A planning method and system for a simultaneous pick-up and delivery vehicle routing problem with multiple time windows under uncertain pickup demand

Abstract

The application provides a planning method and system for a simultaneous pickup and delivery vehicle routing problem with multiple time windows under uncertain pickup demand, comprising the following steps: obtaining distribution information of customer points to be served; constructing a fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle routing problem with multiple time windows based on credibility measure theory; solving the model by using a genetic and harmony search algorithm to obtain a pre-optimized distribution scheme; scheduling a vehicle to travel according to the pre-optimized distribution scheme to determine the pickup demand of each customer; judging whether there is a failure point with a vehicle residual loading capacity less than the actual pickup demand of the customer, if there is a failure point, adopting a failure point service strategy to re-optimize the path planning for the failure point within the remaining serviceable time window; if there is no failure point, storing the determined distribution information of the customer points and outputting an optimal distribution scheme. The application makes full use of the loading capacity of the vehicle and maximizes the economicization, efficiency and greenization of urban logistics distribution.

Description

Technical Field

[0001] This invention relates to the field of logistics and distribution technology, and in particular to a method and system for planning the route of simultaneous pickup and delivery vehicles with multiple time windows under uncertain pickup demand. Background Technology

[0002] In recent years, with the adjustment of urban industrial layout, the upgrading of residents' consumption level, and the promotion of e-commerce models, urban logistics and distribution have gradually become the focus of the logistics industry. However, high distribution costs, waste of transportation resources, and low distribution efficiency remain the three major problems restricting its development.

[0003] Vehicle routing, a crucial component of urban logistics and distribution, is an effective means of addressing the aforementioned challenges. In particular, the Vehicle Routing Problem with Simultaneous Pickup and Delivery (VRPSPD) planning, which combines forward and reverse logistics, can significantly save time and delivery costs, improve logistics efficiency, align with the low-carbon economy concept, and has broad application prospects. Examples include beverage suppliers delivering orders while simultaneously collecting expired beverages and reusable containers (empty bottles); glass factories delivering glass orders while collecting broken glass; and medical supplies delivering orders while collecting expired medications. However, in practical applications, delivery demand can be precisely calculated based on orders, while pickup demand often depends on rough estimates from customers, exhibiting significant uncertainty. This can easily lead to additional vehicle turnarounds and wasted capacity, increasing delivery costs. Furthermore, vehicle emissions are often overlooked during route planning, placing pressure on environmental governance. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for planning the route of simultaneous pickup and delivery vehicles with multiple time windows under uncertain pickup demand. It considers the uncertainty of customer pickup demand, the time window of customer service, and the carbon emissions of the vehicle delivery process, and constructs a VRPSPD optimization model that is more in line with practical applications. By formulating a failure point service strategy, it makes full use of the vehicle's loading capacity and maximizes the economy, efficiency, and greenness of urban logistics distribution.

[0005] The present invention achieves the above-mentioned technical objectives through the following technical means.

[0006] A method for planning the routing of delivery vehicles simultaneously with multiple time windows under uncertain pickup demand includes the following steps:

[0007] Obtain delivery information from customers waiting to be served, and reduce the estimation error of pickup demand by historically served customers based on stored data;

[0008] A fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle routing problem with multiple time windows is constructed based on the credibility measurement theory.

[0009] The established model is solved using a genetic and acoustic search algorithm to obtain a pre-optimized delivery plan;

[0010] Dispatch vehicles to travel according to the pre-optimized delivery plan and determine the pickup demand of each customer;

[0011] Determine if there is a failure point where the vehicle's remaining load capacity is less than the customer's actual pickup demand. If a failure point exists, adopt a failure point service strategy and re-optimize the route planning for the failure point within the remaining service time window. If there is no failure point, store the customer's location and determine the delivery information, and output the optimal delivery plan.

[0012] Furthermore, the delivery information of the customer points to be served includes geographical location, multiple non-overlapping service time windows, delivery demand, and estimated pickup demand, wherein the estimated pickup demand is represented by a triangular fuzzy number.

[0013] Furthermore, the step of reducing the estimation deviation of pickup demand by historically served customers based on stored data specifically involves: for historically served customers, comparing the estimated pickup demand obtained from previous services with the determined pickup demand based on stored data to extract estimation features and reduce the estimation deviation of pickup demand in the current service.

[0014] Furthermore, the fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle routing problem with multiple time windows, constructed based on the credibility measure theory, is as follows:

[0015] The objective function is: min TC 0 ,

[0016] Among them, TC 0 The optimistic cost at confidence level α0;

[0017] The constraints are:

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] In the formula:

[0027] The total delivery cost can be expressed as:

[0028]

[0029] in: The total fuzzy fuel consumption generated by all customer nodes serving the vehicle; H represents the set of homogeneous fleets; R represents the set of customer nodes; i=0 or j=0 indicates the origin and destination positions; The fuzzy load represents the distance vehicle h travels from the i-th customer node to the j-th customer node; λ is the fuel calorific value coefficient; d ij The distance traveled by the vehicle from the i-th customer node to the j-th customer node is represented by k; engine friction coefficient is k; engine speed is N; engine displacement is V; v ij γ represents the speed at which the vehicle travels from the i-th customer node to the j-th customer node; γ is the transmission system coefficient; α is the gravitational constant; β is the air resistance coefficient; m0 represents the empty vehicle mass.

