Intercity dynamic carpooling scheduling optimization method, device, equipment and system for multiple types of orders

Abstract

The present invention provides an inter-city dynamic carpooling scheduling optimization method, device, equipment and system for multi-type orders, which performs online vehicle scheduling and offline vehicle scheduling for passenger and freight transport according to the multi-type order information and vehicle information obtained at the current moment; all unmatched orders are inserted into the vehicle set in a way that maximizes profit to form a set of solutions to be divided and conquered, and the set of solutions to be divided and conquered is divided into several groups, each group containing at most a certain number of points; different types of orders are processed in each group; the optimal solution obtained in each group is merged into a global feasible solution, and the global feasible solution is optimized by an adaptive large neighborhood search algorithm for passenger and freight transport to obtain a global optimal solution. The average total time for picking up and dropping off passengers on each route of this method is not much different from that of greedy insertion, but the total profit, average vehicle load rate, total number of carpooling passengers picked up, total number of chartered vehicles and total number of goods picked up are all higher than those of existing methods.

Description

Technical Field

[0001] The present invention relates to the technical field of intercity passenger transport planning, and in particular to an intercity dynamic carpooling scheduling optimization method, device, equipment and system for multiple types of orders. Background Art

[0002] The intercity dynamic carpooling problem is a complex variant of the vehicle routing problem. It can also be considered as a special dynamic dial-up car-hailing problem or a dynamic pickup and delivery problem. The intercity dynamic carpooling problem with a single type of order has only one type of order, the fare for each passenger is fixed, and the problem model is relatively simple. Compared with the intercity dynamic carpooling problem with a single type of order, the intercity dynamic carpooling problem with multiple types of orders is different in the problem input. Specifically, the types of orders are added on the basis of the intercity dynamic carpooling problem with a single type of order, and the types of vehicle capacity are also added accordingly. The types of orders include chartered orders, carpooling orders, and cargo orders, and the vehicle capacity includes the number of vehicle seats and the number of cargo capacities. The difference in problem input leads to different problem models. Specifically, the vehicle needs to pick up and drop off chartered orders separately, and when a carpooling order and a cargo order are mixed, the goods need to be picked up first and then delivered. That is, when picking up orders in the starting city, the cargo orders need to be picked up first and then the carpooling orders. When delivering orders in the destination city, the carpooling orders need to be dropped off first and then the cargo orders. At the same time, the maximum cargo capacity constraint of the vehicle is added.

[0003] Due to the characteristics of multiple order types in the problem model, the original algorithm for solving the inter-city dynamic carpooling problem with single-type orders cannot solve this problem well. The current mainstream approach to solving this type of problem is to add heuristic rules based on the characteristics of the problem to the algorithm for solving the inter-city dynamic carpooling problem with single-type orders. For example, when solving this problem, you can add order priorities, so that chartered orders have the highest priority, carpooling orders have the second highest priority, and cargo orders have the lowest priority. However, this approach still makes it difficult to match orders with vehicles and optimize vehicle routes.

[0004] In view of this, this application is filed. Summary of the invention

[0005] The present invention provides an inter-city dynamic carpooling scheduling optimization method, device, equipment and system for multiple types of orders, which can at least partially improve the above-mentioned problems.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An inter-city dynamic carpooling scheduling optimization method for multiple types of orders, comprising:

[0008] Acquire the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generate an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders;

[0009] Insert all unmatched orders into the driving paths of the vehicle collection in a way that maximizes profits to form a set of solutions to be divided and conquered, and divide the set of solutions to be divided and conquered into multiple groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints;

[0010] Different types of orders in each group are processed separately, unmatched orders are removed from the vehicle's driving path, and the vehicle-passenger matching is re-performed to obtain the optimal solution for each group. The KM algorithm is used to decouple chartered orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match carpooling orders, chartered orders, and vehicles.

[0011] Each group of optimal solutions obtained is combined to generate a global feasible solution, and the global feasible solution is optimized using the adaptive large neighborhood search ALNS algorithm for passenger and freight transport to obtain a global optimal solution.

[0012] The present invention also provides an inter-city dynamic carpooling scheduling optimization device for multiple types of orders, which includes:

[0013] A scheduling unit is used to obtain multi-type order information and vehicle information at the current moment, and call a pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generate an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders;

[0014] A grouping unit, used for inserting all unmatched orders into the driving paths of the vehicle collection in a way that maximizes profits, forming a set of solutions to be divided and conquered, and dividing the set of solutions to be divided and conquered into a plurality of groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints;

[0015] Each group of optimal solution generation units is used to process different types of orders in each group respectively, remove unmatched orders from the vehicle's driving path, and re-match the vehicle and passengers to obtain the optimal solution for each group, where the KM algorithm is used to decouple chartered orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match carpooling orders, chartered orders, and vehicles;

[0016] The optimization unit is used to combine each group of optimal solutions obtained to generate a global feasible solution, and use the adaptive large neighborhood search ALNS algorithm for passenger and freight transport to optimize the global feasible solution to obtain a global optimal solution.

