Path generation method for two-wheeled electric vehicle, electronic device, and storage medium

Abstract

The application relates to a path generation method for a two-wheeled electric vehicle, an electronic device and a storage medium, which comprises the following steps: collecting data of a user terminal to obtain actual driving data and destination information of a user; analyzing the actual driving data through a pre-trained trajectory generation model to obtain a plurality of candidate points; determining a priority destination from the destination information according to a travel control strategy of the current user, performing distance analysis on the plurality of candidate points and the position information of the priority destination, and selecting the candidate point closest to the priority destination as the next motion point of the user; and completing information of the next motion point through a point generation control quantity, obtaining predicted trajectory data and sending the predicted trajectory data to the user. The application provides the best road navigation information for the user through a deep learning model, optimizes the user destination, and finally obtains the nearest point as the optimal next driving position of the user to the destination.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric vehicle travel analysis, in particular to a path generation method for two-wheeled electric vehicles, an electronic device and a storage medium. BACKGROUND

[0002] As a light, economical, low-carbon, and environmentally friendly mode of transportation in urban roads, electric two-wheelers are widely accepted by the public, and their market share is also increasing. Compared with cars, electric two-wheelers gradually show vigorous vitality in instant delivery, short-distance travel, logistics and express delivery, shared riding, and other fields. Therefore, the riding efficiency of users in urban roads has gradually become the focus of these industries. In terms of improving the riding efficiency of users, navigation algorithms are extremely important tools that can help users quickly make path decisions in unfamiliar environments and choose the best route to complete the user's riding purpose.

[0003] Currently, most existing path navigation technologies are based on related information of map roads, which are characterized as vectors or graphs, and use algorithms such as Dijkstra, A-star, D-star, etc. for calculation and guidance. However, these methods are based on established map road information, which can guarantee arrival, but are not optimal solutions. Because electric two-wheelers have a wider road driving area than cars, they can shuttle in buildings, communities, and other alleys, and even some unattainable places marked on the map. Therefore, the traditional regularized path navigation method based on road information does not fully meet the needs of electric two-wheeler users when providing navigation services. The current fast and efficient mode of transportation still depends on the experience of riders, which is particularly evident in the instant delivery industry. For example, some passable areas can quickly reach the destination, but traditional navigation guides users with a more distant route. Some complex road conditions cannot be captured by traditional algorithms, which reduces the riding efficiency of users.

[0004] Therefore, the existing path recommendation algorithm is only suitable for users of four-wheeled vehicles, and there is a lack of a more effective path navigation method for electric two-wheeled vehicle users. SUMMARY

[0005] The embodiments of the present application provide a path generation method for two-wheeled electric vehicles, an electronic device and a storage medium to at least solve the problem of lack of path navigation method for electric two-wheeled vehicle users in related technologies.

[0006] In a first aspect, the embodiments of the present application provide a path generation method for two-wheeled electric vehicles, comprising:

[0007] Data collection is performed on the user terminal to obtain actual driving data and destination information of the user, wherein the destination information includes location information of at least one destination.

[0008] analyzing the actual driving data by a pre-trained trajectory generation model, to obtain a plurality of candidate point positions; wherein the trajectory generation model is obtained by training a large number of riding trajectory data of two-wheeled electric vehicles;

[0009] determining a priority destination from the destination information according to a travel control strategy of a current user, performing distance analysis on the plurality of candidate point positions and the position information of the priority destination, and selecting a candidate point position closest to the priority destination as a next motion point position of the user;

[0010] completing information of the next motion point position by a point position generation control quantity, obtaining predicted trajectory data, and sending the predicted trajectory data to the user.

[0011] In an embodiment, the method further comprises:

[0012] In a process in which the user travels according to the predicted trajectory data, new actual driving data is obtained by real-time data acquisition;

[0013] It is determined whether the user has arrived at all destinations in the destination information, if yes, it is determined that the navigation is ended, and if no, the predicted trajectory data and the new actual driving data are spliced and input into the trajectory generation model to regenerate predicted trajectory data of a next destination.

