Passenger transport service path optimization method and system based on modular electric vehicle

Abstract

The invention provides a passenger transport service path optimization method and system based on a modular electric vehicle, and belongs to the technical field of urban traffic road operation management. Extracting expected arrival time of the passengers; with the fastest delivery efficiency and the minimum cost as targets, distribution balance adjustment is carried out on vehicle service points by emphatically considering an adaptive charging strategy and cooperative arrival constraints among lines, and decision modeling is carried out on operation lines, decompilation and coupling points, module row length and charging duration of modules; and based on an end-to-end space-time similar demand point mining algorithm and a customized adaptive large neighborhood search algorithm, solving the passenger transport service path optimization model in the crowded area, and realizing a large-scale passenger transport system service line optimization scheme. According to the invention, service route planning, vehicle configuration and charging strategy optimization can be efficiently carried out in real time, the service efficiency and the vehicle utilization rate of the passenger transport system are improved, the riding experience of passengers is improved, and unnecessary electric energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of urban traffic road operation management, and in particular to a passenger service path optimization method and system based on modular electric vehicles. Background Art

[0002] In crowded areas like train stations, business districts, or large event venues, such as during rush hour, large numbers of passengers typically disperse from office areas to surrounding and suburban residential areas. Traditional transportation faces challenges in coping with such a large travel demand. The influx of online ride-hailing services will significantly increase local traffic pressure, leading to road congestion and reduced transportation efficiency. Public transportation operates fixed stations and routes, which cannot fully cover passengers' starting and ending points. For short trips between the beginning and end, passengers often need to rely on walking or shared bicycles, which is less convenient and increases travel time. In addition, public transportation in suburban areas generally has low occupancy rates, resulting in wasted resources and inefficient transportation capacity. Existing transportation methods cannot effectively meet the high-density and high-volume travel needs while providing refined door-to-door service.

[0003] Against this backdrop, modular vehicles, boasting large passenger capacities and door-to-door flexibility, have emerged. These vehicles consist of multiple identical electric modules, each of which can operate independently along its route like a car or connect to form a large bus-like platoon. This flexible decoupling allows for rapid adjustments in module platoon size and passenger capacity along the route, making them ideal for passenger transportation in densely populated areas, where modular vehicles have yet to be developed. Furthermore, optimizing passenger transportation routes in densely populated areas requires optimizing transportation resources under the pressure of large-scale, high-density, and time-sensitive traffic scheduling, making this a challenging task.

[0004] The application of the emerging technology of modular vehicles in passenger transportation services currently focuses primarily on changing the number of modules at the origin and destination stations, or appropriately decoupling vehicle modules from established routes to serve a small number of demand points outside the route. Research is lacking on dynamically adjusting module configuration within fully flexible route services based on spatiotemporal demand. Furthermore, existing solutions for solving large-scale instances first cluster demand points and then make decisions. This approach typically relies on forward-looking features such as distance, demand volume, and passenger arrival time, and fails to consider the improvement of backward-looking decision optimization objectives during clustering. Even if high-performance computing equipment is used or a significant amount of time is spent designing precise algorithms to obtain the optimal solution to the subsequent decision optimization problem, the initial clustering results can still lead to performance bottlenecks in the final decision objective.

[0005] In summary, existing technical solutions for passenger service route planning have several shortcomings in practice. First, existing technologies lack the ability to provide dynamic and flexible route services. Currently, module adjustments are primarily tailored to fixed starting and ending stations or specific route requirements. There is a lack of solutions that can dynamically adjust vehicle module configurations based on actual demand under fully flexible spatiotemporal demand conditions. This limitation prevents existing technologies from effectively addressing complex and spatiotemporally varying passenger demand, limiting the potential application of modular vehicle technology in more diverse scenarios. Furthermore, when solving large-scale examples, the clustering of demand points relies on forward-looking features such as demand volume and distance, without considering the optimization objectives of backward-looking decisions. This approach may result in clustering results that cannot fully support subsequent decision-making, thereby affecting the ultimate service quality and performance. Even with the use of high-performance computing equipment or the design of complex algorithms for solution, the initial clustering method may still create a bottleneck in the optimization effect, resulting in limited overall decision-making performance. Summary of the Invention

[0006] The object of the present invention is to provide a method and system for optimizing passenger service routes based on modular electric vehicles to solve at least one technical problem existing in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a method for optimizing passenger service routes based on modular electric vehicles, comprising:

[0009] Based on online ride-hailing order information and road network information, the passenger's pick-up and drop-off points are matched to the nearest intersection, and the passenger's expected arrival time is extracted;

[0010] Based on the flexible decoupling characteristics of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs, this approach balances the distribution of vehicle service points by focusing on adaptive charging strategies and coordinated arrival constraints between routes. This modeling of module routes, decoupling points, module queue lengths, and charging times is then used to establish a passenger service route optimization model for densely populated areas.

[0011] Based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model in crowded areas and realize the service route optimization plan of large-scale passenger transport systems.

