechanism that is more suitable for complex urban traffic conditions to improve computational efficiency, better cope with changes in passenger and freight demand in complex dynamic environments, and show superior flexibility and efficiency in practical applications. Summary of the invention

[0006] The technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a passenger-freight shared transport bus route planning method based on an adaptive large neighborhood search algorithm.

[0007] In order to solve the above technical problems, the solution adopted by the present invention is: A method for planning a passenger-freight shared bus route based on an adaptive large neighborhood search algorithm is provided, comprising the following steps: (1) Receive basic transportation information of passengers and cargo through reservation, which should at least include the number of passengers and cargo, starting and ending points, and time requirements; determine the passenger and cargo co-transportation strategy by obtaining the characteristics of different passenger and cargo demands; (2) Establishing a route optimization model based on the passenger-cargo co-transportation strategy, giving priority to satisfying passenger needs while taking into account cargo transportation services; the route optimization model consists of two parts: constraints and objective functions. The objective function includes the total driving distance and the average passenger travel time; on the premise of satisfying all needs, the model focuses on vehicle capacity and time window constraints to seek the route solution with the lowest cost; (3) The path optimization model is solved based on an adaptive large neighborhood search algorithm. Starting from the initial solution, the current solution is broken by a destruction operation, and then a new solution space is explored by a repair operation. Through continuous iteration, a path solution that meets the cost minimization goal is found, or a preset number of iterations is reached.

[0008] As a preferred solution of the present invention, the passenger and cargo co-transportation strategy includes the following two types: Strategy 1: Treat passenger and freight transport at the same location as one demand, and set each vehicle to visit each demand point only once; when passenger and freight transport exist at the same time, the passenger and freight transport demands at the location must be served by the same vehicle; Strategy 2: Separate the passenger and freight demands at the same location. Different vehicles can visit the location at most twice. When passengers are boarding and cargo is being loaded at the same time, different vehicles will provide services for each location.

[0009] As a preferred solution of the present invention, the constraint conditions include demand service constraint, service time window constraint and vehicle capacity constraint.

[0010] As a preferred solution of the present invention, the demand service constraints specifically include: ; Among them, the symbols in each formula refer to: k is the vehicle serial number; K is the vehicle set; i, j are nodes or locations that provide demand or service; A is the set of paths between different nodes, (i, j)∈ A; P is the set of pick-up / pick-up points; D is the set of drop-off / delivery points; n is the number of pick-up / pick-up points; It means that for any parameter value; x ijk is a 0-1 variable, which is 1 if vehicle k passes through path (i, j), otherwise it is 0; Formula (1) is used to ensure that all the selected passenger and cargo transportation needs can be met; Formulas (2)-(4) are flow balance constraints, which are used to ensure that the vehicle must start from the starting transportation site and return to that point, and meet the requirements of arriving at and leaving each customer point; Formula (5) is a connection constraint, which is used to ensure that the starting point and the end point of the same demand must be served by the same vehicle.

[0011] As a preferred solution of the present invention, the service time window constraint specifically includes: ; Among them, the symbols in each formula refer to: x ijk is the driving path of the vehicle between demand points i and j; s ik is the starting service time of vehicle k at demand point i; i is the service time; t ij is the travel time of the vehicle between demand points i and j; V is the set of all nodes or locations that provide demand or service; a i Set the earliest service start time for demand point i; b i Set the latest service start time for demand point i; Formula (6) ensures that vehicle k can reach the destination point i through path x after completing the service at the destination point i. ijk The time to arrive at the next service point j must not be greater than the service start time of that point; Formula (7) is the vehicle transportation priority constraint, that is, for each passenger and freight demand, the service order of its passenger pick-up demand must be earlier than the passenger drop-off / delivery demand; Formula (8) is the demand point service time window constraint, that is, the time s when vehicle k arrives at demand point i and starts service ik , no earlier than ai , and cannot be later than b i .