[0014] In an embodiment, after the plurality of candidate point positions are obtained, the method further comprises:

[0015] calculating a movement probability of each candidate point position by a pre-trained trajectory generation model;

[0016] obtaining a preset probability threshold, determining whether the movement probability is greater than the probability threshold, if yes, retaining the corresponding candidate point position, and if no, deleting the corresponding candidate point position.

[0017] In an embodiment, the priority destination is determined from the destination information according to a travel control strategy of a current user, comprising:

[0018] obtaining a current position of the user, calculating a distance between each destination in the destination information and the current position, sorting the distances in descending order, and taking a destination ranked first as a priority destination;

[0019] Alternatively, a limited arrival time of each destination in the destination information is obtained, the destinations are sorted in the order of the limited arrival time, and a destination ranked first is taken as a priority destination.

[0020] In an embodiment, the training process of the trajectory generation model comprises:

[0021] constructing a model framework and determining a training strategy to obtain an initial trajectory generation model;

[0022] obtaining user historical driving data and dividing the historical driving data into a training set, a validation set and a test set;

[0023] training the initial trajectory generation model according to the training strategy and the training set, and adjusting parameters of the initial trajectory generation model through the validation set to obtain an intermediate trajectory generation model;

[0024] evaluating performance of the intermediate trajectory generation model by applying the test set, and obtaining the trajectory generation model when the evaluation result meets pre-set user requirements.

[0025] In an embodiment, the model framework comprises:

[0026] a data embedding layer for mapping multi-dimensional actual driving data or historical driving data into a high-dimensional feature vector;

[0027] a position encoding module for applying a sequential label to the high-dimensional feature vector in a trajectory sequence length direction;

[0028] a plurality of encoding modules for deeply mining trajectory features of the high-dimensional feature vector containing the sequential label to obtain multi-dimensional feature data;

[0029] a linear layer for data conversion according to the multi-dimensional feature to obtain relative motion positions of a plurality of candidate point positions;

[0030] a probability distribution calculation module for calculating a moving probability of a user to the candidate point position according to the relative motion positions.

[0031] In an embodiment, the encoding module comprises a first normalization layer, a multi-head attention layer, a second normalization layer and a feedforward layer, wherein,

[0032] the high-dimensional feature vector is normalized by the first normalization layer to obtain first feature data;

[0033] the multi-head attention layer is used to capture a dependency relationship between the first feature data;

[0034] the dependency relationship is fused with the first feature data to obtain second feature data, and the second feature data is normalized by the second normalization layer to obtain third feature data;

[0035] the third feature data is further extracted by the feedforward layer to obtain fourth feature data;

[0036] The second feature data and the fourth feature data are fused to obtain the multi-dimensional feature data.

[0037] In an embodiment, the actual driving data includes trajectory information of a plurality of points and electric two-wheeler information.

[0038] The trajectory information includes generation time, distance, and geocoding of each point; and the electric two-wheeler includes battery type, battery health, battery charge, and battery capacity.

[0039] In a second aspect, the embodiments of the present application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the path generation method for electric two-wheeled vehicles according to the first aspect.

[0040] In a third aspect, the embodiments of the present application provide a computer readable storage medium having a computer program stored thereon, and the program is executable on a processor to implement the path generation method for electric two-wheeled vehicles according to the first aspect.

[0041] The path generation method for electric two-wheeled vehicles provided by the embodiments of the present application has at least the following technical effects:

[0042] The path generation method for electric two-wheeled vehicles provided by the present application is a new navigation method by combining generative deep learning with travel control strategy. In the method of the present application, the spatiotemporal trajectory large model learning from massive user riding experience can provide the best road navigation information for users, and through deep learning, the trajectory data of the electric two-wheeled vehicle user can be provided to take a shortcut (such as a small road between office buildings); at the same time, the destination expected to be reached by the user is combined as a guide target to assist in optimizing the candidate points generated by the large model, and finally the nearest point is obtained as the optimal next driving position for the user to go to the destination, thereby providing the electric two-wheeled vehicle user with a navigation route with short distance and short time consumption.