[0012] As a further limitation of the first aspect of the present invention, the end-to-end spatiotemporal similar demand point mining algorithm uses a two-layer graph sampling and aggregation model to embed the demand point features of the travel network graph, and inputs the embedded input into the improved differentiable density peak clustering algorithm to obtain the cluster category distribution. In each category, a customized adaptive large neighborhood search algorithm is applied to make decisions and solve the established passenger service path optimization model in densely populated areas to obtain vehicle operation routes, decoupled satellite and coupled satellite positions, charging strategies, module platoon leader configurations and objective function values.

[0013] As a further limitation of the first aspect of the present invention, the objective function values ​​of each category are summed to obtain the total objective function value as the loss function of model training, gradient backpropagation is iteratively trained, gradually learning in the direction of reducing the optimization target, adjusting the clustering results, re-optimizing the decision, repeating this process until the set termination condition is reached, and outputting the result.

[0014] As a further limitation of the first aspect of the present invention, large-scale demand points are clustered and divided into multiple areas based on spatiotemporal similarities, and then modular vehicles depart from the source point in the form of a large vehicle platoon to quickly transport passengers to the decoupling satellites of each area. At the same time, the large vehicle platoon is decoupled into multiple small vehicle modules according to the passenger demand of each area and the platoon scale effect. Each small vehicle module undertakes different passenger transportation tasks and delivers the passengers to the destination according to the passengers' desired time. Passengers do not need to get off or transfer during the ride, and door-to-door refined delivery services are completed through different vehicle modules. After the passengers are delivered, the empty small vehicle modules are converged and coupled again to form a large vehicle platoon to return to the source point for the next service.

[0015] As a further limitation of the first aspect of the present invention, the demand points are initially clustered considering three spatiotemporal characteristics: geographical distance, expected delivery time window, and number of passenger demands; in each area, a customized adaptive large neighborhood search algorithm is used to solve the driving route, vehicle platoon leader configuration, and charging strategy of the decision-making modular vehicles; the loss function related to the decision objective is designed as a function of the decision objective, and backpropagation is performed to adjust the clustering results in the direction of reducing the decision objective function, thereby forming iterative training of the clustering decision-making integrated algorithm.

[0016] As a further limitation of the first aspect of the present invention, the improved differentiable density peak clustering algorithm converts the binary variable P in the traditional density peak clustering algorithm into ik , the degree to which point i is assigned to category k is improved to a probability distribution. This is because during neural network training, backpropagation requires the gradient of the loss function with respect to the model parameters to be differentiable, while the binary quantity of hard clustering is not differentiable and is not suitable for direct use in deep learning models. Therefore, the traditional density peak clustering is improved to soft clustering using local density and distance, generating a probability distribution for each point belonging to each cluster.

[0017] In a second aspect, the present invention provides a passenger service route optimization system based on modular electric vehicles, comprising:

[0018] The extraction module is used to match passenger pick-up and drop-off points to the nearest intersection based on online ride-hailing order information and road network information, and extract the passenger's expected arrival time;

[0019] A construction module is used to leverage the flexible decoupling capabilities of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs. By focusing on adaptive charging strategies and coordinated arrival constraints between routes, the module's operating routes, decoupling points, module queue lengths, and charging times are modeled to optimize passenger service routes in densely populated areas.

[0020] The solution module is used to solve the passenger service path optimization model in crowded areas based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm. It integrates the two through gradient backpropagation to realize the service route optimization plan of large-scale passenger transport systems.

[0021] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the passenger service path optimization method based on modular electric vehicles as described in the first aspect is implemented.

[0022] In a fourth aspect, the present invention provides a computer device comprising a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the passenger service path optimization method based on modular electric vehicles as described in the first aspect.

[0023] In a fifth aspect, the present invention provides an electronic device comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to execute instructions for implementing the passenger service path optimization method based on modular electric vehicles as described in the first aspect.

[0024] The beneficial effects of the present invention are: it can carry out service route planning, vehicle configuration and charging strategy optimization in real time and efficiently, improve the service efficiency and vehicle utilization rate of the passenger transport system, enhance the passenger riding experience, reduce unnecessary power and energy consumption, and promote the technological development and sustainable development of smart city transportation.

[0025] Additional advantages of the present invention will be more clearly given in the following description or learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 This is a schematic diagram of the service line optimization process described in an embodiment of the invention.

[0028] Figure 2 This is a flowchart of the modular vehicle passenger service according to an embodiment of the present invention.

[0029] Figure 3 This is a flow chart of the clustering decision integration algorithm described in an embodiment of the present invention.

[0030] Figure 4 This is a box diagram of the occupancy rate of the modular vehicle and the fixed-capacity shuttle bus according to the embodiment of the present invention.

[0031] in, Figure 4 (a) is a modular vehicle, Figure 4 (b) is a shuttle bus. DETAILED DESCRIPTION

[0032] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention.

[0033] Those skilled in the art will understand that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0034] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.

[0035] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0036] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless otherwise contradictory.

[0037] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0038] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.

