Shared electric vehicle return control method and system

Abstract

The present application relates to the field of shared electric vehicle management, and provides a method and system for controlling the return of shared electric vehicles. When the positioning signal of the electric vehicle to be returned is not received, it switches to the comprehensive addressing mode. In this mode, the positioning data of the mobile device bound to the electric vehicle to be returned is retrieved as the position reference benchmark point, and the candidate area is determined in combination with the adjacent parking areas. Then, the available communication types in the candidate area are obtained. When the first, second, and third communication types are included, the preferred communication type is specified according to the signal quality index, environmental parameters, and device status. Finally, based on the preferred communication type, it is determined whether to send a vehicle replacement permission message to the mobile terminal. In this way, the intelligent level of shared electric vehicle return control can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of shared electric vehicle management, and more specifically, to a method and system for controlling the return of shared electric vehicles. Background Art

[0002] Shared electric vehicles have significantly improved travel convenience, solved the "last mile" problem of public transportation, and are faster and more convenient than shared bicycles. However, standardized parking has become a challenge in operation. Usually, it relies on the electric vehicle positioning module to confirm whether the vehicle is parked in the designated area to ensure orderly vehicle return and prevent random parking. If the positioning module fails, users may face the problem of being unable to complete vehicle return, which affects the user experience and also brings inconvenience to operation and management, resulting in a difficult situation for vehicle return.

[0003] For this reason, Patent CN115691000A discloses a method, system, computer device, and storage medium for controlling the return of shared electric vehicles. In response to the situation of shared electric vehicle positioning failure, it proposes a comprehensive addressing mode to achieve efficient vehicle return. First, the position of the associated mobile terminal is used as a reference to identify the adjacent parking area as the candidate area. Then, different measures are taken according to the communication type of the candidate area: confirm that the vehicle enters the parking area through the Bluetooth signal strength (the first type), networking with other electric vehicles (the second type), or reading the electronic tag (the third type) to ensure accurate positioning and smooth vehicle return.

[0004] However, when the parking area supports multiple communication types (such as Bluetooth, networking, RFID) simultaneously, this patent does not clarify the priority or switching rules of different communication types. For example, in an area where both Bluetooth broadcast signals and networking signals of returned electric vehicles can be detected, the system may make misjudgments due to logical conflicts. This ambiguity may lead to incorrect judgments by the system, affecting the user experience and increasing the complexity of operation and management.

[0005] Therefore, an optimized shared electric vehicle scheduling scheme is desired. Summary of the Invention

[0006] This application aims at the deficiencies in the prior art and provides a method and system for controlling the return of shared electric vehicles.

[0007] According to one aspect of the present application, a method for controlling the return of a shared electric vehicle is provided, which includes: in response to not receiving the positioning signal of the electric vehicle to be returned, switching to the comprehensive addressing mode; in the comprehensive addressing mode, retrieving the positioning data of the mobile device bound to the electric vehicle to be returned and setting it as the position reference benchmark point; obtaining candidate areas based on the position reference benchmark point and one or more parking areas adjacent to the position reference benchmark point; obtaining the available communication types of the candidate areas; when the available communication types of the candidate areas include a first communication type, a second communication type, and a third communication type, specifying a preferred communication type from the first communication type, the second communication type, and the third communication type based on signal quality indicators, environmental parameters, and device status, including: based on the environmental parameters and the device status, performing communication adaptability analysis based on the environmental-device association characteristics on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type in the signal quality indicators to obtain the preferred communication type; based on the preferred communication type, determining whether to send a vehicle replacement permission message to the mobile terminal.

[0008] According to another aspect of the present application, a system for controlling the return of a shared electric vehicle is provided, which includes: a signal detection and switching module for switching to the comprehensive addressing mode in response to not receiving the positioning signal of the electric vehicle to be returned; a benchmark point setting module for retrieving the positioning data of the mobile device bound to the electric vehicle to be returned and setting it as the position reference benchmark point in the comprehensive addressing mode; a candidate area generation module for obtaining candidate areas based on the position reference benchmark point and one or more parking areas adjacent to the position reference benchmark point; a communication type acquisition module for obtaining the available communication types of the candidate areas; a communication adaptability analysis module for specifying a preferred communication type from the first communication type, the second communication type, and the third communication type based on signal quality indicators, environmental parameters, and device status when the available communication types of the candidate areas include the first communication type, the second communication type, and the third communication type, wherein the communication adaptability analysis module is configured to: based on the environmental parameters and the device status, perform communication adaptability analysis based on the environmental-device association characteristics on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type in the signal quality indicators to obtain the preferred communication type; a vehicle replacement decision module for determining whether to send a vehicle replacement permission message to the mobile terminal based on the preferred communication type.