[0012] As a preferred solution of the present invention, the vehicle capacity constraint specifically includes: ; The symbols in the formulas refer to: Q 1 ik is the total passenger load of vehicle k when it leaves demand point i; Q 2 ik is the total cargo load of vehicle k when it leaves demand point i; , is the passenger demand at demand points i and j; , is the demand for goods at demand points i and j; q The rated capacity of the vehicle, i.e. the maximum number of passengers and standard cargo items that the vehicle can carry; Formula (9) and Formula (10) respectively ensure that vehicle k can reach the destination through path x after completing the service of demand point j. ijk When arriving at the next service point j, its passenger / cargo load when leaving point j needs to be equal to the sum of the passenger / cargo load after leaving point i and the passenger / cargo loading and unloading volume at point j; formula (11) ensures that the total load of vehicle k at point i shall not be greater than the rated capacity of the vehicle.

[0013] As a preferred solution of the present invention, the objective function is specifically as follows: ; Symbols in the formula: k is the vehicle serial number; K is the vehicle set; i, j are nodes or locations that provide demand or service; V is the set of all nodes or locations that provide demand or service; t ij is the travel time of the vehicle between demand points i and j; x ijk is the driving path of the vehicle between demand points i and j; s jk , s ik The time when vehicle k arrives at demand point j and starts serving i; is the passenger demand at demand point i; T represents the set of passenger boarding and alighting point pairs; T + Represents the boarding point; α is the weight coefficient of the total vehicle travel time; β is the weight coefficient of the average passenger travel time.

[0014] As a preferred solution of the present invention, the method of solving the adaptive large neighborhood search algorithm specifically includes the following steps: Step i: Input a feasible initial path plan and relevant parameters of the algorithm; Step II: define the destruction operator set and the repair operator set corresponding to the destruction operation and the repair operation; Step iii: define the neighborhood size set; Step IV: Initialize the temperature and operation weights; Step V: Construct the initial solution; Step ⅵ: Before the maximum number of iterations is reached, perform the following operations: (ⅵ-1) Check the solution pool. If it is empty, add a new solution and the corresponding cost; (ⅵ-2) Select new solutions from the solution pool to ensure that each iteration starts from a different initial solution; (ⅵ-3) roulette wheel selection operation; (ⅵ-4) Select the neighborhood size and perform damage repair operations; (ⅵ-5) Calculate the cost of the new solution and calculate the acceptance probability by simulated annealing method; (ⅵ-6) accepting the new solution as the current solution according to the acceptance criteria; (ⅵ-7) Update the optimal solution; (ⅵ-8) Update operation weights; (ⅵ-9) Cooling down, so that the algorithm transitions from the exploration stage to the convergence stage; Step ⅶ: When the maximum number of iterations is reached, output the current optimal solution as the final path plan.

[0015] As a preferred embodiment of the present invention, the adaptive large neighborhood search algorithm includes the following nine destruction operators: random request removal (RRE), random route removal (RRO), worst distance removal (WDR), worst time removal (WTR), worst neighborhood removal (WNR), Shaw removal (SR), proximity-based removal (PR), time-based removal (TR), and request-based removal (RR).

[0016] As a preferred solution of the present invention, the adaptive large neighborhood search algorithm includes the following three repair operators: random insertion (RAI), greedy insertion (GI), and regretful insertion (REI).

[0017] The present invention further provides a computer device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor executes any one of the aforementioned passenger-freight shared bus route planning methods based on an adaptive large neighborhood search algorithm.

[0018] The present invention also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute any of the aforementioned passenger-freight shared bus route planning methods based on an adaptive large neighborhood search algorithm.

[0019] Compared with the prior art, the technical effects of the present invention are: 1. Aiming at the problem of unbalanced utilization of transportation resources caused by the imbalance of passenger and freight flows in the existing urban transportation system, the present invention proposes a method for planning the route of a scheduled passenger-freight shared transport bus based on an adaptive large neighborhood search algorithm. According to the different transportation needs of passengers and freight (including passenger / freight volume, transportation start and end points, and transportation time windows, etc.), an efficient vehicle route planning scheme is provided for the passenger and freight transportation problem in the scheduled bus system. Based on the comprehensive consideration of the principle of passenger priority, the scheme can achieve the balance of supply and demand in the passenger-freight shared transport mode, and provides a new perspective and solution for the efficient operation of the scheduled bus system.