[0017] The present invention also provides an inter-city dynamic carpooling scheduling optimization device for multiple types of orders, including a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of the above items.

[0018] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program can be executed by a processor of a device where the computer-readable storage medium is located, so as to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of the above items.

[0019] In summary, the intercity dynamic carpooling scheduling optimization method for multiple types of orders relates to the technical field of intercity passenger transport planning. The method includes constructing a two-stage scheduling problem model based on multiple types of orders, wherein the first-stage scheduling model is divided into an online vehicle scheduling model and an offline scheduling model. Online vehicle scheduling for passenger-freight transport and offline vehicle scheduling for multiple types of orders are performed according to the multiple types of order information and vehicle information obtained at the current moment. All unmatched orders are inserted into the vehicle set in a way that maximizes profit (partial constraints may not be met) to form a set of solutions to be divided and conquered, and the set of solutions to be divided and conquered is divided into several groups, each of which contains at most a certain number of points. Different types of orders are processed in each group, wherein the KM algorithm is used to decouple charter orders from other types of orders, and the carpooling orders, charter orders and vehicles are matched by the adaptive large neighborhood search algorithm for passenger-freight transport. The optimal solutions obtained in each group are merged into a global feasible solution, and the global feasible solution is optimized by the adaptive large neighborhood search algorithm for passenger-freight transport to obtain the global optimal solution. The average total time for picking up and dropping off passengers on each route of this method is not much different from that of greedy insertion, but the total profit, average vehicle load factor, total number of carpool passengers picked up, total number of chartered vehicles, and total number of cargo picked up are all higher than those of existing carpool scheduling methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of the inter-city dynamic carpooling scheduling optimization method for multiple types of orders provided by the first embodiment of the present invention;

[0021] Figure 2 It is a pseudo code schematic diagram of an online vehicle scheduling algorithm for passenger and freight transport provided by the first embodiment of the present invention;

[0022] Figure 3 It is a pseudo code schematic diagram of the multi-type order offline vehicle scheduling algorithm provided by the first embodiment of the present invention;

[0023] Figure 4 is a flow chart of the divide-and-conquer method based on path similarity provided by the first embodiment of the present invention;

[0024] Figure 5 is a flow chart of the optimal solution calculation process for each group provided by the first embodiment of the present invention;

[0025] Figure 6 is a pseudo code schematic diagram of the ALNS algorithm for passenger and freight transport provided by the first embodiment of the present invention;

[0026] Figure 7 It is a module diagram of an inter-city dynamic carpooling scheduling optimization device for multiple types of orders provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] refer to Figure 1 As shown, the first embodiment of the present invention discloses an inter-city dynamic carpooling scheduling optimization method for multiple types of orders, which can be executed by an inter-city dynamic carpooling scheduling optimization device for multiple types of orders (hereinafter referred to as the optimization device), and in particular, executed by one or more processors in the optimization device to implement the following method:

[0029] S101, obtaining multi-type order information and vehicle information at the current moment, and calling a pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generating an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders;

[0030] Specifically, step S101 includes: the objective function formula of the offline vehicle scheduling model is:

[0031] ;

[0032] in, k For the k Vehicles, is the set of vehicles with state 1, q For the q Customer orders, is a set of orders with status 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, For vehicles Service Node moment, For vehicles Service Node moment, For vehicles Passed by A drop-off point. For vehicles Passed by A drop-off point. For vehicles The number of orders received, For vehicles The number of received goods, For vehicles Service Node moment, For vehicles The latitude and longitude of the starting point;

[0033] The objective function formula of the online vehicle scheduling model is:

[0034] ;

[0035] in, is the set of vehicles with state 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise .

[0036] The objective function formula of the two-stage scheduling model is:

[0037] ;

[0038] ;

[0039] ;

[0040] in, N To start scheduling the system Until the system ends scheduling time This period of time is discretized into the number of groups of timestamps, For The vehicles that have not departed at the time are gathered. For The set of orders that have not been matched successfully and have not timed out and those that have been matched successfully. To connect to The node set of To connect to The node set of Orders for passengers The latitude and longitude of the departure point, is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, for time The starting position point set of the vehicle, for time The collection of pickup locations for the order. is the collection of all the city's entry and exit points, for time The drop-off point set for the order, For vehicles Start serving nodes moment, For vehicles Start serving nodes moment.