[0043] The details of one or more embodiments of the present application are presented in the following drawings and description to make other features, objects, and advantages of the present application more apparent. BRIEF DESCRIPTION OF DRAWINGS

[0044] The drawings described herein are intended to provide further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their description serve to explain the present application, and do not constitute an improper limitation of the present application. In the drawings:

[0045] Figure 1 is a flowchart of the path generation method for electric two-wheeled vehicles in an embodiment of the present application;

[0046] Figure 2 is a basic architecture diagram of a shared battery trajectory large model-based embodiment of the present application;

[0047] Figure 3 is a flowchart of a path generation method for a two-wheeled electric vehicle in another embodiment of the present application;

[0048] Figure 4 is a basic architecture diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application is described and explained below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0050] Obviously, the accompanying drawings in the following description are only some examples or embodiments of the present application, and for those of ordinary skill in the art, the present application can be applied to other similar scenarios without creative effort based on these drawings. In addition, it can be understood that although the efforts made in this development process can be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present application, some design, manufacture, or production changes based on the technical content disclosed in the present application are only routine technical means and should not be understood as insufficient disclosure of the present application.

[0051] In the present application, "embodiment" means that the specific features, structures, or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment that is not mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in the present application can be combined with other embodiments without conflict.

[0052] Unless otherwise defined, technical terms and scientific terms used in the present application shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms "a", "an", "one", "this", and similar referents in the context of describing the application are to be construed to be inclusive, not exclusive. For example, the phrases "comprising a", "having a", "including a", "containing a", or "consisting of a" to specify one step, process, method, product, or apparatus that can include additional steps, processes, methods, products, or apparatuses not expressly listed, or can also include additional steps, processes, methods, products, or apparatuses inherent in such step, process, method, product, or apparatus. The words "connected", "coupled", and similar terms are not limited to direct, physical connections or links, but can also include electrical connections, whether direct or indirect. The term "plurality" means two or more. The term "and / or" describes association between name objects. For example, "A and / or B" can mean A alone, A and B together, or B alone. The character " / " generally means "or". The terms "first", "second", "third", and the like are used only to distinguish similar objects, and do not represent a specific order.

[0053] In recent years, with the rapid development of deep learning, a large number of deep learning-based algorithm models have emerged. Using deep learning algorithm models to learn and mine massive data for a long time can learn more output results that fit business scenarios. At present, it plays a major role in the fields of finance, medicine, meteorology, and social analysis. In the path generation navigation field, the path navigation method of the car has been very mature, and a large number of rules or AI-based navigation algorithms not only achieve excellent results in path navigation, but also have great success in automatic guidance of unmanned vehicles. However, these algorithms are all for car users, and electric two-wheeled vehicles are very different from car users in many aspects such as driving road, driving mileage, and driving habits.

[0054] For electric two-wheeled vehicle users, the present application proposes a new navigation method for electric two-wheeled vehicle users traveling by combining a generative deep learning model with a control strategy. In this method, a large model that learns the spatio-temporal trajectory of massive user riding experience can provide the best road navigation information for users; at the same time, the destination expected to be reached by the user is combined as a guide target to assist in optimizing the candidate points generated by the large model, and finally the nearest point is obtained as the optimal next driving position for the user to go to the destination.

[0055] In a first aspect, the present application provides a path generation method for an electric two-wheeled vehicle, Figure 1This is a flowchart of a path generation method for two-wheeled electric vehicles, such as... Figure 1 As shown, the method includes the following steps.

[0056] Step S1 involves collecting data from the user terminal to obtain the user's actual driving data and destination information, wherein the destination information includes the location information of at least one destination. In this embodiment, the actual driving data includes trajectory information from multiple points and information about the electric two-wheeler.