[0020] 2. Due to the particularity of public transportation services, the node visit count constraints commonly used in traditional vehicle path planning models fail to give priority to passenger factors and cannot be applied to public transportation passenger-cargo sharing modes. The present invention innovatively proposes two passenger-cargo sharing strategies, which can be used for public transportation path planning at stations of different sizes and types of passenger and cargo demand, including stations for picking up passengers and loading cargo, picking up passengers and unloading cargo, unloading passengers and loading cargo, and unloading passengers and unloading cargo. Therefore, the potential of the passenger-cargo sharing mode in public transportation in improving transportation efficiency is reflected, and the importance of giving priority to meeting passenger needs in this mode is also emphasized. When necessary, strategy 2 can often achieve better results than strategy 1.

[0021] 3. Based on the new passenger-cargo co-transportation strategy, the present invention proposes a vehicle path optimization model that comprehensively considers the passenger and cargo co-transportation needs. The model is centered on giving priority to satisfying passenger needs while taking into account cargo transportation services. In the objective function of the model, the key indicator of passenger transit time is specially introduced, which is specifically implemented by the calculation of formula (12) to ensure the optimization of passenger experience. In addition, the model also comprehensively considers key constraints such as time window and vehicle capacity, which are crucial to ensure the practicality and feasibility of the model solution.

[0022] 4. The present invention constructs an improved adaptive large neighborhood search algorithm, which effectively addresses the complexity of passenger-freight bus route planning by dynamically selecting and adjusting the destruction and repair operations. The results of the test experiment show that compared with the commercial solver Gurobi, the algorithm has obvious advantages in solution performance, mainly reflected in the solution time. Moreover, as the scale of constraints and decision variables in the model increases, this advantage will become more significant. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Schematic diagram of passenger and freight transport mode under different strategies of the present invention.

[0024] Figure 2 This is a schematic diagram of the execution process of the solution algorithm of the present invention.

[0025] Figure 3 This is a graph showing the weight changes of different operators in the algorithm of the present invention during the solution process. DETAILED DESCRIPTION

[0026] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0027] Firstly, the passenger-freight shared bus route planning method based on the adaptive large neighborhood search algorithm described in the present invention is implemented based on hardware devices of computer devices and computer-readable storage media.

[0028] The computer device includes: at least one processor, and a memory connected to the at least one processor in communication, wherein the memory stores instructions executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the passenger-freight bus path planning method based on the adaptive large neighborhood search algorithm. The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the passenger-freight bus path planning method based on the adaptive large neighborhood search algorithm.

[0029] The passenger-cargo shared transport bus route planning method based on the adaptive large neighborhood search algorithm described in the present invention mainly includes three key parts: determining the passenger-cargo shared transport strategy, establishing a route optimization model, and performing a search algorithm solution. Among them, the formulation of the passenger-cargo shared transport strategy is to divide different transportation strategies according to the characteristics of different passenger / cargo demands; the establishment of the route optimization model aims to seek the lowest cost route solution under the premise of meeting various conditions of the demand; the search algorithm solution is to cope with the computational complexity of large-scale mixed integer programming problems, and propose a computational method suitable for the model structure to achieve the accuracy and speed of model solution.

[0030] 1. Receive basic transportation information of passengers and goods by appointment through the reservation system of the public transportation service system, which should at least include the number of passengers and goods, starting and ending points, and time requirements; determine the passenger and freight co-transportation strategy by obtaining the characteristics of different passenger and freight demands.

[0031] There are two feasible solutions to the route planning problem of the passenger-freight shared bus described in the present invention. The first solution is to regard passenger and freight transport at the same location as one demand, and set each vehicle to visit each demand point only once. This shows that when passenger and freight transport exist at the same time, the passenger and freight transport demands at the location must be served by the same vehicle; the second solution is to handle the passenger and freight transport demands at the same location separately, and different vehicles can visit the location at most twice, that is, when there are passengers boarding and cargo being loaded at the same time, different vehicles can provide services respectively. These two situations correspond to Figure 1 Strategy 1 and Strategy 2 in the above. Usually, the traditional vehicle routing problem is restricted to visit the same node only once to ensure the rational use of resources and minimize the cost. However, in the passenger and freight transport scenario, this definition may lead to an increase in transportation costs. Figure 1 As can be seen from the figure, the difference between strategy 1 and strategy 2 is that it allows different vehicles to visit the same node twice.