[0041] Obtain the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to schedule the multi-type order information and vehicle information, and generate the online vehicle scheduling results of passenger and freight transport and the offline vehicle scheduling results of multi-type orders, specifically:

[0042] Put the data of vehicles that have already started but have not started to pick up passengers into the collection middle;

[0043] When judging the passenger order When placing an order for goods, To point The distance between the passenger and freight transport and the online vehicle dispatching is collected from the vehicles that have already departed at the current time but have not yet started to pick up passengers. Vehicles in k Sort in ascending order, where Representation Node The corresponding longitude, Representation Node The corresponding longitude, Representation Node The corresponding latitude, Representation Node The corresponding latitude, For vehicles The real-time location latitude and longitude, Representation Node The corresponding longitude, Orders for passengers The latitude and longitude of the departure point, Orders for passengers The latitude and longitude of the departure point, Representation Node The corresponding passengers, Representation Node The corresponding passengers, For vehicles Passed by A pick-up point. For vehicles Passed by A pick-up point. For vehicles Number of received goods;

[0044] Sequentially match the set of vehicles that have already departed but have not started to pick up passengers at the current time in the online vehicle scheduling of the passenger and freight transport Vehicles in and passenger orders , until the first vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found;

[0045] When judging the passenger order For carpooling orders, The real-time location latitude and longitude With passenger orders The latitude and longitude of departure point The distance of the vehicle that has been dispatched and the state of 1 is Vehicles in Sort in ascending order;

[0046] Match the vehicle set with status 1 in sequence Vehicles in and passenger orders , until a second vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found, and the online vehicle scheduling result of passenger and freight transport is obtained by combining the first vehicle path and the second vehicle path;

[0047] When judging the passenger order When placing a chartered car order, put the empty cars at the current moment into the collection Find the current empty vehicle set in the offline vehicle scheduling of the multi-type orders Medium Vehicle The latitude and longitude of the starting point With passenger orders The latitude and longitude of departure point The distance and To passenger orders The destination latitude and longitude The vehicle and passenger order with the smallest distance To match, For the city The entrance and exit For the city The entrance and exit For vehicles The city of

[0048] When judging the passenger order For other types of orders, find the vehicle path that meets the constraints and has the highest profit and insert it to obtain the offline vehicle scheduling results for multiple types of orders.

[0049] In this embodiment, the decision variables, objective function, and constraints of the offline vehicle scheduling model based on multiple types of orders are shown in Table 1:

[0050] Table 1 Offline vehicle scheduling model based on multi-type orders

[0051]

[0052] The decision variables, objective functions, and constraints of the online vehicle scheduling model based on multi-type orders are shown in Table 2:

[0053] Table 2 Online vehicle scheduling model based on multi-type orders

[0054]

[0055] The objective function of the two-stage vehicle scheduling model based on multi-type orders is:

[0056] ;

[0057] in:

[0058] ;

[0059] The decision variables of the two-stage vehicle scheduling model based on multi-type orders are shown in Table 3:

[0060] Table 3 Decision variables of the two-stage vehicle scheduling model

[0061]

[0062] The constraints of the two-stage vehicle scheduling model based on multi-type orders are shown in Tables 4 and 5:

[0063] Table 4 Constraints of the two-stage vehicle scheduling model Part 1

[0064]

[0065] Table 5 Constraints of the two-stage vehicle scheduling model Part II

[0066]

[0067] In this embodiment, the online vehicle scheduling for passenger and freight transport: matches the dispatched vehicles with the newly ordered cargo orders and carpooling orders, and inserts the start and end points of the orders into the vehicle travel path in a way that minimizes the travel time according to the online vehicle scheduling model; the offline vehicle scheduling for multiple types of orders: matches the undispatched vehicles with the newly ordered orders, and first matches the chartered orders with the nearest empty vehicles according to the offline vehicle scheduling model, and then inserts the start and end points of other types of orders into the vehicle travel path in a way that maximizes the profit.

[0068] Specifically, Figure 2 The online vehicle dispatching for passenger and freight transport is shown. The vehicles that have already departed but have not started to pick up passengers are first put into the collection. , For a cargo order, Vehicles in According to the point To point The distances are sorted in ascending order, where Representation Node The corresponding passengers, Representation Node The corresponding longitude, Representation Node The corresponding latitudes are matched in sequence Vehicles and , until a vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found; For carpooling, the dispatched vehicles will be gathered together. Vehicles in According to the point and Sort the distance in ascending order and match Vehicles and , until a vehicle path that meets the problem constraints and has the shortest pick-up and drop-off time is found.

[0069] Figure 3 This shows the offline vehicle scheduling of multiple types of orders. First, the empty vehicles at the current moment are put into the collection. In, if Find a car rental order middle and The distance and arrive The vehicle with the smallest distance Match; if For other types of orders, find the vehicle path that meets the constraints and has the highest profit for insertion, where It is the profit function for calculating the scheduling results.