[0039] This invention provides a method for optimizing passenger service routes based on modular electric vehicles. The method includes: 1) data processing and statistical analysis: Based on online ride-hailing order information and road network information, passenger pick-up and drop-off points are matched to the nearest intersection and the passenger's expected arrival time is extracted. 2) Leveraging the flexible decoupling capabilities of modular vehicles, the method balances the distribution of vehicle service points with the goal of maximizing delivery efficiency and minimizing costs by focusing on adaptive charging strategies and coordinated arrival constraints between routes. Decision modeling is then performed for the module's operating routes, decoupling points, module queue lengths, and charging durations. 3) An integrated clustering decision-making solution algorithm is designed. An end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm are proposed, respectively. These two algorithms are integrated through gradient backpropagation to solve service routes for large-scale passenger transport systems. The proposed method for optimizing passenger service routes based on modular electric vehicles can efficiently and in real time optimize service routes, vehicle configuration, and charging strategies, thereby improving the service efficiency and vehicle utilization of the passenger transport system, enhancing the passenger experience, reducing unnecessary energy consumption, and promoting the technological and sustainable development of smart city transportation.

[0040] Example 1

[0041] In this embodiment 1, a passenger service path optimization system based on modular electric vehicles is first provided, including: an extraction module, which is used to match passenger boarding and alighting points to the nearest intersection based on online car-hailing order information and road network information, and extract the passenger's expected delivery time; a construction module, which is used to take the characteristics of flexible decoupling of modular vehicles, with the fastest delivery efficiency and minimum cost as the goal, by focusing on adaptive charging strategies and coordinated arrival constraints between routes to balance the distribution of vehicle service points, make decision-making models for the module's operating routes, decoupling points, module queue lengths and charging times, and establish a passenger service path optimization model for crowded areas; a solution module, which is used to solve the passenger service path optimization model for crowded areas based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, and integrate the two through gradient backpropagation to achieve a large-scale passenger system service route optimization solution.

[0042] In this embodiment, the above-mentioned system is used to implement a passenger service path optimization method based on modular electric vehicles, including: matching passenger boarding and alighting points to the nearest intersection based on online car-hailing order information and road network information, and extracting the passenger's expected delivery time; based on the flexible decoupling characteristics of modular vehicles, with the fastest delivery efficiency and minimum cost as the goal, by focusing on adaptive charging strategies and coordinated arrival constraints between routes to balance the distribution of vehicle service points, decision modeling is performed on the module's operating route, decoupling point, module queue length and charging time, and a passenger service path optimization model for crowded areas is established; based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model for crowded areas, and realize a large-scale passenger system service route optimization solution.

[0043] Among them, the end-to-end spatiotemporal similarity demand point mining algorithm uses a two-layer graph sampling and aggregation model to embed the demand point features of the travel network graph, and inputs the embedded input into the improved differentiable density peak clustering algorithm to obtain the cluster category distribution. In each category, a customized adaptive large neighborhood search algorithm is applied to make decisions and solve the established passenger service path optimization model in densely populated areas, obtaining vehicle operation routes, decoupling satellite and coupling satellite positions, charging strategies, module platoon leader configurations and objective function values.

[0044] The objective function values ​​of each category are summed up to obtain the total objective function value as the loss function of model training. Gradient backpropagation is used for iterative training to gradually learn in the direction of reducing the optimization target, adjust the clustering results, re-optimize the decision, repeat this process until the set termination condition is reached, and output the result.

[0045] Large-scale demand points are clustered and divided into multiple areas based on spatiotemporal similarities. Modular vehicles then depart from the source in the form of a large platoon, quickly transporting passengers to the decoupling satellites in each area. Simultaneously, the large platoon is decoupled into multiple small vehicle modules based on the passenger demand in each area and the platoon size effect. Each small vehicle module undertakes different passenger transportation tasks and delivers passengers to their destinations at the desired time. Passengers do not need to get off or transfer during the ride, and door-to-door refined delivery services are provided through different vehicle modules. After the passengers are delivered, the empty small vehicle modules converge and couple again to form a large platoon, returning to the source for the next service.

[0046] Demand points are initially clustered considering three spatiotemporal characteristics: geographical distance, expected delivery time window, and number of passenger demands. In each area, a customized adaptive large neighborhood search algorithm is used to solve the driving routes, vehicle platoon configuration, and charging strategy of the decision-making modular vehicles. The loss function related to the decision-making objective is designed as a function of the decision-making objective, and backpropagation is used to adjust the clustering results in the direction of reducing the decision objective function, forming an iterative training of the clustering decision-making integrated algorithm.

[0047] The improved differentiable density peak clustering algorithm transforms the traditional density peak clustering algorithm into the degree P of assigning sample point i to category k. ik , this binary variable is replaced with a probability distribution. This is because neural network training relies on the backpropagation mechanism, which requires the loss function to be differentiable with respect to the model parameters. However, the binary variables used in traditional hard clustering are inherently non-differentiable and therefore unsuitable for direct application in deep learning frameworks. To address this issue, the algorithm extends the original density peak clustering method into a soft clustering method based on the local density of the sample and the distance to other points. This provides each sample point with a probability distribution of its belonging to each cluster center.