[0057] Specifically, each actual trajectory data point consists of dozens to hundreds of multi-dimensional trajectory location information points, representing user riding habits across various time periods and spanning several days. Each trajectory location information point contains two main dimensions: trajectory information and electric two-wheeler information. Trajectory information reflects the spatiotemporal attributes of the trajectory and is the result of the trajectory presentation, specifically including: time, distance, geocoding, and the next movement point. Electric two-wheeler information reflects the dynamic characteristics of the riding process and is the basis for the trajectory generation process, specifically including: battery type, battery health, battery charge, and battery level. Before inputting the data into the model, all data is labeled to facilitate feature extraction and identification by the model.

[0058] Step S2: Analyze the actual driving data using a pre-trained trajectory generation model to obtain multiple candidate points, which are used to guide the user to move towards the destination; wherein, the trajectory generation model is trained using a large amount of riding trajectory data of two-wheeled electric vehicles.

[0059] In the embodiments of this application, the training process of the trajectory generation model is as follows:

[0060] A model framework is constructed and a training strategy is determined to obtain an initial trajectory generation model; historical driving data of users is acquired and divided into a training set, a validation set, and a test set; the initial trajectory generation model is trained according to the training strategy and the training set, and the parameters of the initial trajectory generation model are adjusted through the validation set to obtain an intermediate trajectory generation model; the performance of the intermediate trajectory generation model is evaluated using the test set, and when the evaluation result meets the pre-set user requirements, the trajectory generation model is obtained.

[0061] In the embodiment of the present application, the model framework comprises a data embedding layer, a position coding module, N encoding modules (N is a natural number greater than 0), a linear layer, and a probability distribution calculation module. In the training process, the historical driving data (historical trajectory data) is used as the training input, and then the data embedding layer is used to map the multi-dimensional historical driving data into a high-dimensional feature vector; the position coding module is used to apply a sequential mark to the high-dimensional feature vector in the trajectory sequence length direction; the multiple encoding modules are used to deeply mine the trajectory features of the high-dimensional feature vector containing the sequential mark, to obtain multi-dimensional feature data; the linear layer is used to perform data conversion according to the multi-dimensional feature, to obtain the relative motion positions of the multiple candidate point positions; and the probability distribution calculation module is used to calculate the movement probability of the user to the candidate point positions according to the relative motion positions.

[0062] More specifically, the encoding module comprises a first normalization layer, a multi-head attention layer, a second normalization layer, and a feedforward layer. The high-dimensional feature vector is normalized by the first normalization layer to obtain first feature data; the multi-head attention layer is used to capture the dependency relationship between the first feature data; the dependency relationship is fused with the first feature data to obtain second feature data, and the second feature data is normalized by the second normalization layer to obtain third feature data; the third feature data is further extracted by the feedforward layer to obtain fourth feature data; and the second feature data and the fourth feature data are fused to obtain the multi-dimensional feature data.

[0063] After obtaining the trajectory generation model, the actual riding trajectory data of the two-wheeled electric vehicle of the current user (i.e., all trajectory data before the current time) is input into the trajectory generation model to obtain multiple candidate point positions and corresponding movement probabilities.

[0064] In a preferred embodiment, the main framework of the trajectory generation model of the present embodiment is as shown in Figure 2 The model framework mainly comprises a data embedding layer, a position coding module, multiple encoding modules, a linear layer, and a probability distribution calculation module. In the training process, the historical trajectory data is used as the training input, and the data embedding layer is the first processing link of the data input. The labeled multi-dimensional sequence trajectory information is mapped into a high-dimensional feature vector through this layer, to facilitate the subsequent analysis of the network structure. The position coding process applies a sequential mark to the vector in the trajectory sequence length direction, to facilitate the model to recognize the trajectory sequence. The encoding module is mainly used for deeply mining the high-dimensional trajectory feature information. The specific encoding module comprises a normalization layer 1, a Transformer multi-head attention layer, a normalization layer 2, and a feedforward layer.