[0009] Due to the adoption of the above technical solutions, this application has remarkable technical effects: The shared electric vehicle return control method and system provided by this application switch to the comprehensive addressing mode when the positioning signal of the electric vehicle to be returned is not received. In this mode, the positioning data of the mobile device bound to the electric vehicle to be returned is retrieved as the position reference benchmark point, and the candidate area is determined in combination with the adjacent parking areas. Then, the available communication types in the candidate area are obtained. When the first, second, and third communication types are included, the preferred communication type is specified based on the signal quality index, environmental parameters, and device status. Finally, based on this preferred communication type, it is determined whether to send a vehicle replacement permission message to the mobile terminal. In this way, the intelligent level of shared electric vehicle return control can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. They are used together with the embodiments of the present application to explain the present application and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 It is a flowchart of the shared electric vehicle return control method according to an embodiment of the present application.

[0012] Figure 2 It is a flowchart of step S5 in the shared electric vehicle return control method according to an embodiment of the present application.

[0013] Figure 3 It is a flowchart of step S55 in the shared electric vehicle return control method according to an embodiment of the present application.

[0014] Figure 4 It is a flowchart of step S551 in the shared electric vehicle return control method according to an embodiment of the present application.

[0015] Figure 5 It is a flowchart of step S551-3 in the shared electric vehicle return control method according to an embodiment of the present application.

[0016] Figure 6 It is a system block diagram of the shared electric vehicle return control system according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0018] Shared electric vehicles have improved travel convenience, solved the "last mile" problem of public transportation, and are faster than shared bicycles. However, standardizing parking is an operational challenge. Currently, it mainly relies on the positioning module to confirm the parking area. If the positioning fails, it will cause difficulties for users to return the vehicle, affecting the experience and increasing management costs. For this reason, Patent CN115691000A discloses a method, system, computer device, and storage medium for controlling the return of shared electric vehicles. It identifies the adjacent candidate parking areas through mobile terminal positioning and realizes assisted vehicle return based on the regional communication type (Bluetooth signal strength, vehicle networking, or electronic tag reading). However, this solution does not clarify the priority rules when multiple communication types coexist. For example, the simultaneous presence of Bluetooth and networking signals may lead to system logic conflicts, resulting in misjudgments, reducing the user experience, and increasing management complexity.

[0019] In view of the above technical problems, this application proposes a method for scheduling shared electric vehicles. Figure 1 The flowchart of the method for controlling the return of shared electric vehicles according to an embodiment of the present application is as follows. As Figure 1 shown, the method for controlling the return of shared electric vehicles according to an embodiment of the present application includes: S1, in response to not receiving the positioning signal of the electric vehicle to be returned, switching to the comprehensive addressing mode; S2, in the comprehensive addressing mode, retrieving the positioning data of the mobile device bound to the electric vehicle to be returned and setting it as the position reference benchmark point; S3, based on the position reference benchmark point and one or more parking areas adjacent to the position reference benchmark point, obtaining candidate areas; S4, obtaining the available communication types of the candidate areas; S5, when the available communication types of the candidate areas include the first communication type, the second communication type, and the third communication type, specifying the preferred communication type from the first communication type, the second communication type, and the third communication type based on the signal quality index, environmental parameters, and device status; S6, based on the preferred communication type, determining whether to send a vehicle change permission message to the mobile terminal.

[0020] In step S1, in response to not receiving the positioning signal of the electric vehicle to be returned, switching to the comprehensive addressing mode. It should be understood that shared electric vehicles rely on their built-in positioning modules (such as GPS or Beidou systems) to determine their positions. However, in some cases, the positioning module may fail or malfunction, resulting in the inability to obtain accurate position information. Especially in underground parking lots, high-rise dense areas, or other areas with severe signal shielding, satellite signals may be severely interfered, making the positioning module unable to work properly. Therefore, by switching to the comprehensive addressing mode, even when the positioning module cannot provide position information, the goal of efficiently returning the vehicle can still be achieved. In particular, the comprehensive addressing mode is designed to use other forms of technical means to make up for the lack of positioning information, including but not limited to Bluetooth communication, Wi-Fi networking, and radio frequency identification technology (RFID).