[0048] Example 2

[0049] The present invention provides a method for optimizing passenger service routes based on modular electric vehicles, the method comprising: 1) data preparation and statistical analysis to obtain online car-hailing order information, match passenger boarding and alighting points to the nearest intersection, and extract the passenger's desired location and expected service time; 2) based on modular vehicles that can be flexibly decoupled, establish a passenger service route optimization model for densely populated areas, with the fastest delivery efficiency and minimum cost as the goal, by focusing on adaptive charging strategies and coordinated arrival constraints between routes to balance the distribution of vehicle service points, and make decisions on the module's decoupling points, module queue lengths, and charging times. 3) propose an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, and integrate the two through gradient backpropagation, gradually adjust in the direction of improving the optimization target, and achieve a better decision-making solution.

[0050] Establishment of modular vehicle route optimization model: Based on modular vehicles with flexible decoupling, a route optimization solution is proposed for densely populated areas to meet both large passenger flow transportation and door-to-door refined service requirements. Passenger transportation in densely populated areas usually involves a large number of passengers departing from a single source and traveling to multiple destinations. The modular vehicle passenger service process is divided into three stages for decision modeling, such as Figure 2 First, the large-scale demand points are clustered and divided into multiple regions based on spatiotemporal similarity, as shown in Figure 2 As shown in (a), the modular vehicles then depart from the source in the form of a large vehicle platoon and quickly transport the passengers to the decoupling satellites in each region (first phase), as shown in Figure 2 As shown in (b), the large vehicle platoon is decoupled into multiple small vehicle modules according to the passenger demand in each area and the platoon size effect. Each small vehicle module undertakes different passenger transportation tasks and delivers passengers to their destination according to the passengers' expected time, as shown in Figure 2 (c) As shown. Passengers do not need to get off or transfer during the ride, and can complete the door-to-door refined delivery service through different vehicle modules (second stage). After the passengers are delivered, the empty small vehicle modules are re-converged and coupled to form a large vehicle platoon to return to the source for the next service (third stage), as shown in Figure 2 (d) shown.

[0051] Design of an integrated clustering and decision-making solution algorithm: An end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm are proposed and integrated. The loss function of the integrated clustering and decision-making algorithm is designed to be a function of the decision objective. Iterative training is performed through a gradient backpropagation mechanism. This ensures that clustering is not just a static partitioning of passenger demand data, but rather a gradual learning process toward reducing the optimization objective, automatically adjusting clustering results, and dynamically responding to changes in the optimization objective. The clustering process is tightly integrated with the global optimization objective. The clustering results not only consider the internal structure of spatiotemporal travel data but also significantly optimize the decision objective function, thereby finding more advantageous solutions to complex travel optimization problems.

[0052] The specific operations are:

[0053] Step S1: Data Processing and Statistical Analysis. Prepare online ride-hailing order data, which primarily includes order number, origin and destination locations (latitude and longitude), boarding and alighting times, and number of passengers. Cleanse travel time and distance data according to the 3σ principle. Extract the locations of passenger demand points and map them to the nearest intersections. Statistically analyze the expected service time of passengers.

[0054] Step S2: Establish a modular vehicle route optimization model. The modular vehicle passenger service process is modeled in three stages. The basic assumptions of the model are: (1) The passenger demand at each demand point is met at one time and can only be visited once by an electric modular vehicle. (2) The vehicle has a maximum allowable module platoon length limit. The maximum vehicle power that can be achieved with the maximum allowable platoon length is sufficient to support the power consumption between any two points in each area. The platoon scale effect will save energy and other resource consumption, so the power consumption of large vehicles in the first and second stages is ignored. (3) Each passenger demand point can be charged. The independent operability of modular vehicles enables them to provide door-to-door personalized services and pick up passengers in each community. There are charging stations near each community. The key problem in the first stage of route planning is to determine the optimal route for a group of modular vehicles with an indefinite number of platoon lengths to deliver and disassemble the modules required for each area to the decoupling satellite. This route generation problem can be regarded as a variant of the open capacity-constrained vehicle routing problem with time windows. The number of vehicle modules required for each category, the location of the decoupling satellite, and the time window are determined in the second stage. To ensure service timeliness, the minimum expected arrival time for all passengers in each category is taken as the expected service time for that decoupled satellite. By setting the return distance to zero, the modular vehicle can be directly disassembled and transported to the satellite without the need for a towed vehicle to return.

[0055] The second phase of route planning uses the decoupled satellites in each region as the starting point, using multiple disassembled small vehicle modules to provide simultaneous service to passengers in the region on multiple routes, ultimately converging at the marshaled satellite. Key decision points include the small vehicle module service routes, the selection of decoupled and marshaled satellite locations based on time coordination and route allocation, the vehicle module platoon length plan based on passenger demand and queue size effects, and the charging decision based on time coordination and travel distance, including whether the vehicle module should be charged at the customer point and the charging duration. The second phase problem modeling is as follows:

[0056]

[0057]