[0065] The data processed by the single encoding module has the same dimension as the input feature data. The data output by the encoding module is sequentially input into the encoding modules of the same structure. After a series of encoding modules perform deep featureization on the data, the data is output to a linear layer, which is used to convert the multi-dimensional feature data into a set of feature vectors that only reflect the relative motion position of the next point. The dimension of this feature vector is the same as the number of labels of the next point. Finally, this set of vector information is input into a probability distribution calculation module to analyze the probability distribution of approaching each geographical location. The training output is the next point result, including the coordinates of the place the user is likely to go to and the possibility of going to that place (i.e., the moving probability).

[0066] During the training process of the model, the target mask is used to mask the original trajectory sequence length in the Transformer multi-head attention layer of the encoding module, which fully utilizes the data in the original trajectory sequence and improves the training efficiency of the next point prediction result. The cross-entropy loss operation is performed on the maximum probability result of a series of next points obtained by training and the position label of the real next point, and the result of this loss function is used as the basis for updating the model forward propagation parameters.

[0067] In step S3, the priority destination is determined from the destination information according to the current user's travel control strategy, the distance between the multiple candidate point positions and the priority destination information is analyzed, and the candidate point position closest to the priority destination is selected as the next motion point position of the user.

[0068] After obtaining the multiple candidate point positions, the moving probability of each candidate point position is calculated by the pre-trained trajectory generation model. A preset probability threshold is obtained, and it is determined whether the moving probability is greater than the probability threshold. If yes, the corresponding candidate point position is retained; if no, the corresponding candidate point position is deleted.

[0069] To determine the priority destination from the destination information according to the current user's travel control strategy, the current position of the user can be obtained, the distance between each destination in the destination information and the current position can be calculated, and the distances can be sorted in descending order. The destination ranked first is the priority destination. Alternatively, the arrival time limit of each destination in the destination information can be obtained, and the arrival time limits can be sorted in chronological order. The destination ranked first is the priority destination.

[0070] In a preferred embodiment, for the trajectory generation model that has completed training, a piece of historical trajectory is input, and the model generates the motion position information of the next point. However, the next point information alone cannot guide the user to move towards a specific target location. In order to establish a connection between the position information and the user's navigation towards the destination, a set of control strategies are used to assist in navigation trajectory generation.

[0071] In step S4, the next motion point is supplemented with information by the point generation control quantity, the predicted trajectory data is obtained and sent to the user. In the embodiment of the present application, the model generation method combined with control only obtains the position information of the next point. In order to supplement other dimensional information of the trajectory information, the point generation control quantity is combined to supplement, for example, the complete trajectory and the multi-dimensional information of the electric scooter are calculated by the motion process power consumption, motion average time consumption and other control quantities.

[0072] In step S5, during the user driving according to the predicted trajectory data, real-time data acquisition is performed to obtain new actual driving data. Finally, it is judged whether the user has arrived at all the destinations in the destination information. If yes, it is determined that the navigation is ended. If no, the predicted trajectory data and the new actual driving data are spliced and input into the trajectory generation model to regenerate the predicted trajectory data of the next destination.

[0073] Specifically, the supplemented trajectory information is judged whether all the destination points are reached. If all the set destinations are reached, it is determined that the navigation is ended. If all the set destinations are not reached, the inferred trajectory point information (i.e. new user trajectory data) and the current trajectory data of the user are spliced, combined with the remaining destination information, and input into the trajectory generation model for the next inference cycle. The destination information is deleted in this repeated cycle process until all the destinations are reached, and the entire path inference process is ended.