[0070] S102, inserting all unmatched orders into the driving paths of the vehicle collection in a way that maximizes profits to form a set of solutions to be divided and conquered, and dividing the set of solutions to be divided and conquered into multiple groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints;

[0071] Specifically, step S102 includes: using a divide-and-conquer method based on path similarity as a divide-and-conquer method to group the set of solutions to be divided and conquered;

[0072] Each time the paths are grouped together, the paths with the smallest distance similarity are grouped together until the group contains at least Points, paths With path The distance similarity function The calculation formula is:

[0073] ;

[0074] in, for A random number that follows a uniform distribution between and Respectively represent the path and path The number of nodes in Indicates the driving path The corresponding vehicle, For the node To Node distance;

[0075] Among them, the non-compliant partial constraints include vehicle capacity constraints, chartered vehicle constraints, order hard time window constraints, maximum insertion limit time constraints and maximum time expenditure constraints.

[0076] In this embodiment, all unmatched orders are inserted into the vehicle set in a way that maximizes profits (some constraints may not be met) to form a set of solutions to be divided and conquered, and the set of solutions to be divided and conquered is divided into several groups, each of which contains at most a certain number of points. Some constraints that may not be met include: vehicle capacity constraints, charter order constraints (charter orders are considered as carpooling orders), order hard time window constraints, maximum insertion limit time constraints, and maximum time expenditure constraints. The set of solutions to be divided and conquered is divided into several groups by a divide-and-conquer method based on path similarity. Figure 4 Demonstrated a divide-and-conquer approach based on path similarity, is a list of groups, each group contains at least points, and divides the more similar paths into a group according to the path similarity. and The distance similarity function The calculation formula is:

[0077] ;

[0078] in: for A random number that follows a uniform distribution; and Respectively represent the path and The number of nodes in ; Indicates the driving path The corresponding vehicle. The smaller the value, the more similar the two paths are, and the more suitable they are to be grouped into the same group.

[0079] S103, processing different types of orders in each group respectively, removing unmatched orders from the vehicle's driving path, and re-matching the vehicle and the passengers to obtain the optimal solution for each group, wherein the KM algorithm is used to decouple chartered orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match carpooling orders, chartered orders, and vehicles;

[0080] Specifically, step S103 includes: filtering out and removing the unmatched orders in each group to obtain the currently feasible vehicle path solution. S ;

[0081] Classify the filtered unmatched orders into chartered orders and other types of orders;

[0082] Each vehicle set is required to pick up one more chartered car order each time until all chartered car orders are traversed, and the edge weights of each vehicle and each chartered car order to be picked up are input into the KM algorithm as the weight matrix to obtain the maximum weight matching and matching point pairs;

[0083] All the orders that the vehicle that picks up the chartered order originally wanted to pick up are inserted into other vehicles through the adaptive large neighborhood search ALNS algorithm for passenger and freight transport. If the insertion is incomplete, it means that the chartered order cannot be picked up at present, and the profit of all vehicles picking up the chartered order is set to zero;

[0084] Insert other single-passenger and freight intermodal transport adaptive large neighborhood search ALNS algorithm into the path of the current vehicle set to obtain the optimal solution for each group.

[0085] Preferably, the removal algorithm of the adaptive large neighborhood search ALNS algorithm for passenger and freight transport includes a random removal operator , Shaw removal operator based on cargo list , Shaw removal operator based on carpooling orders and the worst order removal operator The insertion operator of the adaptive large neighborhood search ALNS algorithm for passenger and freight transport includes a random insertion operator , optimal insertion operator , perturbation greedy insertion operator and regret-2 insertion operator .

[0086] In this embodiment, all previously unmatched orders are removed from each group to obtain the currently feasible vehicle path solution set. , and get the unmatched orders of this group, and divide them into chartered orders and other types of orders. Starting from the first chartered order, let the vehicle set pick up one more chartered order each time until all chartered orders are traversed. The edge weights of each vehicle and each chartered order to be picked up are input into the KM algorithm as the weight matrix to obtain the maximum weight matching. The matching point pair obtained is the matching relationship between the vehicle and the chartered order, where the edge weight of the vehicle and the chartered order is: the profit obtained by the vehicle when it does not pick up other orders but only picks up the chartered order. If the path does not meet the problem constraints, the profit is 0. Among them, all the orders that the vehicle that picks up and drops off the chartered order are originally inserted into other vehicles through the adaptive large neighborhood search algorithm of passenger and freight transport. If the insertion is incomplete, it means that the chartered order cannot be picked up and dropped off at present, and the profit of all vehicles picking up and dropping off the chartered order is set to zero. Finally, other orders are inserted into the path of the current vehicle set through the adaptive large neighborhood search algorithm of passenger and freight transport.