[0058] The objective function (1) aims to minimize the travel distance cost ∑ i,j,k,p d ij W p x ijk z kp , fleet size cost∑ k, p pU p z kp , collaborative time difference penalty Penalty for violating passenger expected service time window∑ i,k y ik w ik , Maximum charging time and total service time The weighted sum of constraints (2) and (3) restricts each area to have only one decoupling satellite and one cooperating satellite. Constraints (4-6) restrict that if vehicle k is used, it must start from the decoupling satellite and eventually reach the cooperating satellite. Constraint (7) indicates that each passenger demand point must be visited. Constraint (8) indicates that each passenger demand point can only be visited once. Constraint (9) indicates that the vehicle module must have both entry and exit at the demand point. Constraint (10) indicates that the vehicle module cannot drive directly from the decoupling satellite to the cooperating satellite. Constraint (11) is the passenger capacity constraint of the small vehicle module. Constraint (12) indicates that a small vehicle module can only have one module platoon leader selected. Constraints (13-18) are power-related constraints. Constraint (13) indicates that each small vehicle module is fully charged when it departs from the decoupling satellite. Constraints (14) and (15) are power consumption constraints for path i to j. Constraints (16) and (17) are charging constraints for point j. The power of leaving point j must not be less than the power of arriving at point j. The power is greater than 0 and less than full. Constraint (18) extracts the maximum value of the total charging time of each small vehicle module during the entire service journey. Constraints (19) and (20) are the travel time constraints for path i to j, and constraint (21) requires that the vehicle must depart from the decoupling satellite at time 0. Constraint (22) is the violation of the service time window constraint. Constraints (23) and (24) are time coordination constraints, which respectively extract the maximum and minimum values ​​of the time for each small vehicle module to arrive at the coupling satellite. The remaining constraints (25-29) define the variables. The parameters and variable symbols used in the model are shown in Table 1.

[0059] Table 1 Parameters and variable symbols

[0060]

[0061]

[0062] The third phase aims to determine an optimal set of paths, sequentially visiting the marshaling satellites, marshaling the small vehicle modules at the satellites into the vehicle platoon, and returning to the source point for the next service. This process can be understood as the reverse of the first phase, except that the point set is no longer the set of decoupling satellites, but the set of marshaling satellites. Furthermore, the third phase requires that the vehicle service each marshaling satellite be later than the time it takes for all small vehicle modules in each area to converge at the marshaling satellite. Small vehicle modules depart directly from the marshaling satellites without the need for a tow vehicle. This is achieved by setting the distance from the source point to any marshaling satellite to zero.

[0063] Step S3: Propose an integrated clustering decision-making algorithm. This algorithm integrates the end-to-end spatiotemporal similarity demand point mining with a customized adaptive large neighborhood search algorithm through gradient backpropagation. Figure 3The algorithm's workflow is outlined. First, a data-driven approach is used to initially cluster demand points, taking into account spatiotemporal characteristics such as geographic distance, expected arrival time window, and passenger demand. Then, within each region, a customized adaptive large neighborhood search algorithm is used to determine the driving routes, platoon configuration, and charging strategy for the decision-making modular vehicles. Finally, a loss function is designed that is correlated with the decision objective. Backpropagation is then used to adjust the clustering results in a way that minimizes the decision objective function. This results in iterative training of an integrated clustering and decision-making algorithm to further enhance the quality of the solution.

[0064] The end-to-end spatiotemporal similarity demand point mining algorithm first embeds demand point features using a data-driven graph sampling and aggregation method. The demand point representation is then updated through message passing between demand points. By introducing a sampling and aggregation strategy, the graph sampling and aggregation model only samples a fixed number of neighbors for each node, rather than aggregating all neighbors. This effectively improves the model's scalability and computational efficiency on large graphs. The output continuous embedding space is then fed into a differentiable layer of improved density peak clustering to generate cluster centers and generate soft assignments for the demand points.

[0065] A customized adaptive large neighborhood search algorithm is designed for application within each group to determine vehicle service routes, module number configuration, and charging strategy. This algorithm, based on the large-area search algorithm framework, embeds vehicle platoon leader configuration and charging strategy optimization within the routes generated in each iteration. Two key aspects of the customized adaptive large neighborhood search algorithm are initial solution generation and the design of removal and insertion operators. A reasonable initial solution is crucial for algorithm search efficiency and result quality. The removal and insertion operators flexibly adjust the solution structure, helping the algorithm avoid local optima and accelerate convergence to the global optimal solution. This embodiment selects coupling / decoupling satellites based on a probabilistic selection mechanism and generates initial routes using a greedy strategy. The removal and insertion operators are designed to address charging and time coordination. The platooning nature of vehicle modules can reduce vehicle operating costs related to driving distance and fleet size. Based on the feasible routes generated in each iteration, this embodiment uses numerical analysis to find that vehicle operating costs are a checksum function of the vehicle platoon leader configuration. The checksum function properties are then used to determine the vehicle platoon leader configuration. The charging strategy optimization of the line is embedded as a subroutine in the customized adaptive large neighborhood search algorithm to balance the distribution of vehicle charging time and coordinate the time difference of each module arriving at the coupling satellite.

[0066] The clustering decision integration algorithm constructs a travel network diagram based on residents' spatiotemporal travel order data and road network structure. The forward propagation process uses a two-layer graph sampling and aggregation model to embed the demand point features of the travel network diagram, and then inputs the embedded data into the improved differentiable density peak clustering algorithm to obtain the soft clustering results. A customized adaptive large-scale search algorithm is applied to each category to determine the second-stage model's decision-making and solution, obtaining the decoupling and marshaling satellite positions, charging strategies, module platoon leader configurations, and objective function values. Backpropagation sums the objective function values ​​for each category, using the resulting total second-stage objective function value as the loss function. Gradient backpropagation is used for iterative training, gradually learning to reduce the optimization target and adjusting the clustering results. Finally, based on the determined decoupling and marshaling satellite positions, the paths and module decoupling strategies for the first and third stages are determined.