[0021] In step S2, in the integrated addressing mode, the positioning data of the mobile device bound to the electric vehicle to be returned is retrieved and set as the position reference benchmark point. Accordingly, when the built-in positioning module of the shared electric vehicle fails or cannot provide accurate position information, an alternative method is needed to determine the approximate position of the vehicle. Most users operate shared electric vehicles through applications on mobile devices such as smartphones. Therefore, mobile devices are usually carried around and have good network connection capabilities, which facilitates obtaining position information. Specifically, users' mobile devices often have strong signal reception capabilities and, in many cases, are more adaptable to complex environments (such as the city center with high-rise buildings or underground parking lots) than the positioning systems of electric vehicles themselves. Therefore, using the positioning data of mobile devices can significantly improve the reliability of position determination. In this way, by using the positioning data of mobile devices as the position reference benchmark point, even if there are problems with the electric vehicle's own positioning module, users can still be effectively guided to find a suitable parking area and complete the vehicle return operation. This ensures the continuity and efficiency of the entire vehicle return process.

[0022] In step S3, one or more parking areas adjacent to the position reference benchmark point are used to obtain candidate areas based on the position reference benchmark point. That is, by identifying nearby parking areas based on the reference position provided by the mobile terminal and marking them as potential vehicle return locations, not only the problem of positioning module failure is overcome, but also efficient navigation guidance is provided to help users quickly find suitable candidate parking areas.

[0023] The following is a detailed description of a specific implementation process of "using one or more parking areas adjacent to the position reference benchmark point to obtain candidate areas based on the position reference benchmark point": First, the position reference benchmark point needs to be obtained through the mobile device bound to the electric vehicle to be returned. The mobile device generates real-time geographical coordinate data containing longitude and latitude information through the built-in Global Positioning System (GPS), Beidou satellite navigation system, or hybrid positioning technology based on cellular network base stations and Wi-Fi access points. After denoising processing (such as filtering coordinate points with abnormal fluctuations) of these positioning data, they are confirmed by the system as the position reference benchmark point, serving as the core coordinates for subsequent area screening. The acquisition process of this benchmark point needs to meet multi-scenario applicability. For example, in an indoor environment, when satellite signals are blocked, it automatically switches to Wi-Fi fingerprint positioning or Bluetooth beacon positioning to ensure stable and effective position information can be obtained in different environments.

[0024] Next, the operator needs to pre - construct a complete basic data system for the parking areas in the system. The geographical information of each parking area includes a unique identifier, spatial range, permitted vehicle types, real - time status, and additional attributes. Among them, the definition of the spatial range is divided into two forms: for regular areas (such as rectangular parking lots), the boundary is described by a set of polygon vertex coordinates; for open areas (such as circular parking areas demarcated by the roadside), the center coordinate and radius combination is used to define it. The permitted vehicle types need to be clearly marked as "shared electric vehicles" to exclude areas only applicable to shared bicycles or motor vehicles. The real - time status includes "available", "closed", "full", etc., which are updated in real - time through Internet of Things sensors (such as geomagnetic induction devices) or manual reports to ensure the consistency between the system data and the physical scenario. In addition, the parking area data needs to be connected to the map service API for dynamic calibration, and information such as road planning changes and temporary controls is synchronized regularly to avoid incorrect area judgments caused by outdated basic data.

[0025] After obtaining the location reference benchmark point and pre - processing the parking area data, the system enters the spatial screening stage for adjacent parking areas. Based on the spatial analysis function of the geographic information system, first, the geometric center coordinates or boundary ranges of all pre - stored parking areas are extracted, and the spatial distance between each parking area and the location reference benchmark point is calculated. The distance calculation uses the spherical distance formula (such as the Haversine formula), which takes into account the curvature of the earth and accurately calculates the shortest distance between two points on the earth's surface through longitude and latitude coordinates. To improve the screening efficiency, the system sets a dynamic distance threshold, which is adjusted according to the characteristics of urban areas: in high - density urban areas (such as commercial centers), the threshold is set to 300 meters to ensure that users can obtain candidate areas within walking distance; in low - density suburbs, the threshold is extended to 1000 meters to cover a wider range of potential parking areas. During the screening process, all parking areas whose distance from the benchmark point exceeds the threshold are initially excluded, forming a preliminary candidate set containing adjacent areas.