[0030] R0 = R{0} represents the set of customer nodes that do not include the start and end points; x 0jh Let represent the decision variables for vehicle h traveling from the starting point to the j-th customer node; c1 is the unit fixed cost of the vehicle; c2 is the unit fuel consumption cost; c3 is the unit carbon emission cost; e is the carbon emission coefficient; c4 is the unit waiting time cost for the vehicle arriving early; c5 is the unit penalty cost for the vehicle arriving late; W j Let j be the set of time windows for the j-th customer node; The earliest service start time that the j-th customer node can accept is the q-th time. T represents the qth latest service start time acceptable to the j-th customer node. j Let the time when the vehicle arrives at the j-th customer node be denoted as . s i Let t be the service duration of the i-th customer node. ij This represents the travel time for a vehicle to travel from the i-th customer node to the j-th customer node.

[0031] Cr is a credibility measure; α0 is the confidence level predetermined by the company's decision-makers.

[0032] β0 represents the confidence level predetermined by the firm's decision-makers; d j Let M be the delivery demand of the j-th customer node; M is the maximum load capacity of the vehicle. p represents the estimated pickup demand for the j-th customer node. 1j p represents the minimum pickup requirement for the j-th customer node. 2j p represents the most likely pickup demand at the j-th customer node. 3j Let x be the maximum pickup demand of the j-th customer node. j0h M represents the decision variable for vehicle h returning from the j-th customer node to the destination; 0jh This represents the initial load of vehicle h from the origin to the j-th customer node; v l Minimum speed; v m That is the maximum speed.

[0033] Furthermore, the established model is solved using a genetic harmony search algorithm to obtain a pre-optimized delivery plan, specifically including the following steps:

[0034] S031: Initialize the algorithm parameters, including population size T and crossover probability P. c Probability of mutation P m The number of iterations G, the size of the sound memory bank HMS, the range of the probability of the sound memory bank value HMCR, the range of the fine-tuning probability PCR, the current creation count t, and the number of creations NI;

[0035] S032: Generating the initial memory bank HM using a genetic algorithm:

[0036] Customer points are represented by 1, 2, 3, ... r using natural number encoding. A random customer sequence is generated and sorted. A total of g+1 zeros representing distribution centers are inserted at the beginning, end, and different positions in the sequence to generate the initial chromosome.

[0037] Then, screening is performed based on fuzzy capacity opportunity constraints and fuzzy cost opportunity constraints. If the constraints are met, the chromosome is retained. This process is repeated until the population size reaches T, and the initial population is denoted as P. T =X1,X2,X m ,…X T m = 1, 2, ..., T, where individual X m Represent a feasible delivery plan; calculate the fitness value for each individual. Among them, TC(X) m ) 0 For individual X m Given a confidence level α 0 The minimum expected total cost is given, and a higher fitness value indicates a better individual.

[0038] A combination of roulette wheel selection and elite retention strategies is used to select superior individuals for the next generation. Then, two parent chromosomes are randomly selected from these individuals for crossover. The crossover probability P is set as follows: c Compare P with a randomly generated random number q1 within (0,1). c If P > q1, then perform sequential crossover on the two parent chromosomes; if P c If ≤q1, then remain unchanged;

[0039] Perform a mutation operation on any selected chromosome: if the mutation probability P m If a random number q2 is randomly generated within the range (0, 1), then two gene exchange positions on that chromosome are randomly selected to generate a new chromosome after mutation. It is then determined whether the new chromosome satisfies the fuzzy capacity chance constraint and the fuzzy cost chance constraint. If it does, it is retained; otherwise, the mutation probability P... m A random number q2 generated within the range ≤ (0, 1) remains unchanged;

[0040] Repeated crossover and mutation operations are used to update the offspring population. When the number of iterations G is reached, the genetic algorithm stops. Each individual in the final population is sorted in descending order of fitness value, and the best HMS individual is selected as the initial solution for the harmony memory bank.

[0041] S033: Update memory access probability (HMCR) and fine-tuning probability (PCR):

[0042] The HMCR (Hypermeability Code Rate) is dynamically adjusted using the following formula:

[0043]

[0044] Among them, HMCR min Minimum probability of a value being retrieved from the acoustic memory bank; HMCR max The maximum probability of a value being retrieved from the sound memory bank; t is the current creation count;

[0045] The fine-tuning probability PCR is dynamically adjusted using the following formula:

[0046]

[0047] Among them, PCR min This represents the minimum value of the fine-tuning probability; PCR max This represents the maximum value of the fine-tuning probability;

[0048] S034: Generate a new harmony, specifically:

[0049] Generate a random number q3 of rand(0,1). If q3≤HMCR, then randomly select a harmony from the harmony memory HM; otherwise, randomly generate a harmony outside the harmony memory HM.

[0050] The selected or generated harmony is perturbed by PCR probability, and a neighborhood search of two-point exchange and single-point insertion is performed in sequence to generate new harmony with full fuzzy capacity opportunity constraints and fuzzy cost opportunity constraints.