[0067] In summary, the path optimization method for a modular electric vehicle passenger service system for areas with dense passenger flow proposed in this embodiment aims to solve the problems of modular electric vehicle service route planning and charging strategy. By analyzing online car-hailing travel data, extracting passengers' demand points and expected service times, establishing a route optimization model based on time collaboration, and proposing a solution algorithm that integrates clustering and optimization decision-making based on gradient backpropagation, the path optimization of modular vehicles, the selection of vehicle platoon leaders and decoupling points, and the optimization of charging strategies are achieved. This provides solutions for passenger services in areas with dense passenger flow, improves transportation efficiency, and provides personalized door-to-door services.

[0068] Example 3

[0069] like Figure 1 As shown, this embodiment provides a service route optimization method based on modular vehicles, the method comprising:

[0070] Step S1: data processing and statistical analysis. In this embodiment, taking a certain city in Beijing as an example, the following data need to be prepared: the online car-hailing order data information within the Fifth Ring Road of a certain city in Beijing in June and July 2018, the road network information within the Fifth Ring Road of a certain city in Beijing, and the intersection data information. The travel time and travel distance are cleaned according to the 3σ principle. The data mainly includes the order number, the location of the origin and destination (latitude and longitude), the boarding and alighting time, the number of passengers, etc. The present invention matches the passenger boarding and alighting locations to the nearest intersection, aggregates the passengers with the same origin and destination, and takes the average of all trip delivery times in each origin-destination pair as the expected service time for the passenger. Data analysis shows that large-scale events will trigger large-scale passenger flows. For example, after the concert held at the Workers' Stadium in a certain city in Beijing ended at 21:30 on July 14, 2018, there were 195 online ride-hailing orders departing from the venue at the same time, with a passenger demand of 470 people. A large number of people and vehicles gathered at the exit of the venue at the same time, and the road occupancy rate was low, which easily caused congestion and put pressure on the transportation system.

[0071] Step S2: Establish a modular vehicle routing optimization model. By constructing auxiliary variables and linearizing the established model using the Big M method, nonlinear equality constraints are converted into equivalent linear inequality constraints, enabling the model to be solved using standard linear programming methods. This linearization process not only improves computational efficiency but also ensures the accuracy and feasibility of the model solution. The final model is represented as follows:

[0072]

[0073] st(2)-(3)

[0074]

[0075] (7)-(12)

[0076]

[0077]

[0078] (18)-(22)

[0079]

[0080] (25)-(29)

[0081]

[0082] Constraints (31-35) are used to construct auxiliary variables g k Constraints (4-6) are linearized. Constraints (36-42) are obtained by linearizing constraints (13-17) using the Big M method. Constraints (43-46) are obtained by linearizing constraints (23-24). Objective function (1) is linearized into the new objective function (30) and constraints (47-51). Constraints (52-54) are variable definitions.

[0083] Step S3: Propose a clustering decision-making integrated solution algorithm. A clustering decision-making integrated algorithm is proposed, which uses the optimization decision objective to represent the loss function. Iterative training is performed using a gradient backpropagation mechanism, allowing clustering to dynamically adjust toward the optimization decision objective, significantly improving decision results and optimizing the global objective. Table 2 uses numerical examples to compare the optimization performance of the traditional clustering-first-then-decision method and the clustering decision-making integrated method proposed in this embodiment at different case sizes. As can be seen from the table, when the case size is 50, the target value for direct decision-making without clustering is 2203. First, using the GraphSAGE-DPC algorithm to cluster the region division results and then solving the decision for each sub-region, the target value is 1591. The clustering decision-making integrated solution, after iterative training, achieves a target value of 1342, which is 39.1% lower than the non-clustering and clustering-first-then-decision methods and 15.6% lower than the non-clustering and clustering-first-then-decision methods. First, using the GTN-DPC algorithm to cluster the region division results and then solving the decision for each sub-region, the target value is 1926. The clustering decision-making integrated solution, after iterative training, achieves a target value of 1565. When the case size is 150, the clustering-decision integration algorithm solves the target value of 6100, which is 26.4% and 14.1% lower than the 8285 and 7100 of no clustering and clustering-first-then-decision respectively.