[0087] Specifically, in this embodiment, different types of orders are processed separately in each group, wherein charter orders are decoupled from other types of orders through the KM algorithm, and carpooling orders, charter orders and vehicles are matched through an adaptive large neighborhood search algorithm for passenger and freight transport. Figure 5 The calculation process of the optimal solution for each group is shown. ,Bundle Remove all previously unmatched orders and put the chartered order in In the middle, other single middle, The corresponding vehicle set , unmatched order set .calculate If you do not pick up other orders, only pick up The profit obtained when the vehicle path does not meet the constraints is 0. OK Matrix of columns The part in the dotted box is the decoupling of chartered orders from other types of orders based on the KM algorithm, where the weight matrix Input into the KM algorithm for maximum weight matching, and the KM algorithm outputs the matching point pairs. Figure 5 The symbols in mean as follows:

[0088] is a grouping result obtained by divide and conquer method; For grouping The charter list collection in; For grouping A collection of other types of orders except chartered orders; For grouping The corresponding vehicle set; For grouping The set of orders that were not successfully matched; for OK The weight matrix of the columns; For one The weight matrix of ; For grouping The set of vehicles matched to non-zero weight edges when performing the KM algorithm to decouple chartered orders from other types of orders; For grouping In the KM algorithm, when decoupling chartered orders from other types of orders, The collection of other vehicles of the vehicle; for The collection of orders that the vehicle originally picked up and dropped off.

[0089] Figure 6 The ALNS algorithm for passenger and freight transport is presented, where epn is the maximum number of iterations, ntn is the maximum number of times the solution has not been improved, upn is the number of iterations for the operator to update the weights, TP is the initial temperature of simulated annealing, TK is the temperature change rate, the reaction factor To control the speed at which the algorithm responds to changes in the scores of each operator, If 1, the choice of operator depends only on the score obtained during the iteration just ended. If it is 0, the selection of the operator depends only on the weight obtained in the previous iteration, without considering the score obtained in the iteration just ended. The operators included are shown in Table 6:

[0090] Table 6 Operators included in the adaptive large neighborhood search algorithm for passenger and freight transport

[0091]

[0092] in is a set of orders that have not been successfully matched. , represents the vehicle path set The number of orders in For path The number of points passed is the number of removal orders of the removal operator , , is a parameter that controls the number of removed orders. The order inserted into the vehicle path in the insertion operator must meet the problem constraints. The order is calculated by the order similarity function and The similarity of the order is:

[0093] ;

[0094] in is the order distance similarity weight, is the order time similarity weight; express The normalized distance between two points; Indicates order The scheduled pickup time to the order The normalized time of the scheduled boarding time. The distance and time are each normalized using the min-max method, the formula is: ,in, is the sample data, is the sample minimum value, is the maximum value of the sample.

[0095] S104, merging each group of optimal solutions obtained to generate a global feasible solution, and using the adaptive large neighborhood search ALNS algorithm for passenger and freight transport to optimize the global feasible solution to obtain a global optimal solution.

[0096] In this embodiment, the optimal solutions obtained by each group are merged into a global feasible solution, and the global feasible solution is optimized by the large neighborhood search algorithm for passenger and freight transport to obtain the global optimal solution. That is, after all the optimal solutions for the groups are obtained, the set of vehicles that have not been dispatched at the current moment and the set of orders that have not been successfully matched are input into the large neighborhood search algorithm for passenger and freight transport for optimization.

[0097] Specifically, in this embodiment, in order to verify the effectiveness of this method, the vehicle and order data of two routes of a certain AAA Technology Co., Ltd. from Anxi County to Xiamen City on June 24, 2023 and from Zhangzhou City to Xiamen City on March 31, 2023 were selected as the research objects. Among them, the number of carpooling orders from Anxi County to Xiamen City is 718, the number of chartered orders is 138, the number of cargo orders is 173, and the number of vehicles is 64; the number of carpooling orders from Zhangzhou City to Xiamen City is 3910, the number of chartered orders is 729, the number of cargo orders is 897, and the number of vehicles is 398.

[0098] The evaluation method is to schedule vehicles on a daily basis, and put each trip information into the trip set The dispatch accuracy is comprehensively evaluated by total profit, average vehicle load factor, total number of carpool passengers, total number of chartered vehicles, total number of cargo pickups, and average total time for passenger pickup and drop-off. Evaluation indicators include:

[0099] Total profit (unit: yuan): ; Average vehicle load factor: ; Total carpool passenger pickups: ;Total number of chartered vehicles: ;Total number of goods received: ; Average total time for picking up and dropping off passengers (unit: minutes): Where: Collection for itinerary The number of trips in For a particular trip; For itinerary The total order fare obtained, ;

[0100] For itinerary Total time spent on picking up and dropping off passengers; For itinerary Number of passengers carried; For itinerary The number of completed charter orders; For itinerary The experimental parameter settings are shown in Table 7, the algorithm parameter settings are shown in Table 8, and the experimental results are shown in Table 9;

[0101] Table 7 Constraint parameter settings

[0102]

[0103] Table 8 Algorithm parameter settings

[0104]

[0105] Table 9 Comparison of experimental results between the proposed method and the common greedy insertion algorithm

[0106]

[0107] Greedy insertion means that whenever there is a new order, it is inserted into the vehicle itinerary in a way that maximizes profit and meets the problem constraints. It can be seen that in the experimental results of the two routes, the average total time for picking up and dropping off passengers on each route of this method is not much different from that of greedy insertion, but the total profit, average vehicle load factor, total number of carpooling passengers, total number of chartered vehicles, and total number of goods picked up are all higher than those of the greedy insertion method, indicating that this method is superior to the current mainstream algorithm for solving the optimization of inter-city dynamic carpooling scheduling for multiple types of orders in terms of scheduling accuracy, verifying that the proposed method can better match multiple types of orders with vehicles and optimize vehicle routes.