[0026] After the preliminary candidate set is formed, the system needs to perform validity filtering on the real - time status of the parking areas. First, parking areas with the status of "closed" are excluded. Such areas may be temporarily prohibited from returning vehicles due to construction, event control, etc. Recommending them to users will result in failed vehicle returns. Secondly, parking areas with the remaining number of parking spaces being 0 are filtered out to prevent users from going to areas without available spaces and improve the experience efficiency. In the vehicle type verification link, by matching the "shared electric vehicle" attribute permitted in the parking area, areas that only support other vehicle types are excluded to ensure the pertinence of the screening results. The real - time nature of the status data is guaranteed through a two - way data synchronization mechanism: Internet of Things devices collect the parking space status in real - time and upload it to the cloud, and operators can manually update abnormal statuses through the management background to ensure that the system only retains parking areas that can be used normally during screening.

[0027] After the completion status filtering, the system needs to perform geofence matching and regional boundary verification to address potential errors in positioning data. Since the positioning accuracy of mobile devices is typically in the range of 5 - 50 meters (affected by the environment), a buffer mechanism needs to be introduced in the processing of the parking area boundary. For a parking area defined by a polygon, the system calculates the minimum distance between the reference point and each side of the polygon. If this distance is less than the preset buffer value (such as 10 meters), it is considered that the reference point is within the effective associated range of the parking area; for a circular parking area, the distance between the reference point and the center of the circle is compared with the radius, and at the same time, the radius is allowed to expand by 10% to cover the boundary offset that may be caused by positioning errors. The core of this link is to consider the positioning errors through the spatial buffer algorithm, avoid excluding compliant parking areas due to minor position deviations, and ensure that users can correctly obtain the area when they are actually near the parking area.

[0028] After spatial screening, status filtering, and boundary verification, the eligible parking areas form a set of candidate areas. For the convenience of subsequent processing and user guidance, the system needs to perform a logical sorting on this set. The sorting rules include distance priority (from near to far according to the distance from the reference point), parking space priority (from more to less according to the remaining number of parking spaces), or comprehensive priority (combining distance, the number of parking spaces, and historical usage frequency). The information of the sorted candidate areas is transmitted to the user-side APP through the application programming interface (API) and presented in the form of map annotations or lists. The annotation content includes the name of the parking area, distance, remaining parking spaces, and the walking navigation path. Users can select the target area according to their own needs, and the system synchronously sends a verification request to the communication detection module of this area to enter the subsequent communication type adaptation analysis process.

[0029] During the implementation process, the real-time data is achieved through the combination of Internet of Things technology and manual verification to ensure that the parking area status is consistent with the actual situation; the positioning error compensation is achieved through the setting of the buffer range to cover the possible positioning deviation of the device and improve the inclusiveness of area screening; the dynamic threshold adjustment is based on urban planning and user habits to flexibly adapt to the needs in different scenarios and avoid the screening results being too broad or too narrow. In addition, the system needs to have an exception handling mechanism. For example, when there are no eligible parking areas around the reference point, the screening threshold is automatically expanded or the user is prompted to go to the nearest compliant area to ensure the integrity of the process.

[0030] In step S4, obtain the available communication types of the candidate area. It should be understood that the available communication types of the candidate area specifically include information such as the communication technology categories supported by the candidate area and the corresponding hardware or signal interaction methods, and are specifically manifested as follows: the first communication type corresponds to the Bluetooth broadcast signal deployed in the parking area, the second communication type corresponds to the wireless networking signal composed of the already-returned electric vehicles in the parking area, and the third communication type corresponds to the radio frequency electronic tag set in the parking area. In short, by obtaining and analyzing the available communication types of the candidate area, the model can understand the applicability of different communication types, and then select the optimal communication type most suitable for the current situation and decide whether to send the vehicle return permission information. This connection enables the system to dynamically select the appropriate communication type according to the actual situation, achieve accurate judgment, ensure the smooth progress of the vehicle return process, and improve the intelligence and accuracy of the shared electric vehicle return control.