[0051] S035: Update the harmony memory bank;

[0052] If the objective function value of the new harmony is less than the worst harmony in the HM, then replace it and re-sort the HM; otherwise, do not make any changes.

[0053] S036: If t > NI, stop the iteration and output the optimal pre-optimized delivery plan in HM; otherwise, t = t + 1 and return to S033.

[0054] Furthermore, the failure point service strategy is that if the vehicle travels along the pre-optimized route and the remaining load cannot meet the pickup needs of the i-th customer node, then only the i-th customer node will be provided with delivery service, and then simultaneous pickup and delivery services will be provided to all subsequent customer nodes.

[0055] Furthermore, the failure point re-optimization path planning is a pickup vehicle path problem with a time window, and a re-optimization service penalty cost is added to the objective function when constructing the model.

[0056] A system for planning the route of simultaneous pickup and delivery vehicles with multiple time windows under uncertain pickup demand includes a data transmission and reception module, a road condition prediction module, a route planning module I, a judgment module, a route planning module II, a delivery management module, and a vehicle scheduling module.

[0057] The data transceiver module is used to receive delivery information from multiple customers on the user end and send early warning information about the deviation of the customer's estimated pickup demand; the road condition prediction module is connected to the data transceiver module and is used to obtain the expected vehicle speed on the path between the delivery center and the customer, between customers, and between the customer and the delivery center.

[0058] The route planning module I is connected to the traffic condition prediction module and is used to pre-optimize the route planning for the received customer points; the route planning module I inputs the pre-optimized route planning into the vehicle dispatching module, which is used to dispatch multiple delivery vehicles to execute the delivery plan; the vehicle dispatching module inputs the delivery plan into the judgment module, which is used to determine whether there are any service failure points;

[0059] The route planning module II is used to re-optimize the route planning at failure points; the route planning module II inputs the re-optimized route planning into the vehicle scheduling module, which is used to schedule multiple delivery vehicles to execute the optimized delivery plan; the vehicle scheduling module inputs the optimized delivery plan into the delivery management module, which is used to improve the accuracy of customers' estimation of pickup demand; the delivery management module is connected to the data transceiver module.

[0060] Furthermore, the route planning module I includes a model building unit I, an algorithm solving unit, and a delivery plan generation unit I; the model building unit I is used to construct the fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle route problem; the algorithm solving unit uses the genetic and acoustic search algorithm to solve the model; and the delivery plan generation unit I is used to generate a pre-optimized delivery plan.

[0061] The route planning module II includes a model building unit II and a delivery plan generation unit II. The model building unit II is used to build a re-optimization model for failure points; the delivery plan generation unit II is used to generate a re-optimized delivery plan.

[0062] Furthermore, the delivery management module includes a storage unit, a retrieval unit, and an early warning unit. The storage unit is used to store actual historical data of customers who have been served; the retrieval unit is used to calculate the deviation of customers' estimated pickup demand; and the early warning unit has a built-in threshold to alert customers.

[0063] The beneficial effects of this invention are as follows:

[0064] The method and system for planning the route of delivery vehicles with multiple time windows under uncertain pickup demand described in this invention can fully utilize the remaining capacity of delivery vehicles and effectively avoid waste of transportation capacity, while also taking into account carbon emissions during vehicle delivery. This can maximize the economy, efficiency and greenness of urban logistics distribution. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings described below are some embodiments of the present invention. For those skilled in the art, it is obvious that other drawings can be obtained from these drawings without creative effort.

[0066] Figure 1 This is a flowchart illustrating the method for planning the route of delivery vehicles with multiple time windows under uncertain pickup demand, as described in this invention.

[0067] Figure 2 This is a flowchart of the genetic and acoustic search algorithm described in this invention.

[0068] Figure 3 This is a schematic diagram of the simultaneous pickup and delivery vehicle route planning system with multiple time windows under uncertain pickup demand as described in this invention.

[0069] Figure 4This is a schematic diagram illustrating the working principle of the simultaneous pickup and delivery vehicle route planning system with multiple time windows under uncertain pickup demand, as described in this invention. Detailed Implementation

[0070] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0071] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0072] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "axial," "radial," "vertical," "horizontal," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0073] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0074] like Figure 1 As shown, the method for planning the route of simultaneous pickup and delivery vehicles with multiple time windows under uncertain pickup demand, as described in this invention, includes the following:

[0075] S01: Obtain delivery information for customers waiting to be served, and reduce the estimation error of pickup demand by historically served customers based on stored data;

[0076] Specifically, delivery companies obtain delivery information for customers to be served, including geographical location, multiple non-overlapping service time windows, delivery demand, and estimated pickup demand. The estimated pickup demand is represented by triangular fuzzy numbers, namely the minimum pickup demand, the most likely pickup demand, and the maximum pickup demand. If the customer is a previously served customer, the estimated pickup demand obtained from past services is compared with the confirmed pickup demand based on stored data to extract estimation features and reduce the estimation deviation of pickup demand in the current service, thereby reducing the probability of re-optimization from the customer's perspective.