[0084] Table 2 Comparison of clustering decision integration, direct decision without clustering, and clustering first and then decision-making algorithms

[0085]

[0086] In a real-world case study, various large-scale event scenarios were explored, encompassing diverse event types, venues, times, and scales. These events included concerts, exhibitions, and sporting events, covering various time periods, morning, noon, and evening, with event sizes ranging from approximately 50 to approximately 200 people. To explore the advantages of modular vehicles' flexible decoupling, they were compared with fixed-capacity shuttle buses and online ride-hailing services, as shown in Table 3. A fixed-capacity shuttle bus is a vehicle with a fixed maximum capacity and a module count of 1. The maximum service time is the time it takes for the last passenger to arrive at their destination, while the average service time is the average of all passengers' arrival times. The occupancy rate for each route is calculated as the number of passengers on board divided by the number of seats. The table shows that the platooning nature of modular vehicles significantly reduces travel distance. For event 4, for example, the fixed-capacity shuttle bus traveled 556 km, the online ride-hailing service traveled 2065 km, and the modular vehicle traveled 394 km. Compared to traditional fixed-capacity shuttle buses and online ride-hailing services, the modular vehicle reduced travel distance by 29.1% and 80.9% respectively. Modular vehicles can flexibly decouple and recouple vehicle units to adjust vehicle configurations based on varying needs. Compared to fixed-capacity shuttle buses, this dynamic deployment method allows for full utilization of each vehicle's passenger capacity and reduces vehicle empty loads. During the three major passenger flows of the evacuation activity, the average occupancy rate for modular vehicles across all routes was 66.1%, compared to only 51.3% for traditional fixed-capacity buses. Furthermore, the flexibility of modular vehicles significantly reduced overall and average service times, improving overall transportation efficiency. In Activity 4, it took 70 minutes for all evacuees to their destinations using modular vehicles, with an average service time of 21 minutes. Fixed-capacity buses took 96 minutes, with an average service time of 27 minutes, while online ride-hailing vehicles took 120 minutes, with an average service time of 21 minutes. Although the average service time for modular vehicles was slightly higher than that for online ride-hailing vehicles, the increase was small, and modular vehicles significantly reduced road occupancy. Figure 4 The distribution of occupancy rates of modular vehicles and fixed-capacity shuttle buses in various service categories in various large-scale event scenarios is depicted. It can be seen that the box type of modular vehicles is generally higher than that of fixed-capacity shuttle buses. For example, the occupancy rate of fixed-capacity shuttle buses in Activity 3 was mainly distributed between 10% and 40%, while that of modular vehicles was distributed between 60% and 90%. The modular characteristics of vehicles can significantly improve seat utilization.

[0087] Table 3 Analysis of the modular advantages of modular vehicles compared to fixed-capacity shuttle buses and online ride-hailing vehicles

[0088]

[0089] Example 4

[0090] This embodiment 4 provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method for optimizing passenger service routes based on modular electric vehicles as described above is implemented. The method includes:

[0091] Based on online ride-hailing order information and road network information, the passenger's pick-up and drop-off points are matched to the nearest intersection, and the passenger's expected arrival time is extracted;

[0092] Based on the flexible decoupling characteristics of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs, this approach balances the distribution of vehicle service points by focusing on adaptive charging strategies and coordinated arrival constraints between routes. This modeling of module routes, decoupling points, module queue lengths, and charging times is then used to establish a passenger service route optimization model for densely populated areas.

[0093] Based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model in crowded areas and realize the service route optimization plan of large-scale passenger transport systems.

[0094] Example 5

[0095] This embodiment 5 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor calls the program instructions to execute the above-mentioned method for optimizing passenger service routes based on modular electric vehicles, the method comprising:

[0096] Based on online ride-hailing order information and road network information, the passenger's pick-up and drop-off points are matched to the nearest intersection, and the passenger's expected arrival time is extracted;

[0097] Based on the flexible decoupling characteristics of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs, this approach balances the distribution of vehicle service points by focusing on adaptive charging strategies and coordinated arrival constraints between routes. This modeling of module routes, decoupling points, module queue lengths, and charging times is then used to establish a passenger service route optimization model for densely populated areas.

[0098] Based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model in crowded areas and realize the service route optimization plan of large-scale passenger transport systems.

[0099] Example 6

[0100] This embodiment 6 provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the above-mentioned method for optimizing passenger service routes based on modular electric vehicles. The method includes:

[0101] Based on online ride-hailing order information and road network information, the passenger's pick-up and drop-off points are matched to the nearest intersection, and the passenger's expected arrival time is extracted;

[0102] Based on the flexible decoupling characteristics of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs, this approach balances the distribution of vehicle service points by focusing on adaptive charging strategies and coordinated arrival constraints between routes. This modeling of module routes, decoupling points, module queue lengths, and charging times is then used to establish a passenger service route optimization model for densely populated areas.

[0103] Based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model in crowded areas and realize the service route optimization plan of large-scale passenger transport systems.

[0104] In summary, the passenger service route optimization method and system based on modular electric vehicles, described in embodiments of the present invention, constructs a ride-hailing travel topology network based on factors such as the location of passenger demand points and their expected service times, establishes a time-coordinated route optimization mathematical model, and proposes a solution algorithm that integrates clustering and optimization decision-making based on gradient backpropagation to achieve modular vehicle route optimization, vehicle platoon length and decoupling point selection, and charging strategy optimization. Traditional clustering-then-decision-making methods often partition data based on empirical rules or metrics such as Euclidean distance before optimizing each subregion. Clustering and optimization decision-making are independent processes. Even if optimal solutions are obtained for each subregion using high-performance computing equipment or time-consuming, precise algorithm design, the final decision-making objective may still face performance bottlenecks due to the initial clustering results. However, an integrated algorithm, trained iteratively with the same objective, gradually adjusts the clustering results toward reducing the optimization objective, dynamically responding to changes in the optimization objective. The clustering process is tightly integrated with the global optimization objective, resulting in a more advantageous solution for complex travel optimization problems. The present invention not only ensures that the service route planning of modular electric vehicles is close to optimal, but also optimizes the vehicle's module configuration, decoupling position, charging location and charging time. While ensuring passenger delivery efficiency, it also realizes door-to-door personalized service without transfers and reduces unnecessary power resource consumption.