[0031] In step S5, when the available communication types of the candidate area include the first communication type, the second communication type, and the third communication type, specify the optimal communication type from the first communication type, the second communication type, and the third communication type based on the signal quality index, environmental parameters, and device status. In particular, the first communication type is the Bluetooth broadcast signal, the second communication type is the networking signal; the third communication type is the electronic tag. It should be understood that when the parking area supports multiple communication types at the same time (such as Bluetooth broadcast signal, networking signal, and electronic tag), it is necessary to make a priority decision to determine the optimal communication type. Each communication method has its own advantages and disadvantages in practical applications and is affected by environmental conditions and device status: the Bluetooth broadcast signal is convenient to detect and widely deployed, but it is easily blocked and interfered in a complex environment, resulting in unstable signal quality; the networking signal depends on the number of already-returned electric vehicles in the surrounding area and the network environment. If the number of vehicles in the area is insufficient or the network condition is poor, its effect will be limited; the electronic tag (RFID) provides stable and reliable identification ability, but highly depends on specific hardware deployment. Once the device fails or the tag is damaged, it will not be able to work properly. Therefore, by comprehensively evaluating the signal quality index, environmental parameters, and device status, the system can dynamically select the communication method most suitable for the current situation, ensure efficient and reliable position confirmation and vehicle return operation, thereby maximizing the advantages of each communication technology and minimizing the impact of potential adverse factors, and improving the flexibility and user experience of the overall system.

[0032] To this end, in response to the technical problems in the above-mentioned background technology, the technical concept of this application is: first, the signal quality indicators such as Bluetooth signal strength, networking device density and electronic tag reading status are converted into quantifiable and comparable coding vectors, and environmental parameters (such as weather conditions, electromagnetic interference intensity) and device status (such as mobile terminal power, vehicle sensor health) are mapped into structured coding vectors. Subsequently, the characteristics of communication type signal quality, environment and device status are interactively analyzed through deep neural encoding and decoding to explore the potential correlation between the three - for example, in a rainstorm environment, the Bluetooth signal attenuates significantly but the electronic tag is more stable, or when there are sufficient cars returned nearby, the reliability of the networking signal is better than other types. Finally, the model outputs a comprehensive fitness score for each communication type, and automatically selects the type with the highest score as the basis for verification. This mechanism effectively solves the problem of logical conflicts when multiple communication types coexist. By dynamically weighing the real-time changes in environmental interference, device status and signal quality, it ensures that the system prioritizes the use of communication methods with strong anti-interference and high data consistency in the current scenario to complete the vehicle return verification. It not only avoids the lack of adaptability of traditional fixed priority rules in complex environments, but also reduces the misjudgment rate through intelligent decision-making, thereby improving the smoothness of the vehicle return process and the accuracy of operational management.

[0033] Specifically, in the embodiment of the present application, the step S5 includes: based on the environmental parameters and the device status, performing communication adaptability analysis based on the environment-device association characteristics on the first communication type signal quality indicator, the second communication type signal quality indicator and the third communication type signal quality indicator in the signal quality indicators to obtain the preferred communication type. More specifically, Figure 2 FIG. 5 is a flow chart of step S5 in the shared electric vehicle return control method according to an embodiment of the present application. Figure 2As shown, step S5 includes: S51, extracting the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type from the signal quality indicators; S52, performing one-hot embedding encoding on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type respectively to obtain the one-hot embedding encoding vectors of the signal quality indicators of the first to third communication types; S53, performing structured mapping encoding on the environmental parameters and the device status to obtain the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector; S54, fusing the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector to obtain the environmental parameter-device status structured mapping joint encoding vector; S55, performing communication adaptability deep neural encoding and decoding on the one-hot embedding encoding vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain the first adaptability, the second adaptability, and the third adaptability; S56, taking the communication type corresponding to the maximum value among the first adaptability, the second adaptability, and the third adaptability as the preferred communication type.