[0077] S02: Constructing a fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle routing problem with multiple time windows based on the credibility measurement theory;

[0078] Specifically, the total delivery cost is the sum of vehicle fixed costs, fuel consumption and carbon emission costs, and time window penalty costs. Fuel consumption and carbon emission costs are directly proportional to fuel consumption, and the calculation of fuel consumption fully considers the impact of vehicle load and speed. As shown in the following formula:

[0079]

[0080] In the formula:

[0081] The total fuzzy fuel consumption generated by all customer nodes serving the vehicle; H represents the set of homogeneous fleets; R represents the set of customer nodes; i=0 or j=0 indicates the origin and destination positions; The fuzzy load represents the distance vehicle h travels from the i-th customer node to the j-th customer node; λ is the fuel calorific value coefficient; d ij The distance traveled by the vehicle from the i-th customer node to the j-th customer node is represented by k; engine friction coefficient is k; engine speed is N; engine displacement is V; v ij γ represents the speed at which the vehicle travels from the i-th customer node to the j-th customer node; γ is the transmission system coefficient; α is the gravitational constant; β is the air resistance coefficient; m0 represents the empty vehicle mass.

[0082] The total delivery cost is expressed as follows:

[0083]

[0084] In the formula: R0 = R{0} represents the set of customer nodes that do not include the start and end points; x 0jh Let represent the decision variables for vehicle h traveling from the starting point to the j-th customer node; c1 is the unit fixed cost of the vehicle; c2 is the unit fuel consumption cost; c3 is the unit carbon emission cost; e is the carbon emission coefficient; c4 is the unit waiting time cost for the vehicle arriving early; c5 is the unit penalty cost for the vehicle arriving late; W jLet j be the set of time windows for the j-th customer node; The earliest service start time that the j-th customer node can accept is the q-th time. T represents the qth latest service start time acceptable to the j-th customer node. j Let the time when the vehicle arrives at the j-th customer node be denoted as . s i Let t be the service duration of the i-th customer node. ij This represents the travel time for a vehicle to travel from the i-th customer node to the j-th customer node.

[0085] Based on the theory of credibility measurement, the objective can only be achieved if the constraints are met at a certain confidence level. Therefore, fuzzy cost-opportunity constraints are introduced:

[0086]

[0087] In the formula: Cr is the credibility measure; α0 is the confidence level predetermined by the enterprise decision-maker; TC 0 This represents the optimistic cost at a confidence level of α0.

[0088] Based on the above analysis, the following fuzzy chance-constrained programming model for the simultaneous pickup and delivery vehicle routing problem with multiple time windows is constructed based on the reliability measure theory:

[0089] The objective function is: min TC 0 ,

[0090] Among them, TC 0 The optimistic cost at confidence level α0;

[0091] The constraints are:

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100] In the formula:

[0101] Cr is a credibility measure; α0 is the confidence level predetermined by the company's decision-makers.

[0102] β0 represents the confidence level predetermined by the firm's decision-makers; d j Let M be the delivery demand of the j-th customer node; M is the maximum load capacity of the vehicle. p represents the estimated pickup demand for the j-th customer node. 1j p represents the minimum pickup requirement for the j-th customer node. 2j p represents the most likely pickup demand at the j-th customer node. 3j Let x be the maximum pickup demand of the j-th customer node. j0h M represents the decision variable for vehicle h returning from the j-th customer node to the destination; 0jh This represents the initial load of vehicle h from the origin to the j-th customer node; v l Minimum speed; v m That is the maximum speed.

[0103] Formula (1) represents the total cost No more than the optimistic cost TC 0 The confidence level is greater than or equal to the preset confidence level α0; Formula (2) indicates that the confidence level of vehicle h after serving customer j is greater than or equal to the preset confidence level β0; Formula (3) is the path node constraint, indicating that each customer has one and only one vehicle to serve once, and the vehicle has only one service path when it is used and satisfies the condition of starting from the starting point and returning to the destination; Formula (4) is the vehicle load constraint, that is, the initial load of the vehicle starting from the starting point must meet the delivery demand of the customer being served and must not exceed the maximum load limit of the vehicle; Formula (5) indicates that the customer can only be selected in one time window; Formula (6) is the vehicle speed limit constraint; Formulas (7) and (8) represent the decision variable attribute constraints.

[0104] S03: The established model is solved using a genetic harmony search algorithm to obtain a pre-optimized delivery plan;

[0105] Specifically, the genetic harmony search algorithm utilizes the inherent parallelism of genetic algorithms and the characteristic of rapidly improving solutions in the early stages to generate an initial memory bank, thereby compensating for the high dependence of the harmony search algorithm on the initial solution. The steps are as follows: Figure 2 As shown:

[0106] S031: Initialize the algorithm parameters, including population size T and crossover probability P. c Probability of mutation P m The number of iterations G, the size of the sound memory bank HMS, the range of the probability of the sound memory bank value HMCR, the range of the fine-tuning probability PCR, the current creation count t, and the number of creations NI;

[0107] S032: Generating the initial memory bank HM using a genetic algorithm:

[0108] Customer points are represented by 1, 2, 3, ... r using natural number encoding. A random customer sequence is generated and sorted. A total of g+1 zeros representing distribution centers are inserted at the beginning, end, and different positions in the sequence to generate the initial chromosome.