[0032] The planning method of the present invention can be used for different optimization goals, such as minimum time cost, minimum expense cost and maximum profit, etc. The path planning method provides suitable transportation solutions for users with different needs, meets the passenger / cargo needs of customers, realizes the reasonable allocation of transportation resources, realizes the cost reduction and efficiency improvement of the transportation system, and proposes a paradigm for the passenger and cargo co-transportation mode of scheduled public transportation.

[0033] In Tables 1 and 2, the objective function values ​​of the two strategies are compared respectively. Table 1 shows the compact time window constraint conditions based on ten demand point examples, and Table 2 shows the loose time window constraint conditions based on ten demand point examples.

[0034] Table 1 ; Table 2 ; As can be seen from Tables 1 and 2, Strategy 1 performs well under tight time windows: When the time window is tight, Strategy 1 performs better in reducing vehicle travel time and total cost. This is mainly due to the strict time window constraints, which leads to higher costs due to the priority given to passenger demand, especially when additional vehicles need to be dispatched.

[0035] In addition, Strategy 2 has more advantages under loose time windows: under loose time window conditions, Strategy 2 can better balance vehicle and passenger time and reduce the waiting time of passengers in the car. This shows that under the condition of greater time flexibility, Strategy 2 can more effectively dispatch resources and optimize overall operations.

[0036] 2. Establish a path optimization model based on the passenger and freight transport strategy; on the premise of meeting all needs, focus on vehicle capacity and time window constraints to seek the lowest cost path solution.

[0037] The route optimization model of the present invention includes two parts: constraint conditions and objective function, wherein the constraint conditions mainly include demand service constraint, service time window constraint and vehicle capacity constraint. The objective function mainly takes time cost as the optimization target.

[0038] (1) Demand service constraints: Since the transportation task is to meet the needs of the reservation request after being processed by the platform, it is necessary to ensure that the reservation-based bus can serve all the screened needs, and the setting of formula (1) meets this constraint. Formulas (2)-(4) are classic flow balance constraints, which ensure that the vehicle must start from the starting transportation site and return to that point, and meet the requirements of arriving at and leaving each customer point. In addition, formula (5) is a connection constraint, which ensures that the starting point and end point of the same demand must be served by the same vehicle.

[0039] ;

[0040] (2) Service time window constraints: In order to ensure that the vehicle meets the demand within the expected time period, it is necessary to add appropriate time constraints to the scheduled bus. Among them, formula (6) expresses that after vehicle k finishes serving demand point i, it takes path x ijk The time to arrive at the next service point j must not be greater than the service start time of that point. Formula (7) is the vehicle transportation priority constraint, that is, for each passenger and freight demand, the service order of its passenger pick-up demand must be earlier than the passenger drop-off (delivery) demand. Formula (8) is the demand point service time window constraint, that is, the time s when vehicle k arrives at demand point i to start service ik , no earlier than a i , and cannot be later than b i .

[0041] ; (3) Vehicle capacity constraints: As a means of travel by appointment, scheduled public transportation needs to ensure that there is enough personal space in its carriage (i.e. it cannot exceed the rated capacity of the vehicle). q, that is, the maximum number of passengers and standard cargo items that can be loaded cannot exceed the vehicle's limited capacity). Therefore, the number of passengers and cargo in the vehicle needs to be limited in the model, and equations (9) and (10) respectively represent the number of vehicles k passing through path x after completing the service of demand point i. ijk When arriving at the next service point j, the passenger / cargo load when leaving point j needs to be equal to the sum of the passenger / cargo load after leaving point i and the passenger loading (loading and unloading) at point j. Formula (11) shows that the total load of vehicle k at point i shall not be greater than the rated capacity of the vehicle.

[0042] ; (4) Objective function: Considering the reality that passenger and cargo transportation needs coexist, the priority of passenger transportation in the public transportation system is higher than that of cargo transportation. Therefore, in addition to the total travel distance, the objective function of the present invention also adds the average passenger travel time to improve the priority of passenger travel in the system. Therefore, the objective function is as follows: ; It should be noted that, since vehicle operation needs to meet time window constraints, the average passenger travel time item may further include the time passengers spend waiting for departure at the station.

[0043] The meanings of the parameters and variables in formulas (1) to (12) of the present invention are specifically shown in Table 3.

[0044] Table 3:

[0045] ;

[0046] It means that for any parameter value (such as k∈K means that for any k belongs to K).