[0108] See also Figure 7 The second embodiment of the present invention further provides an inter-city dynamic carpooling scheduling optimization device for multiple types of orders, which includes:

[0109] The scheduling unit 201 is used to obtain the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generate an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders;

[0110] The grouping unit 202 is used to insert all the unmatched orders into the driving paths of the vehicle collection in a way that maximizes the profit, to form a set of solutions to be divided and conquered, and to divide the set of solutions to be divided and conquered into a plurality of groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints;

[0111] Each group of optimal solution generating units 203 is used to process different types of orders in each group respectively, remove unmatched orders from the vehicle driving path, and re-match the vehicle and the passengers to obtain the optimal solution of each group, wherein the KM algorithm is used to decouple the chartered car orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match the carpooling orders, chartered car orders and vehicles;

[0112] The optimization unit 204 is used to combine each group of optimal solutions obtained to generate a global feasible solution, and use the adaptive large neighborhood search ALNS algorithm for passenger and freight combined transport to optimize the global feasible solution to obtain a global optimal solution.

[0113] The third embodiment of the present invention also provides an inter-city dynamic carpooling scheduling optimization device for multiple types of orders, including a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of the above items.

[0114] The fourth embodiment of the present invention also provides a computer-readable storage medium storing a computer program, which can be executed by a processor of a device where the computer-readable storage medium is located to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of the above items.

[0115] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An inter-city dynamic carpooling scheduling optimization method for multiple types of orders, characterized in that: include: Acquire the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generate an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders; Insert all unmatched orders into the driving paths of the vehicle collection in a way that maximizes profits to form a set of solutions to be divided and conquered, and divide the set of solutions to be divided and conquered into multiple groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints; Different types of orders in each group are processed separately, unmatched orders are removed from the vehicle's driving path, and the vehicle-passenger matching is re-performed to obtain the optimal solution for each group. The KM algorithm is used to decouple chartered orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match carpooling orders, chartered orders, and vehicles. Each group of optimal solutions obtained is combined to generate a global feasible solution, and the global feasible solution is optimized using the adaptive large neighborhood search ALNS algorithm for passenger and freight transport to obtain a global optimal solution; The objective function formula of the offline vehicle scheduling model is: ； in, k For the k Vehicles, is the set of vehicles with state 1, q For the q Customer orders, is a set of orders with status 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, For vehicles Service Node moment, For vehicles Service Node moment, For vehicles Passed by A drop-off point. For vehicles Passed by A drop-off point. For vehicles The number of orders received, For vehicles The number of received goods, For vehicles Service Node moment, For vehicles The latitude and longitude of the starting point; The objective function formula of the online vehicle scheduling model is: ； in, is the set of vehicles with state 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise ; The objective function formula of the two-stage scheduling model is: ； ； ； in, N To start scheduling the system Until the system ends scheduling time This period of time is discretized into the number of groups of timestamps, For The vehicles that have not departed at the time are gathered. For The set of orders that have not been matched successfully and have not timed out and those that have been matched successfully. To connect to The node set of To connect to The node set of Orders for passengers The latitude and longitude of the departure point, is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, for time The starting position point set of the vehicle, for time The collection of pickup locations for the order. is the collection of entry and exit points of all cities, for time The drop-off point set for the order, For vehicles Start serving nodes moment, For vehicles Start serving nodes moment; Obtain the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to schedule the multi-type order information and vehicle information, and generate the online vehicle scheduling results of passenger and freight transport and the offline vehicle scheduling results of multi-type orders, specifically: Put the data of vehicles that have already started but have not started to pick up passengers into the collection middle; When judging the passenger order When placing an order for goods, To point The distance between the passenger and freight transport and the online vehicle dispatching is collected from the vehicles that have already departed at the current time but have not yet started to pick up passengers. Vehicles in k Sort in ascending order, where Representation Node The corresponding longitude, Representation Node The corresponding longitude, Representation Node The corresponding latitude, Representation Node The corresponding latitude, For vehicles The real-time location latitude and longitude, Representation Node The corresponding longitude, Orders for passengers The latitude and longitude of the departure point, Orders for passengers The latitude and longitude of the departure point, Representation Node The corresponding passengers, Representation Node The corresponding passengers, For vehicles Passed by A pick-up point. For vehicles Passed by A pick-up point. For vehicles Number of received goods; Sequentially match the set of vehicles that have already departed but have not started to pick up passengers at the current time in the online vehicle scheduling of the passenger and freight transport Vehicles in and passenger orders , until the first vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found; When judging the passenger order For carpooling orders, The real-time location latitude and longitude With passenger orders The latitude and longitude of departure point The distance of the vehicle that has been dispatched and the state of 1 is Vehicles in Sort in ascending order; Match the vehicle set with status 1 in sequence Vehicles in and passenger orders , until a second vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found, and the online vehicle scheduling result of passenger and freight transport is obtained by combining the first vehicle path and the second vehicle path; When judging the passenger order When placing a chartered car order, put the empty cars at the current moment into the collection Find the current empty vehicle set in the offline vehicle scheduling of the multi-type orders Medium Vehicle The latitude and longitude of the starting point With passenger orders The latitude and longitude of departure point The distance and To passenger orders The destination latitude and longitude The vehicle and passenger order with the smallest distance To match, For the city The entrance and exit For the city The entrance and exit For vehicles The city of When judging the passenger order For other types of orders, find the vehicle path that meets the constraints and has the highest profit to insert, and obtain the offline vehicle scheduling results for multiple types of orders; The set of solutions to be divided and conquered is divided into multiple groups, specifically: A divide-and-conquer method based on path similarity is adopted as a divide-and-conquer method to group the set of solutions to be divided and conquered; Each time the paths are grouped together, the paths with the smallest distance similarity are grouped together until the group contains at least Points, paths With path The distance similarity function The calculation formula is: ； in, for A random number that follows a uniform distribution between and Respectively represent the path and path The number of nodes in Indicates the driving path The corresponding vehicle, For the node To Node distance; Among them, the non-compliant partial constraints include vehicle capacity constraints, chartered vehicle constraints, order hard time window constraints, maximum insertion limit time constraints and maximum time expenditure constraints.