[0109] Then, screening is performed based on fuzzy capacity opportunity constraints and fuzzy cost opportunity constraints. If the constraints are met, the chromosome is retained. This process is repeated until the population size reaches T, and the initial population is denoted as P. T =X1,X2,X m ,…X T m = 1, 2, ..., T, where individual X m Represent a feasible delivery plan; calculate the fitness value for each individual. Among them, TC(X) m ) 0 For individual X m Given a confidence level α 0 The minimum expected total cost is given, and a higher fitness value indicates a better individual.

[0110] A combination of roulette wheel selection and elite retention strategies is used to select superior individuals for the next generation. Then, two parent chromosomes are randomly selected from these individuals for crossover. The crossover probability P is set as follows: c Compare P with a randomly generated random number q1 within (0,1). c If P > q1, then perform sequential crossover on the two parent chromosomes. Sequential crossover involves randomly selecting two crossover points. If the gene at the crossover point is not 0, the crossover point is shifted left or right until the gene at the crossover point is 0. Then, the sequence segment between the two crossover points where the gene is 0 in the first parent chromosome is retained. The portion of the second parent chromosome that overlaps with the retained sequence in the first parent chromosome is removed. The remaining genes are then sequentially inserted into the first chromosome to form a daughter chromosome. Similarly, a daughter chromosome of the second chromosome is generated. If P c If ≤q1, then remain unchanged;

[0111] Perform a mutation operation on any selected chromosome: if the mutation probability P m If a random number q2 is randomly generated within the range (0, 1), then two gene exchange positions on that chromosome are randomly selected to generate a new chromosome after mutation. It is then determined whether the new chromosome satisfies the fuzzy capacity chance constraint and the fuzzy cost chance constraint. If it does, it is retained; otherwise, the mutation probability P... m A random number q2 generated within the range ≤ (0, 1) remains unchanged;

[0112] Repeated crossover and mutation operations are used to update the offspring population. When the number of iterations G is reached, the genetic algorithm stops. Each individual in the final population is sorted in descending order of fitness value, and the best HMS individual is selected as the initial solution for the harmony memory bank.

[0113] S033: Update memory access probability (HMCR) and fine-tuning probability (PCR):

[0114] The HMCR (Hypermeability Code Rate) is dynamically adjusted using the following formula:

[0115]

[0116] Among them, HMCR min Minimum probability of a value being retrieved from the acoustic memory bank; HMCR max The maximum probability of a value being retrieved from the sound memory bank; t is the current creation count;

[0117] The fine-tuning probability PCR is dynamically adjusted using the following formula:

[0118]

[0119] Among them, PCR min This represents the minimum value of the fine-tuning probability; PCR max This represents the maximum value of the fine-tuning probability;

[0120] S034: Generate a new harmony, specifically:

[0121] Generate a random number q3 of rand(0,1). If q3≤HMCR, then randomly select a harmony from the harmony memory HM; otherwise, randomly generate a harmony outside the harmony memory HM.

[0122] The selected or generated harmony is perturbed by PCR probability, and a neighborhood search of two-point exchange and single-point insertion is performed in sequence to generate new harmony with full fuzzy capacity opportunity constraints and fuzzy cost opportunity constraints.

[0123] S035: Update the harmony memory bank;

[0124] If the objective function value of the new harmony is less than the worst harmony in the HM, then replace it and re-sort the HM; otherwise, do not make any changes.

[0125] S036: Algorithm termination judgment: If t > NI, stop the iteration and output the optimal pre-optimized delivery plan in HM; otherwise, t = t + 1 and return to S033.

[0126] S04: Dispatch vehicles to drive according to the pre-optimized delivery plan and determine the pickup demand of each customer;

[0127] Specifically, delivery companies dispatch vehicles according to pre-optimized delivery plans, serving each customer in turn while simultaneously determining their pickup needs.

[0128] S05: Determine if there is a failure point, i.e., whether the remaining load capacity of the vehicle is less than the customer's actual pickup demand. If so, adopt the failure point service strategy and re-optimize the route planning for the customer in the remaining service time window. Otherwise, store the customer's point to determine the delivery information and output the optimal delivery plan.

[0129] The failure point service strategy is that if the vehicle travels along the pre-optimized route and the remaining load cannot meet the pickup needs of the i-th customer node, then only the i-th customer node will be delivered, and then simultaneous pickup and delivery services will be provided to all subsequent customer nodes.

[0130] Finally, the number of failure points was counted, and to meet the pickup service demand, a unified re-optimization route planning was performed. The problem then transforms into a pickup vehicle routing problem with time windows. A re-optimization model is constructed with the objective of minimizing the sum of vehicle fixed costs, fuel consumption and carbon emission costs, time window penalty costs, and re-optimization service penalty costs. Here, the customer pickup demand is fixed, and the vehicle's fuel consumption is calculated using the following formula:

[0131]

[0132] Where: F is the total fuel consumption resulting from all service failure points of the vehicle; R 1 This represents the set of failure points; when s=0 or u=0, it represents the start and end positions. d su v represents the distance the vehicle travels from the s-th failure point to the u-th failure point; su M represents the speed at which the vehicle travels from failure point s to failure point u. suh Let h be the load on vehicle h as it travels from failure point s to failure point u.