[0105] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide the functions for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0109] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solutions disclosed in the present invention without the need for creative work should be included in the scope of protection of the present invention.

Claims

1. A passenger service route optimization method based on modular electric vehicles, characterized in that: include: Based on online ride-hailing order information and road network information, the passenger's pick-up and drop-off points are matched to the nearest intersection, and the passenger's expected arrival time is extracted; Based on the flexible decoupling characteristics of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs, this approach balances the distribution of vehicle service points by focusing on adaptive charging strategies and coordinated arrival constraints between routes. This modeling of module routes, decoupling points, module queue lengths, and charging times is then used to establish a passenger service route optimization model for densely populated areas. Based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm, the two are integrated through gradient backpropagation to solve the passenger service path optimization model in crowded areas and realize the service route optimization plan of large-scale passenger transport systems.

2. The passenger service route optimization method based on modular electric vehicles according to claim 1, characterized in that: The end-to-end spatiotemporal similarity demand point mining algorithm uses a two-layer graph sampling and aggregation model to embed the demand point features of the travel network graph, and inputs the embedded input into the improved differentiable density peak clustering algorithm to obtain the cluster category distribution. In each category, a customized adaptive large neighborhood search algorithm is applied to make decisions and solve the established passenger service path optimization model in crowded areas, obtaining vehicle operation routes, decoupling satellite and coupling satellite locations, charging strategies, module platoon leader configurations and objective function values.

3. The passenger service route optimization method based on modular electric vehicles according to claim 2, characterized in that: The objective function values ​​of each category are summed up to obtain the total objective function value as the loss function of model training. Gradient backpropagation is used for iterative training to gradually learn in the direction of reducing the optimization target, adjust the clustering results, re-optimize the decision, repeat this process until the set termination condition is reached, and output the result.

4. The passenger service route optimization method based on modular electric vehicles according to claim 2, characterized in that: Large-scale demand points are clustered and divided into multiple areas based on spatiotemporal similarities. Modular vehicles then depart from the source in the form of a large platoon, quickly transporting passengers to the decoupling satellites in each area. Simultaneously, the large platoon is decoupled into multiple small vehicle modules based on the passenger demand in each area and the platoon size effect. Each small vehicle module undertakes different passenger transportation tasks and delivers passengers to their destinations at the desired time. Passengers do not need to get off or transfer during the ride, and door-to-door refined delivery services are provided through different vehicle modules. After the passengers are delivered, the empty small vehicle modules converge and couple again to form a large platoon, returning to the source for the next service.

5. The passenger service route optimization method based on modular electric vehicles according to claim 2, characterized in that: Demand points are initially clustered considering three spatiotemporal characteristics: geographical distance, expected delivery time window, and number of passenger demands. In each area, a customized adaptive large neighborhood search algorithm is used to solve the driving routes, vehicle platoon configuration, and charging strategy of the decision-making modular vehicles. The loss function related to the decision-making objective is designed as a function of the decision-making objective, and backpropagation is used to adjust the clustering results in the direction of reducing the decision objective function, forming an iterative training of the clustering decision-making integrated algorithm.

6. The passenger service route optimization method based on modular electric vehicles according to claim 2, characterized in that: In order to adapt to the requirements of gradient calculation for differentiability in deep learning, the traditional density peak clustering algorithm is improved. Based on the local density of the point and its distance from other points, the binary variable used to indicate whether point i belongs to cluster k is replaced by the probability distribution of the possibility of belonging to each cluster.

7. A passenger service route optimization system based on modular electric vehicles, characterized in that: include: The extraction module is used to match passenger pick-up and drop-off points to the nearest intersection based on online ride-hailing order information and road network information, and extract the passenger's expected arrival time; A construction module is used to leverage the flexible decoupling capabilities of modular vehicles, with the goal of maximizing delivery efficiency and minimizing costs. By focusing on adaptive charging strategies and coordinated arrival constraints between routes, the module's operating routes, decoupling points, module queue lengths, and charging times are modeled to optimize passenger service routes in densely populated areas. The solution module is used to solve the passenger service path optimization model in crowded areas based on an end-to-end spatiotemporal similarity demand point mining algorithm and a customized adaptive large neighborhood search algorithm. It integrates the two through gradient backpropagation to realize the service route optimization plan of large-scale passenger transport systems.

8. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by the processor, the passenger service path optimization method based on modular electric vehicles as described in any one of claims 1 to 6 is implemented.

9. A computer device, characterized in that: It includes a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the passenger service path optimization method based on modular electric vehicles as described in any one of claims 1 to 6.

10. An electronic device, characterized in that: include: A processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to execute instructions for implementing the passenger service path optimization method based on modular electric vehicles as described in any one of claims 1-6.

Images