[0074] The implementation process of the control strategy is as follows Figure 3As shown, in the present embodiment, the user's current trajectory data and destination are taken as input information, the current trajectory data is analyzed by the trajectory generation model (i.e. the pre-trained model) to obtain a plurality of candidate points, and at the same time, the input destination has multiple, so the destination to be visited first is selected from the multiple destinations. For the destination to be visited first, the current geographical position of the current two-wheeled electric vehicle can be obtained from the current trajectory data, and the destination closest to the geographical position is selected as the destination to be visited first from the multiple destinations. Then, the multiple candidate points are combined and analyzed with the destination to be visited first, the candidate point closest to the destination to be visited first is selected as the next motion point of the user, and the trajectory information of the next motion point and the current geographical position of the user is completed by generating the control amount of the point, to obtain the next trajectory data. Then it is judged whether all the input destinations have been reached, if yes, the present path generation is ended; if not, the new user trajectory data and the current user trajectory data are spliced, and the unvisited destination (i.e. the remaining destination) is combined as the input information of the next round of path generation, so as to obtain the trajectory data of the remaining destination. In this way, the cycle is repeated until all the destinations have been reached.

[0075] Specifically, in the actual navigation process, the user's current trajectory data and destination information are input into the entire algorithm, and the destination can be a plurality of geographical latitude and longitude points to be reached by the user. The trajectory sequence information is analyzed by the pre-trained model (i.e. the trajectory generation module) that has been trained. The model generates a series of candidate points after operation, which indicate the probability of the user moving to the next point. Since there are many next points, we use a specific probability threshold to filter these candidate points here, and only keep the next candidate motion point with a larger probability recommended by the model. The destination information input by the user usually has multiple, so it is necessary to determine the order of visiting these destinations first. Here, we take the user's current position in the user's current trajectory data as a reference point, and select the point closest to the user among the multiple destinations as the destination to be visited first. This destination to be visited first is combined with multiple candidate points to determine which candidate point is closest to this destination, and the nearest next candidate point is taken as the motion position of the user.

[0076] In the use of the path generation method of the present application, the user can select multiple target locations on the APP or web platform and start the navigation generation process. At the same time, the APP or terminal hardware provides the user's current GPS or location information and battery-related information, which are combined with the user's target point information and returned to the embedded computing terminal or cloud for data characterization and path generation. After predicting the path generation, the embedded edge calculator or cloud will return the navigation path result to the user's interaction platform, thereby showing the user the navigation result. In the process of providing this service, the trajectory generation model will continuously learn and update the version based on the user's personalized information, thereby better providing the user with more matched navigation results and improving the user's riding efficiency.

[0077] Compared with the traditional navigation method, the present application fully utilizes the deep mining of the big model of the rider's riding experience in the real scene, so that in the navigation process, a completely different effect from the commercial navigation based on the map road framework is output. The road navigation information generated by the present application has a more efficient path than the conventional navigation, and most of the generated roads pass through some shortcut alleys and alleys between communities, with fewer traffic lights, so that the target location can be reached more quickly and conveniently. In addition, since the model learns the behavior patterns of riders at different times and the riding strategies under different battery states during the training process, the routes planned for the user at different times are different, which can automatically reduce the area passing through the poor road conditions; at the same time, the model generates trajectory information to prompt the user to replace the battery immediately during long-distance riding, reducing the user's "mileage anxiety", which is more advantageous for delivery personnel who need to quickly plan roads in complex roads.

[0078] In summary, the path generation method for two-wheeled electric vehicles provided by the embodiments of the present application uses the powerful learning ability of the deep neural network, and learns the relevant habits of users in terms of electric two-wheeled vehicle trajectory and battery performance by means of massive historical spatiotemporal trajectory data of electric two-wheeled vehicle users, thereby providing the user with a more real-scene-conforming generative path navigation capability. This method solves the problem that the traditional rule-based navigation algorithm cannot learn the riding experience of riders in complex scenarios and is difficult to adapt to the navigation of electric two-wheeled vehicle users with stronger flexibility. The new deep learning model enables users with massive learned implicit riding experience of users, and provides electric two-wheeled vehicle users and related industry staff with more efficient riding efficiency.