2. The inter-city dynamic carpooling scheduling optimization method for multiple types of orders according to claim 1 is characterized in that: Process different types of orders in each group separately, remove unmatched orders from the vehicle's driving path, and re-match the vehicle and passengers to obtain the optimal solution for each group, specifically: Filter out and remove the unmatched orders in each group to obtain the currently feasible vehicle path solution. S ; Classify the filtered unmatched orders into chartered orders and other types of orders; Each vehicle set is required to pick up one more chartered car order each time until all chartered car orders are traversed, and the edge weights of each vehicle and each chartered car order to be picked up are input into the KM algorithm as the weight matrix to obtain the maximum weight matching and matching point pairs; All the orders that the vehicle that picks up the chartered order originally wanted to pick up are inserted into other vehicles through the adaptive large neighborhood search ALNS algorithm for passenger and freight transport. If the insertion is incomplete, it means that the chartered order cannot be picked up at present, and the profit of all vehicles picking up the chartered order is set to zero; Insert other single-passenger and freight intermodal transport adaptive large neighborhood search ALNS algorithm into the path of the current vehicle set to obtain the optimal solution for each group.

3. The inter-city dynamic carpooling scheduling optimization method for multiple types of orders according to claim 1 is characterized in that: The removal algorithm of the adaptive large neighborhood search ALNS algorithm for passenger and freight transport includes a random removal operator , Shaw removal operator based on cargo list , Shaw removal operator based on carpooling orders and the worst order removal operator The insertion operator of the adaptive large neighborhood search ALNS algorithm for passenger and freight transport includes a random insertion operator , optimal insertion operator , perturbation greedy insertion operator and regret-2 insertion operator .