[0047] 3. Solve the path optimization model based on the adaptive large neighborhood search algorithm. Starting from the initial solution, use the destruction operation to break the current solution, and then use the repair operation to explore the new solution space; through continuous iteration, find the path plan that meets the cost minimization goal, or reach the preset number of iterations.

[0048] The adaptive large neighborhood algorithm of the present invention belongs to a class of neighborhood search algorithms and can be used to solve the pick-up and delivery vehicle routing problem with a time window. In recent years, the adaptive large neighborhood algorithm has achieved remarkable results in dealing with large-scale vehicle routing problems, and many public documents have been documented. Its core mechanism starts with a set of initial solutions and iteratively optimizes them through a series of destruction and repair operations. Among them, the destruction operation breaks the current state by removing part of the current solution, while the repair operation attempts to rebuild the solution on the basis of the destruction in order to improve the quality of the solution. Through this iterative process, the adaptive large neighborhood algorithm can gradually approach the optimal or approximately optimal solution to the problem.

[0049] The solution process of the adaptive large neighborhood algorithm of the present invention is specifically as follows: Step i: Input a feasible initial path plan x0 and relevant parameters of the algorithm;

[0050] These include: maximum number of iterations, maximum neighborhood range percentage maxPercentageNHB, minimum neighborhood size minSizeNHB, etc.

[0051] Step II: define a set of destruction operators and a set of repair operators corresponding to the destruction operations and repair operators;

[0052] Among them: the destruction operator set D is mainly responsible for removing some path elements from the current solution, such as vehicle tasks and stations. The repair operator set R is mainly responsible for re-adding the path elements removed by the destruction operator to the solution to obtain a new feasible solution.

[0053] Step III: Define the neighborhood size set.

[0054] Set the neighborhood size set Q =﹛ q min ,q1,q2,q3,…,q max ﹜, used to control the number of tasks processed in each step. max = max﹛1,|maxPercentageNHB The number of requests |﹜ is the maximum size of the neighborhood; and q min The minimum size of the neighborhood is determined by the parameter minSizeNHB. The neighborhood size is used to control the scope of the destruction and repair operations, thereby achieving different degrees of search depth.

[0055] Step IV: Initialize the temperature and operation weights.

[0056] Set the initial temperature T and cooling rate φ to control the acceptance criterion of simulated annealing. Initialize the weight w of each destruction and repair operation.

[0057] Step V: Construct the initial solution.

[0058] Generate an initial solution x0 and set it as the current optimal solution The initial solution can be generated by random assignment or heuristic method to ensure its feasibility.

[0059] Step ⅵ: Iteration process.

[0060] Before the maximum number of iterations is reached, perform the following operations: (ⅵ-1) Check whether the solution pool is empty: If the solution pool is empty, the regret value insertion method is called to generate a new solution x and the corresponding cost Cost by maximizing the regret value of the insertion point x , and add it to the solution pool.

[0061] (ⅵ-2) Select a new solution from the solution pool: If the solution pool is not empty, select a solution x from it and remove it from the solution pool, so that each iteration starts with a different initial solution.

[0062] (ⅵ-3) Roulette selection operation: Perform a roulette operation and randomly select a destruction operator d and a repair operator r according to the weight of the operation. This will make the operation with good performance have a higher probability of being selected.

[0063] (ⅵ-4) Select the neighborhood size and perform damage repair operations: Randomly select a neighborhood size q, perform a destruction repair operation, and generate a new solution x new The destruction operation first removes part of the path from the current solution, and then the repair operation reinserts the removed part into the solution to ensure the feasibility of the solution.

[0064] (ⅵ-5) Calculate the cost of the new solution: New calculation solution x new Cost s new ; At the same time, the new solution is calculated by simulated annealing method x new The acceptance probability is: ; in, P accept represents the probability of acceptance; x new represents the new solution; Δ h( x new ,x) represents the cost difference between the new solution and the current solution; T refers to the current temperature.

[0065] (ⅵ-6) Acceptance criteria: Generate a random number in the range of 0 to 1. If the random number is less than the acceptance probability P accept , then accept the new solution x new as the current solution x.