[0079] In a second aspect, the embodiments of the present application provide an electronic device, Figure 4 is a block diagram of an electronic device according to an example embodiment. As Figure 4 shown, the electronic device can include a processor 11 and a memory 12 storing computer program instructions.

[0080] In particular, the processor 11 can comprise a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to perform the functions of the embodiments of the application.

[0081] The memory 12 can include a mass storage for data or instructions. By way of example and not limitation, the memory 12 can include a Hard Disk Drive (HDD), floppy disk drive, a Solid State Drive (SSD), flash memory, a Compact Disc Read Only Memory (CDROM), a Digital Versatile Disk (DVD), a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. The memory 12 can be removable and / or non-removable (or fixed) as appropriate. The memory 12 can be internal or external as appropriate. In particular embodiments, the memory 12 is a Non-Volatile memory. In particular embodiments, the memory 12 includes a Read-Only Memory (ROM) and a Random-Access Memory (RAM). The ROM can be a mask-programmed ROM, a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), an Electrically Alterable ROM (EAROM), or a FLASH, or a combination of two or more of these, as appropriate. The RAM can be a Static Random-Access Memory (SRAM) or a Dynamic Random-Access Memory (DRAM), which can be a Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), an Extended Data Out Dynamic Random-Access Memory (EDODRAM), a Synchronous Dynamic Random-Access Memory (SDRAM), or the like, as appropriate.

[0082] The memory 12 can be used to store or buffer various data files required for processing and / or communication, and possible computer program instructions executed by the processor 11.

[0083] The processor 11 reads and executes the computer program instructions stored in the memory 12 to implement any one of the path generation methods for the two-wheeled electric vehicle in the above embodiments.

[0084] In an embodiment, the electronic device can further include a communication interface 13 and a bus 10. As shown, the processor 11, the memory 12, and the communication interface 13 are connected through the bus 10 and complete communication with each other. Figure 4

[0085] The communication interface 13 is used to realize the communication between the modules, devices, units and / or equipment in the embodiments of the present application. The communication interface 13 can also realize data communication with other components, such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations, etc.

[0086] ​Bus 10 includes hardware, software, or both, to couple components of electronic device to each other. Bus 10 includes, but is not limited to, at least one of a data bus, an address bus, a control bus, an expansion bus, a local bus, and the like. By way of example and not limitation, bus 10 can include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand (IB) interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or the like, or a combination of two or more of these. Where appropriate, bus 10 can include one or more buses. Although the present application is described and illustrated with a particular bus, it is contemplated that any appropriate bus or interconnect can be used.

[0087] In a third aspect, the embodiments of the present application provide a computer readable storage medium, having a program stored thereon, where the program is executed by a processor to implement the path generation method for two-wheeled electric vehicles according to the first aspect.

[0088] More specifically, the readable storage medium can include, but is not limited to, a portable disc, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any appropriate combination of the above.

[0089] In possible implementation manners, the present application can also be implemented in the form of a program product, which comprises program codes for causing a terminal device to execute steps of the path generation method for a two-wheeled electric vehicle provided by the first aspect when the program product is run on the terminal device.

[0090] The program codes for executing the present application can be written in any combination of one or more programming languages, and can be executed completely on a user device, partially on a user device, as an independent software package, partially on a user device and partially on a remote device, or completely on a remote device.

[0091] The technical features of the above-described embodiments can be combined in any manner. To make the description concise, all possible combinations of the technical features in the above-described embodiments are not described, however, as long as the combinations of the technical features do not exist in contradiction, they should be considered as falling within the scope of the present application.

[0092] The above-described embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the patent protection scope of the present application should be subject to the appended claims.