[0133] The objective function is:

[0134]

[0135] The constraints are:

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144]

[0145] In the formula: This represents the set of failure points excluding the start and end points; x 0uh The decision variable represents the journey of vehicle h from the starting point to the u-th failure point. This represents the set of remaining optional time windows for the u-th failure point; The qth earliest service start time acceptable at the uth failure point; T is the qth latest service start time acceptable at the uth failure point; u Let the time when the vehicle reaches the u-th failure point be denoted as . s s The service duration at the s-th failure point, t su This represents the travel time of the vehicle from the s-th failure point to the u-th failure point; c6 represents the penalty cost for optimizing the unit weight of goods; Let x be the pickup demand at failure point u; uoh M represents the decision variable for vehicle h to travel from the u-th failure point to the destination; 0uh Let h be the initial load from the starting point to the u-th failure point; Formula (9) indicates that each failure point can only be served by one vehicle; Formula (10) indicates that failure points on the same route are served by the same delivery vehicle and each failure point is served only once. When a vehicle is used, it must start from the starting point and return to the destination; Formula (11) indicates that the initial load of vehicle h when leaving the starting point is 0; Formula (12) indicates that the sum of the pickup demand of vehicle h at failure points does not exceed the maximum load limit of the vehicle; Formula (13) indicates that only one time window can be selected for a failure point; Formula (14) indicates the vehicle speed limit constraint; Formulas (15), (16) and (17) are decision variable attribute constraints.

[0146] like Figure 3 As shown, a system for planning the route of a delivery vehicle with multiple time windows under uncertain pickup demand includes a data transmission and reception module, a road condition prediction module, a route planning module I, a judgment module, a route planning module II, a delivery management module, and a vehicle scheduling module.

[0147] The data transceiver module is used to receive delivery information from multiple customers on the user end and send early warning information about the deviation of the customer's estimated pickup demand; the road condition prediction module is connected to the data transceiver module and is used to obtain the expected vehicle speed on the path between the delivery center and the customer, between customers, and between the customer and the delivery center.

[0148] The route planning module I is connected to the traffic condition prediction module and is used to pre-optimize the route planning for the received customer points; the route planning module I inputs the pre-optimized route planning into the vehicle dispatching module, which is used to dispatch multiple delivery vehicles to execute the delivery plan; the vehicle dispatching module inputs the delivery plan into the judgment module, which is used to determine whether there are any service failure points;

[0149] The route planning module II is used to re-optimize the route planning at failure points; the route planning module II inputs the re-optimized route planning into the vehicle scheduling module, which is used to schedule multiple delivery vehicles to execute the optimized delivery plan; the vehicle scheduling module inputs the optimized delivery plan into the delivery management module, which is used to improve the accuracy of customers' estimation of pickup demand; the delivery management module is connected to the data transceiver module.

[0150] The route planning module I includes a model building unit I, an algorithm solving unit, and a delivery plan generation unit I. The model building unit I is used to build the fuzzy chance constraint planning model for the simultaneous pickup and delivery vehicle route problem. The algorithm solving unit uses the genetic and acoustic search algorithm to solve the model. The delivery plan generation unit I is used to generate a pre-optimized delivery plan.

[0151] The path planning module II includes a model building unit II, an algorithm solving unit, and a delivery plan generation unit II. The model building unit II is used to build a re-optimization model for failure points; the delivery plan generation unit II is used to generate re-optimized delivery plans.

[0152] The delivery management module includes a storage unit, a retrieval unit, and an early warning unit. The storage unit stores actual historical data of customers who have been served. The retrieval unit calculates the deviation of customers' estimated pickup demand. The early warning unit has a threshold value to alert customers.

[0153] The working principle of the simultaneous pickup and delivery vehicle route planning system with multiple time windows under uncertain pickup demand described in this invention is as follows: Figure 4As shown, the data transceiver module is used to receive delivery information from multiple customers on the user end, including geographical location, multiple non-overlapping serviceable time windows, delivery demand, and estimated pickup demand. The road condition prediction module obtains the customer's geographical location information through the data transceiver module and generates the expected vehicle speed on the paths between the delivery center and the customer, between customers, and between the customer and the delivery center based on real-time road conditions. The path planning module I can reduce the estimation deviation of the customer's pickup demand based on the historical data stored in the delivery management module. The model building unit I constructs a fuzzy chance constraint programming model for the simultaneous pickup and delivery vehicle path problem with multiple time windows under uncertain pickup demand. The algorithm solving unit uses the genetic harmony search algorithm to solve the problem. The delivery plan generation unit I outputs a pre-optimized delivery plan and transmits it to the vehicle scheduling module. The vehicle dispatch module schedules vehicles to travel according to the pre-optimized delivery plan, determines the actual pickup demand of each customer, and transmits it to the delivery management module. The judgment module determines whether there are service failure points; if so, it adopts the failure point service strategy. The path planning module II constructs the failure point re-optimization model through the model building unit II, solves the algorithm, generates a re-optimized delivery plan in the delivery plan generation unit II, and then executes it by the vehicle dispatch module to meet the pickup demand of the failure point. The delivery management module receives the actual pickup demand of customers served by the vehicle dispatch module and stores it and other delivery information in the storage unit. The extraction unit calculates the deviation value of the customer's estimated pickup demand and compares it with the built-in threshold of the early warning unit. If it is greater than the threshold, it feeds back to the customer through the data transceiver module.