[0066] (ⅵ-7) Update the optimal solution (ⅵ-8) Update operation weight: The weights w of the destruction and repair operations are updated based on the performance of the operation (i.e., whether the generated solution is accepted or improves the current optimal solution). Generally speaking, operations with good performance will increase their weights, while operations with poor performance will decrease their weights.

[0067] (ⅵ-9) Cooling: After each round of iteration, the formula T = φ T Lowering the temperature T gradually reduces the acceptance probability of the solution, making the algorithm transition from the exploration stage to the convergence stage.

[0068] Step 7: Output the optimal solution ; When the maximum number of iterations is reached, the algorithm outputs the current optimal solution , as the final path solution.

[0069] Compared with conventional algorithms, the adaptive large neighborhood algorithm used in the present invention has made significant improvements in the operation steps, including the introduction of an adaptive destruction and repair weight update mechanism, a flexible neighborhood size control method, a solution generation strategy based on regret value, the fusion of simulated annealing and ALNS, and a diversified initial solution selection strategy based on a solution pool. These improvements improve the intelligence and efficiency of the algorithm, significantly enhance the exploration capability of the solution space and the local optimization effect, thereby achieving higher solution efficiency, better solution quality, and stronger adaptability and robustness.

[0070] In the adaptive large neighborhood algorithm used in the present invention, adaptive changes are mainly made to the structure of the destruction operator, including the introduction of a selection strategy based on task characteristics, dynamic adjustment of the destruction range, destruction operators that integrate historical information, diversification of destruction operator types, adaptive weight adjustment mechanism, and dynamic decision-making combined with the solution state after destruction. These improvements significantly improve the pertinence and flexibility of the destruction operator, allowing the algorithm to more efficiently explore the solution space, focus on key areas for optimization, and show significant advantages in solution efficiency, adaptability, and robustness.

[0071] In view of the characteristics of the optimization model and the differences in passenger and freight demand, the destruction operators designed in the algorithm of the present invention include the following nine types: (1) Random Request Removal (RRE): This operation randomly removes requests from the current solution. It is a basic operator and is often used in neighborhood search algorithms.

[0072] (2) Random Route Removal (RRO): This destruction operation is used to randomly select routes containing requests, by randomly filtering invalid routes to ensure that the selected routes contain at least enough requests for operation.

[0073] (3) Worst Distance Removal (WDR): The core idea of ​​this operation is to remove the distance factors that contribute the worst to the overall solution quality from the current solution. These factors usually significantly increase the cost of the solution or affect the efficiency of the solution.

[0074] (4) Worst Time Removal (WTR): This is a destruction operation that considers the worst time and is used to identify time window constraint problems. It can help optimize the time constraints in route scheduling problems by finding the request with the largest difference between the service start time and the time window start time.

[0075] (5) Worst Neighbor Removal (WNR): This operation effectively identifies the requests that have the greatest impact on the path cost based on the relative contributions of neighboring requests. By calculating the distance contribution of each request relative to its previous and next locations, the request with the highest percentage contribution to the overall path distance is identified. The entire path can then be optimized by removing or re-optimizing this request.

[0076] (6) Shaw Removal (SR): This operation selects the next service request that is most relevant to the current location by taking into account the factors in the model (in this invention, distance, time window, and demand are considered).

[0077] (7) Proximity-based Removal (PR): This method identifies the next request based on proximity and selects the closest location to disrupt the current solution to achieve further optimization.

[0078] (8) Time-based Removal (TR): This is to select the next most suitable request based on the time window difference. Its goal is to minimize the time window difference between requests to satisfy the time window constraint and optimize the service order in the scheduling problem.

[0079] (9) Request-based Removal (RR): This is a heuristic destruction operation based on demand differences, which can optimize load balancing in routing or scheduling problems by selecting the next location with the closest demand.

[0080] The repair operators used in the adaptive large neighborhood search algorithm of the present invention include the following three types: (1) Random Insertion (RAI): This operation generates new candidate solutions by randomly reintroducing elements removed in the destruction phase into the solution. This repair operation does not prioritize the best placement, but emphasizes randomness to enhance the diversity of the search space and promote a wider exploration of potential solutions.

[0081] (2) Greedy Insertion (GI): The goal of this operation is to improve the quality of the solution with each insertion, or at least prevent it from deteriorating. However, while this approach may be effective in the short term, it may also accelerate the algorithm's convergence to a local optimal solution because it prioritizes immediate gains over long-term optimization potential.