Claims

1. A path generation method for a two-wheeled electric vehicle, characterized by, The method comprises: collecting data of a user terminal to obtain actual driving data and destination information of the user, wherein the destination information comprises position information of at least one destination; analyzing the actual driving data by using a pre-trained trajectory generation model to obtain a plurality of candidate points; wherein the trajectory generation model is trained by using a large amount of riding trajectory data of electric two-wheel vehicles; determining a priority destination from the destination information according to a travel control strategy of the current user, analyzing distances between the plurality of candidate points and the position information of the priority destination, and selecting a candidate point closest to the priority destination as a next motion point of the user; completing information of the next motion point by using a point generation control quantity to obtain predicted trajectory data and sending the predicted trajectory data to the user; wherein trajectory information and electric two-wheel vehicle information are calculated by using motion process power consumption and motion average time consumption as control quantities to complete information of the motion point; wherein, after obtaining the plurality of candidate points, a movement probability of each candidate point is calculated by using the pre-trained trajectory generation model; a preset probability threshold is obtained, and it is determined whether the movement probability is greater than the probability threshold; if yes, the corresponding candidate point is retained; if no, the corresponding candidate point is deleted.

2. The method of claim 1, wherein, The method further comprises: collecting data in real time during travel of the user according to the predicted trajectory data to obtain new actual driving data; determining whether the user has arrived at all destinations in the destination information; if yes, it is determined that navigation is completed; if no, the predicted trajectory data and the new actual driving data are spliced and input into the trajectory generation model to generate predicted trajectory data of a next destination again.

3. The method of claim 1, wherein, The method of determining a priority destination from the destination information according to a travel control strategy of the current user comprises: obtaining a current position of the user, calculating distances between each destination in the destination information and the current position, sorting the distances in descending order, and taking a destination ranked first as the priority destination; or, obtaining a limited arrival time of each destination in the destination information, sorting the destinations according to the order of the limited arrival time, and taking a destination ranked first as the priority destination.

4. The method of claim 1, wherein, The training process of the trajectory generation model comprises: constructing a model framework and determining a training strategy to obtain an initial trajectory generation model; obtaining historical driving data of the user and dividing the historical driving data into a training set, a validation set and a test set; training the initial trajectory generation model according to the training strategy and the training set, adjusting parameters of the initial trajectory generation model by using the validation set, and obtaining an intermediate trajectory generation model; evaluating performance of the intermediate trajectory generation model by using the test set, and obtaining the trajectory generation model when the evaluation result meets a pre-set user requirement.

5. The method of claim 4, wherein, The model framework comprises: a data embedding layer for mapping multi-dimensional actual driving data or historical driving data into a high-dimensional feature vector; a position encoding module for applying a sequential label to the high-dimensional feature vector in a trajectory sequence length direction; A plurality of encoding modules are configured to perform trajectory feature deep mining on the high-dimensional feature vector containing sequential labels to obtain multi-dimensional feature data; A linear layer is configured to perform data conversion according to the multi-dimensional feature to obtain relative motion positions of a plurality of candidate point locations; A probability distribution calculation module is configured to calculate a movement probability of a user to the candidate point locations according to the relative motion positions.

6. The method of claim 5, wherein, The encoding module includes a first normalization layer, a multi-head attention layer, a second normalization layer, and a feedforward layer, wherein The high-dimensional feature vector is normalized by the first normalization layer to obtain first feature data; The multi-head attention layer is used to capture the dependency relationship between the first feature data; The dependency relationship is fused with the first feature data to obtain second feature data, and the second feature data is normalized by the second normalization layer to obtain third feature data; The third feature data is further extracted by the feedforward layer to obtain fourth feature data; The second feature data and the fourth feature data are fused to obtain the multi-dimensional feature data.

7. The method of claim 1, wherein, The actual driving data includes trajectory information of a plurality of point locations and electric two-wheeler information; The trajectory information includes generation time, distance, and geocoding of each point location; and the electric two-wheeler includes battery type, battery health, battery charge level, and battery capacity.

8. An electronic device, comprising: The program is executed by the processor to implement the path generation method for the two-wheeled electric vehicle according to any one of claims 1 to 7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the path generation method for the two-wheeled electric vehicle according to any one of claims 1 to 7.

Images