[0154] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0155] The detailed descriptions listed above are merely specific illustrations of feasible embodiments of the present invention and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for planning a multi-time windowed simultaneous pick-up and delivery vehicle routing problem under uncertain pickup demand, characterized in that, The method comprises the following steps: obtaining delivery information of a to-be-served customer point, and reducing the estimation deviation of the historical served customer on the pickup demand according to the stored data, specifically: for the historical served customer, the estimated pickup demand obtained in the past service is compared with the determined pickup demand according to the stored data to extract the estimation feature, so as to reduce the estimation deviation of the pickup demand in the current service; a fuzzy chance-constrained programming model of the simultaneous pickup and delivery vehicle routing problem with multiple time windows is constructed based on the credibility measure theory, as follows: Objective function is: minTC 0 , where TC 0 is the optimistic cost at a confidence level a0; The constraint condition is: In the formula: The total distribution cost can be expressed as: wherein: the total amount of fuzzy fuel consumption generated by all customer nodes served by the vehicle; H represents a homogeneous fleet set; R represents a customer node set; i = 0 or j = 0 represents the start and end point locations; represents the fuzzy load of vehicle h from the i-th customer node to the j-th customer node; λ is the fuel heat value coefficient; d ij represents the distance of vehicle from the i-th customer node to the j-th customer node; k is the engine friction coefficient; N is the engine speed; V is the engine displacement; v ij represents the driving speed of vehicle from the i-th customer node to the j-th customer node; γ is the transmission system coefficient; α is the gravity constant; β is the air resistance coefficient; m0 represents the empty car mass; R0= R\{0} represents a customer node set not containing the start and end point positions; x 0jh is a decision variable representing the vehicle h driving from the start point to the jth customer node; c1 is the unit fixed cost of the vehicle; c2 is the unit fuel consumption cost; c3 is the unit carbon emission cost; e is the carbon emission coefficient; c4 is the unit time waiting cost of the vehicle arriving early; c5 is the unit time penalty cost of the vehicle arriving late; W j is a time window set of the jth customer node; is the qth earliest start service time acceptable to the jth customer node; is the qth latest start service time acceptable to the jth customer node; T j is the time when the vehicle arrives at the jth customer node, represented as s i is the service duration of the ith customer node, t ij represents the driving time of the vehicle from the ith customer node to the jth customer node; Cr is the credibility measure; and a0 is the confidence level determined by the enterprise decision maker in advance. β0 is the confidence level predetermined by the decision maker; d j is the delivery demand of the jth customer node; M is the maximum load of the vehicle; is the estimated pickup demand of the jth customer node, p 1j is the minimum pickup demand of the jth customer node, p 2j is the most likely pickup demand of the jth customer node, p 3j is the maximum pickup demand of the jth customer node, x j0h is the decision variable indicating that vehicle h returns to the terminal from the jth customer node; M 0jh is the initial load of vehicle h from the starting point to the jth customer node; v l is the minimum vehicle speed; v m is the maximum vehicle speed; The genetic and harmony search algorithm is used to solve the established model to obtain a pre-optimized delivery scheme. The vehicle is dispatched to travel according to the pre-optimized delivery scheme, and the pickup demand of each customer is determined. It is judged whether there is a failure point with a vehicle remaining loading capacity less than the actual pickup demand of the customer, if there is a failure point, a failure point service strategy is adopted, and a re-optimized path planning is performed for the failure point within the remaining serviceable time window; if there is no failure point, the delivery information of the customer point is stored, and an optimal delivery scheme is output.

2. The method of claim 1, wherein, The delivery information of the to-be-served customer point includes a geographic location, multiple non-overlapping serviceable time windows, a delivery demand and an estimated pickup demand, wherein the estimated pickup demand is represented by a triangular fuzzy number.