4. An inter-city dynamic carpooling scheduling optimization device for multiple types of orders, characterized in that: include: A scheduling unit is used to obtain multi-type order information and vehicle information at the current moment, and call a pre-trained two-stage scheduling problem model to perform scheduling processing on the multi-type order information and vehicle information, and generate an online vehicle scheduling result for passenger and freight transport and an offline vehicle scheduling result for multi-type orders, wherein the two-stage scheduling problem model includes a one-stage scheduling model and a two-stage scheduling model, the one-stage scheduling model includes an offline vehicle scheduling model and an online vehicle scheduling model, and the online vehicle scheduling result for passenger and freight transport and the offline vehicle scheduling result for multi-type orders are vehicle path results for successfully matched orders; A grouping unit, used for inserting all unmatched orders into the driving paths of the vehicle collection in a way that maximizes profits, forming a set of solutions to be divided and conquered, and dividing the set of solutions to be divided and conquered into a plurality of groups, wherein each group contains at least a preset number of points, and the driving paths of the vehicle collection contain paths that do not meet some constraints; Each group of optimal solution generation units is used to process different types of orders in each group respectively, remove unmatched orders from the vehicle's driving path, and re-match the vehicle and passengers to obtain the optimal solution for each group, where the KM algorithm is used to decouple chartered orders and other types of orders, and the adaptive large neighborhood search ALNS algorithm for passenger and freight transport is used to match carpooling orders, chartered orders, and vehicles; An optimization unit, used for merging each group of optimal solutions obtained to generate a global feasible solution, and optimizing the global feasible solution using the adaptive large neighborhood search ALNS algorithm for passenger and freight transport to obtain a global optimal solution; The objective function formula of the offline vehicle scheduling model is: ； in, k For the k Vehicles, is the set of vehicles with state 1, q For the q Customer orders, is a set of orders with status 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, For vehicles Service Node moment, For vehicles Service Node moment, For vehicles Passed by A drop-off point. For vehicles Passed by A drop-off point. For vehicles The number of orders received, For vehicles The number of received goods, For vehicles Service Node moment, For vehicles The latitude and longitude of the starting point; The objective function formula of the online vehicle scheduling model is: ； in, is the set of vehicles with state 0, is a 0-1 decision variable, vehicle Pickup and delivery orders but ,otherwise ; The objective function formula of the two-stage scheduling model is: ； ； ； in, N To start scheduling the system Until the system ends scheduling time This period of time is discretized into the number of groups of timestamps, For The vehicles that have not departed at the time are gathered. For The set of orders that have not been matched successfully and have not timed out and those that have been matched successfully. To connect to The node set of To connect to The node set of Orders for passengers The latitude and longitude of the departure point, is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , is a 0-1 decision variable. If the vehicle Through the arc but ,otherwise , For orders The order price, C is the fixed cost of using each vehicle, SC is the cost per kilometer traveled by the vehicle, sp is the speed of the vehicle, the vehicle travels at a constant speed, for time The starting position point set of the vehicle, for time The collection of pickup locations for the order. is the collection of entry and exit points of all cities, for time The drop-off point set for the order, For vehicles Start serving nodes moment, For vehicles Start serving nodes moment; Obtain the multi-type order information and vehicle information at the current moment, and call the pre-trained two-stage scheduling problem model to schedule the multi-type order information and vehicle information, and generate the online vehicle scheduling results of passenger and freight transport and the offline vehicle scheduling results of multi-type orders, specifically: Put the data of vehicles that have already started but have not started to pick up passengers into the collection middle; When judging the passenger order When placing an order for goods, To point The distance between the passenger and freight transport and the online vehicle dispatching is collected from the vehicles that have already departed at the current time but have not yet started to pick up passengers. Vehicles in k Sort in ascending order, where Representation Node The corresponding longitude, Representation Node The corresponding longitude, Representation Node The corresponding latitude, Representation Node The corresponding latitude, For vehicles The real-time location latitude and longitude, Representation Node The corresponding longitude, Orders for passengers The latitude and longitude of the departure point, Orders for passengers The latitude and longitude of the departure point, Representation Node The corresponding passengers, Representation Node The corresponding passengers, For vehicles Passed by A pick-up point. For vehicles Passed by A pick-up point. For vehicles Number of received goods; Sequentially match the set of vehicles that have already departed but have not started to pick up passengers at the current time in the online vehicle scheduling of the passenger and freight transport Vehicles in and passenger orders , until the first vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found; When judging the passenger order For carpooling orders, The real-time location latitude and longitude With passenger orders The latitude and longitude of departure point The distance of the vehicle that has been dispatched and the state of 1 is Vehicles in Sort in ascending order; Match the vehicle set with status 1 in sequence Vehicles in and passenger orders , until a second vehicle path that meets the problem constraints and has the shortest passenger pick-up and drop-off time is found, and the online vehicle scheduling result of passenger and freight transport is obtained by combining the first vehicle path and the second vehicle path; When judging the passenger order When placing a chartered car order, put the empty cars at the current moment into the collection Find the current empty vehicle set in the offline vehicle scheduling of the multi-type orders Medium Vehicle The latitude and longitude of the starting point With passenger orders The latitude and longitude of departure point The distance and To passenger orders The destination latitude and longitude The vehicle and passenger order with the smallest distance To match, For the city The entrance and exit For the city The entrance and exit For vehicles The city of When judging the passenger order For other types of orders, find the vehicle path that meets the constraints and has the highest profit to insert, and obtain the offline vehicle scheduling results for multiple types of orders; The set of solutions to be divided and conquered is divided into multiple groups, specifically: A divide-and-conquer method based on path similarity is adopted as a divide-and-conquer method to group the set of solutions to be divided and conquered; Each time the paths are grouped together, the paths with the smallest distance similarity are grouped together until the group contains at least Points, paths With path The distance similarity function The calculation formula is: ； in, for A random number that follows a uniform distribution between and Respectively represent the path and path The number of nodes in Indicates the driving path The corresponding vehicle, For the node To Node distance; Among them, the non-compliant partial constraints include vehicle capacity constraints, chartered vehicle constraints, order hard time window constraints, maximum insertion limit time constraints and maximum time expenditure constraints.

5. An inter-city dynamic carpooling scheduling optimization device for multiple types of orders, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: A computer program is stored, and the computer program can be executed by a processor of the device where the computer-readable storage medium is located to implement the inter-city dynamic carpooling scheduling optimization method for multiple types of orders as described in any one of claims 1 to 3.

Images