[0082] (3) Regret Insertion (REI): The advantage of the regret insertion based approach is that it takes into account potential future losses, leading to more cautious decision making. This approach helps prevent the placement of critical nodes in suboptimal locations, which may otherwise result in significant cost increases.

[0083] Figure 3 The weight change diagram of different operators in the algorithm of the present invention during the solution process, where (a) shows the weight change in the destruction operation, and (b) shows the weight change in the repair operation.

[0084] In order to demonstrate that the solution based on the adaptive large neighborhood search algorithm for the path optimization model in the present invention has technical advantages, a comparative calculation is set up for verification.

[0085] The verification conditions are set as follows: In the comparative experiment, the Gurobi solver and the adaptive large neighborhood search algorithm (ALNS) are used to solve the path optimization model. For Gurobi, Python is used to call its 11.0 version, and the solution time limit is set to 5000 seconds. For ALNS, the specific parameters are set as follows: the minimum neighborhood size minSizeNHB is 1, the maximum neighborhood range percentage maxPercentageNHB is 5%, the temperature drop rate φ of simulated annealing is 0.9994 and the initial acceptance probability parameter is 0.03, the decay parameter of the dynamic adjustment operator weight is 0.30, and the number of iterations is 2000. Under the same example conditions, the performance difference of the two methods in solving the path optimization problem is evaluated through comparative experiments. The results of the comparative experiment are recorded in Table 4.

[0086] Table 4

[0087] ;

[0088] From the content of Table 4, it can be seen that compared with the commercial solver Gurobi, the algorithm based on adaptive large neighborhood search in the present invention has obvious advantages in solution performance, mainly reflected in the solution time. When the number of demand points doubles, the relatively saved computing time is more obvious, and the solution efficiency is significantly improved.

Claims

1. A passenger-freight bus route planning method based on an adaptive large neighborhood search algorithm, characterized in that: The following steps are involved: (1) Receive basic transportation information of passengers and cargo through reservation, which should at least include the number of passengers and cargo, starting and ending points, and time requirements; determine the passenger and cargo co-transportation strategy by obtaining the characteristics of different passenger and cargo demands; (2) Establishing a route optimization model based on the passenger-cargo co-transportation strategy, giving priority to satisfying passenger needs while taking into account cargo transportation services; the route optimization model consists of two parts: constraints and objective functions. The objective function includes the total driving distance and the average passenger travel time; on the premise of satisfying all needs, the model focuses on vehicle capacity and time window constraints to seek the route solution with the lowest cost; (3) The path optimization model is solved based on an adaptive large neighborhood search algorithm. Starting from the initial solution, the current solution is broken by a destruction operation, and then a new solution space is explored by a repair operation. Through continuous iteration, a path solution that meets the cost minimization goal is found, or a preset number of iterations is reached.

2. The method according to claim 1, characterized in that The passenger and cargo co-transportation strategies include the following two: Strategy 1: Treat passenger and freight transport at the same location as one demand, and set each vehicle to visit each demand point only once; when passenger and freight transport exist at the same time, the passenger and freight transport demands at the location must be served by the same vehicle; Strategy 2: Separate the passenger and freight demands at the same location. Different vehicles can visit the location at most twice. When passengers are boarding and cargo is being loaded at the same time, different vehicles will provide services for each location.

3. The method according to claim 1, characterized in that The constraints include demand service constraints, service time window constraints and vehicle capacity constraints.

4. The method according to claim 3, characterized in that The demand service constraints specifically include: ； ； Among them, the symbols in each formula refer to: k is the vehicle serial number; K is the vehicle set; i, j are nodes or locations that provide demand or service; A is the set of paths between different nodes, (i, j) ∈ A; P is the set of pick-up / pick-up points; D is the set of drop-off / delivery points; n is the number of pick-up / pick-up points; It means for any parameter value; x ijk is a 0-1 variable, which is 1 if vehicle k passes through path (i, j), otherwise it is 0; Formula (1) is used to ensure that all the selected passenger and cargo transportation needs can be met; Formulas (2)-(4) are flow balance constraints, which are used to ensure that the vehicle must start from the starting transportation site and return to that point, and meet the requirements of arriving at and leaving each customer point; Formula (5) is a connection constraint, which is used to ensure that the starting point and the end point of the same demand must be served by the same vehicle.