3. The method of claim 1, wherein, The genetic and harmony search algorithm is used to solve the established model to obtain a pre-optimized delivery scheme, specifically comprising the following steps: S031: initialize parameters of the algorithm, including population size T, crossover probability P c , mutation probability P m , iteration number G, sound memory bank size HMS, memory bank value probability HMCR value range, fine-tuning probability PCR value range, current creation number t and creation number NI; S032: generating an initial memory bank HM using a genetic algorithm: In the form of natural number coding, the customer point is represented by 1, 2, 3, … r, a customer sequence order is randomly generated, and g+1 0s representing the distribution center are inserted at the beginning and end of the sequence and at different positions in the sequence to generate an initial chromosome; Then according to the fuzzy capacity opportunity constraint and the fuzzy cost opportunity constraint, if the constraint is satisfied, the chromosome is reserved; repeat the above operation until the population size reaches T, and record the initial population as P T = X1, X2, X m ,…X T ,m = 1, 2,…T, wherein, the individual X m represents a feasible distribution scheme; calculate the fitness value of each individual Wherein TC(X m ) 0 is the minimum expected total cost of the individual X m under a given confidence level α 0 , the greater the fitness value indicates that the individual is better; The roulette wheel and the elite reservation strategy are combined to select excellent individuals into the next generation, and two parent chromosomes are randomly selected from the excellent individuals to perform the crossover operation: if the crossover probability P c is compared with a random number q1 generated in (0, 1), and P c > q1, the sequential crossover operation is performed on the two parent chromosomes; if P c ≤ q1, the two parent chromosomes are kept unchanged. Any chromosome selected to perform mutation operation: if the mutation probability P m is greater than or equal to the random number q2 randomly generated within (0, 1), then randomly select two gene exchange positions on the chromosome to generate a new chromosome after mutation, and determine whether the new chromosome satisfies the fuzzy capacity chance constraint and the fuzzy cost chance constraint. If it satisfies, it is retained; if the mutation probability P m is less than the random number q2 randomly generated within (0, 1), it remains unchanged; The crossover operation and mutation operation are repeated to update the child population, and the genetic algorithm stops when the number of iterations G is reached, and the final population is arranged in descending order according to the fitness value, and the best HMS individuals are selected as the initial solution of the harmony memory bank; S033: updating the memory bank value probability HMCR and the fine-tuning probability PCR: The memory bank value probability HMCR is dynamically adjusted by the following formula: where HMCR min HMCR max HMCR The fine-tuning probability PCR is dynamically adjusted by the following formula: where PCR min is the minimum value of the fine-tuning probability; PCR max is the maximum value of the fine-tuning probability; S034: generating a new harmony, specifically: Generate a rand(0, 1) random number q3, if q3HMCR, then a set of harmonies is randomly selected from the harmony memory bank HM, otherwise a set of harmonies is randomly generated outside the harmony memory bank HM; The selected or generated harmonies are disturbed with a PCR probability, and two-point crossover and single-point insertion neighborhood search are sequentially performed to generate new harmonies that meet the fuzzy capacity opportunity constraint and the fuzzy cost opportunity constraint; S035: updating the harmony memory bank; If the objective function value of the new harmony is less than the worst harmony in HM, replace it and reorder HM, otherwise, do not modify; S036: if t>NI, stop iteration, output the optimal pre-optimized delivery scheme in HM, otherwise t=t+1, return to S033.

4. The method of claim 3, wherein, The failure point service strategy is that the vehicle travels along the pre-optimized path, and if the remaining load cannot meet the pickup demand of the ith customer node, only the ith customer node is delivered, and then all subsequent customer nodes of the ith customer node are simultaneously picked up and delivered.

5. The method of claim 1, wherein, The failure point re-optimization path planning is a pickup vehicle path problem with a time window, and a re-optimization service penalty cost is added to the objective function when the model is constructed.

6. A system for planning a method of the simultaneous pick-up and delivery vehicle routing problem with multiple time windows under uncertain demand according to any one of claims 1-5, characterized in that, The data transceiver module, the road condition prediction module, the path planning module I, the judgment module, the path planning module II, the distribution management module and the vehicle scheduling module are included. The data transceiver module is used to receive the distribution information of multiple customers of the user end and send the early warning information of the deviation of the customer's pickup demand estimation; the road condition prediction module is connected with the data transceiver module and is used to obtain the predicted vehicle speed on the path between the distribution center and the customer, the customer and the customer, and the customer and the distribution center; The path planning module I is connected with the road condition prediction module and is used to pre-optimize the path planning of the received customer node; the path planning module I inputs the pre-optimized path planning into the vehicle scheduling module for scheduling multiple distribution vehicles to execute the distribution scheme; The vehicle scheduling module inputs the distribution scheme into the judgment module for judging whether there is a service failure point; The path planning module II is used for re-optimization path planning of the failure point; the path planning module II inputs the re-optimized path planning into the vehicle scheduling module for scheduling multiple distribution vehicles to execute the optimized distribution scheme; The vehicle scheduling module inputs the optimized distribution scheme into the distribution management module for improving the accuracy of the customer's pickup demand estimation; the distribution management module is connected with the data transceiver module.

7. The system for the method of planning of simultaneous pick-up and delivery vehicle routing problem with multiple time windows under uncertain pickup demand according to claim 6, wherein, The path planning module I is provided with a model construction unit I, an algorithm solving unit and a distribution scheme generation unit I; the model construction unit I is used to construct the fuzzy chance constraint programming model of the simultaneous pickup and delivery vehicle path problem; the algorithm solving unit solves the model by using the genetic and sound search algorithm; and the distribution scheme generation unit I is used to generate a pre-optimized distribution scheme; The path planning module II is provided with a model construction unit II and a distribution scheme generation unit II; the model construction unit II is used to construct a re-optimization model of the failure point; and the distribution scheme generation unit II is used to generate a re-optimized distribution scheme.

8. The system for the method of planning of simultaneous pick-up and delivery vehicle routing problem with multiple time windows under uncertain pickup demand according to claim 6, wherein, The distribution management module is provided with a storage unit, an extraction unit and a warning unit; the storage unit is used to store the actual historical data of the served customers; the extraction unit is used to calculate the deviation of the customer's pickup demand estimation; and the warning unit is provided with a threshold for prompting the customer.

Images