5. The method according to claim 3, characterized in that: The service time window constraints specifically include: ； Among them, the symbols in each formula refer to: x ijk is the driving path of the vehicle between demand points i and j; s ik is the starting service time of vehicle k at demand point i; i is the service time; t ij is the travel time of the vehicle between demand points i and j; V is the set of all nodes or locations that provide demand or service; a i Set the earliest service start time for demand point i; b i Set the latest service start time for demand point i; Formula (6) ensures that vehicle k can reach the destination point i through path x after completing the service at the destination point i. ijk The time to arrive at the next service point j must not be greater than the service start time of that point; Formula (7) is the vehicle transportation priority constraint, that is, for each passenger and freight demand, the service order of its passenger pick-up demand must be earlier than the passenger drop-off / delivery demand; Formula (8) is the demand point service time window constraint, that is, the time s when vehicle k arrives at demand point i to start service ik , no earlier than a i , and cannot be later than b i .

6. The method according to claim 3, characterized in that The vehicle capacity constraints specifically include: ； The symbols in the formulas refer to: Q 1 ik is the total passenger load of vehicle k when it leaves demand point i; Q 2 ik is the total cargo load of vehicle k when it leaves demand point i; , is the passenger demand at demand points i and j; , is the demand for goods at demand points i and j; q The rated capacity of the vehicle, i.e. the maximum number of passengers and standard cargo items that the vehicle can carry; Formula (9) and Formula (10) respectively ensure that vehicle k can reach the destination through path x after completing the service of demand point j. ijk When arriving at the next service point j, its passenger / cargo load when leaving point j needs to be equal to the sum of the passenger / cargo load after leaving point i and the passenger / cargo loading and unloading volume at point j; formula (11) ensures that the total load of vehicle k at point i shall not be greater than the rated capacity of the vehicle.

7. The method according to claim 1, characterized in that The objective function is specifically as follows: ； Symbols in the formula: k is the vehicle serial number; K is the vehicle set; i, j are nodes or locations that provide demand or service; V is the set of all nodes or locations that provide demand or service; t ij is the travel time of the vehicle between demand points i and j; x ijk is the driving path of the vehicle between demand points i and j; s jk , s ik The time when vehicle k arrives at demand point j and starts serving i; is the passenger demand at demand point i; T represents the set of passenger boarding and alighting point pairs; T + Represents the boarding point; α is the weight coefficient of the total vehicle travel time; β is the weight coefficient of the average passenger travel time.

8. The method according to claim 1, characterized in that The method of solving the adaptive large neighborhood search algorithm specifically includes the following steps: Step i: Input a feasible initial path plan and relevant parameters of the algorithm; Step II: define the destruction operator set and the repair operator set corresponding to the destruction operation and the repair operation; Step iii: define the neighborhood size set; Step IV: Initialize the temperature and operation weights; Step V: Construct the initial solution; Step ⅵ: Before reaching the maximum number of iterations, perform the following operations: (ⅵ-1) Check the solution pool. If it is empty, add a new solution and the corresponding cost; (ⅵ-2) Select new solutions from the solution pool to ensure that each iteration starts from a different initial solution; (ⅵ-3) roulette wheel selection operation; (ⅵ-4) Select the neighborhood size and perform damage repair operations; (ⅵ-5) Calculate the cost of the new solution and calculate the acceptance probability by simulated annealing method; (ⅵ-6) accepting the new solution as the current solution according to the acceptance criteria; (ⅵ-7) Update the optimal solution; (ⅵ-8) Update operation weights; (ⅵ-9) Cooling down, so that the algorithm transitions from the exploration stage to the convergence stage; Step ⅶ: When the maximum number of iterations is reached, output the current optimal solution as the final path plan.

9. The method according to claim 1, characterized in that: The adaptive large neighborhood search algorithm includes the following nine destruction operators: random request removal, random route removal, worst distance removal, worst time removal, worst neighborhood removal, Shaw removal, proximity-based removal, time-based removal, and request-based removal.

10. The method according to claim 1, characterized in that The adaptive large neighborhood search algorithm includes the following three repair operators: random insertion, greedy insertion, and regretful insertion.

Images