[0034] In step S51, the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type are extracted from the signal quality indicators. It should be understood that for the first communication type (Bluetooth broadcast signal), the signal quality indicators include parameters such as the strength, signal-to-noise ratio, and signal stability of the Bluetooth broadcast signal, which can reflect the signal transmission quality and detectability, and these parameters can reflect the signal interaction ability between the electric vehicle to be returned and the Bluetooth module in the selected area; for the second communication type (networking signal), the signal quality indicators cover parameters such as the strength, connection success rate, data transmission rate, signal delay, and packet loss rate of the networking signal, which can characterize the networking connection reliability and stability between the electric vehicle to be returned and the returned electric vehicles, and are used to measure the quality and efficiency of networking communication; for the third communication type (electronic tag reading signal), the signal quality indicators include parameters such as the strength of the radio frequency signal, the tag reading success rate, and the signal anti-interference ability, which can reflect the effectiveness of the signal interaction between the radio frequency module of the electric vehicle to be returned and the electronic tag in the selected area, and determine whether the electronic tag information can be accurately read. In short, by extracting the signal quality indicators of each communication type, the actual availability of each communication type in the current scenario can be objectively evaluated, providing a key basis for comprehensively determining the preferred communication type based on signal quality, environmental parameters, and device status, ensuring that the system can select the communication method with the best and most stable signal quality to determine whether the electric vehicle to be returned enters the parking area when multiple communication types coexist, thereby avoiding misjudgment or missed judgment caused by poor signal quality and ensuring the accuracy and reliability of the shared electric vehicle return control.

[0035] In step S52, one-hot embedding encoding is respectively performed on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type to obtain one-hot embedding encoding vectors of the signal quality indicators of the first to third communication types. Accordingly, considering that the signal quality indicators of different communication types have significant heterogeneity and dimensional differences. The intensity of Bluetooth broadcast signals is in dBm, the quality of networking signals depends on the number of neighboring devices and the network latency time, and the recognition success rate of electronic tags is expressed as a binary state or a percentage of error rate. If these multi-dimensional heterogeneous data are directly input into the model for joint analysis, it will lead to a chaotic feature space. For example, a high-intensity Bluetooth signal and a high-density networking device may be numerically similar, but their actual physical meanings and reliability impact factors are completely different. Therefore, in this application, one-hot embedding encoding is respectively performed on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type to obtain one-hot embedding encoding vectors of the signal quality indicators of the first to third communication types. That is, by introducing the one-hot embedding encoding technology, the signal quality indicators of each communication type are mapped into high-dimensional sparse vectors. Specifically, each type of signal quality indicator (such as Bluetooth signal strength is divided into three levels: weak / medium / strong) is discretized and graded, and a unique one-hot encoding position is assigned to each level. On this basis, the one-hot encoding vector is transformed into a dense vector representation through the embedding layer, and this process is essentially to project the original signal quality indicator from the physical measurement space to the learnable feature space.

[0036] In step S53, the environmental parameters and the device status are subjected to structured mapping encoding to obtain an environmental parameter structured mapping encoding vector and a device status structured mapping encoding vector. Specifically, in the embodiment of the present application, the step S53 includes: performing structured mapping encoding on the environmental parameters and the device status based on a multi-layer perceptron to obtain the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector. Correspondingly, considering that the environmental parameters and the device status contain various types of data, such as weather conditions (sunny, rainy, snowy, etc.) and electromagnetic interference intensity (strong, medium, weak, etc.) in the environmental parameters; mobile terminal battery power (specific battery percentage) and in-vehicle sensor health (normal, minor fault, serious fault, etc.) in the device status. These data are not only diverse in type, but also the relationships between them are relatively complex and not simple linear relationships. Therefore, in order to effectively capture these complex data features and relationships, the present application performs structured mapping encoding on the environmental parameters and the device status based on a multi-layer perceptron to obtain the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector. Specifically, a multi-layer perceptron (MLP) is used for structured mapping encoding, and through the cascading effect of its stacked fully connected layers and non-linear activation functions (such as ReLU), the original environmental parameter and device status data are projected into a low-dimensional dense vector space. Specifically, for the categorical variables (such as weather type) and continuous variables (such as temperature value) in the environmental parameters, the MLP first performs normalization and feature crossing processing. For example, the combination of "heavy rain + high humidity" is mapped to an activation state of a specific dimension; for the multi-source sensor data (such as gyroscope offset and battery internal resistance value) in the device status, the MLP realizes automatic focusing on abnormal states through the weight allocation of hidden layer nodes. This processing is essentially constructing a learnable feature distiller to extract abstract representations strongly related to the communication quality decision from the original data, so as to more accurately reflect the actual situation of the environment and the device.

[0037] In step S54, the environmental parameter structured mapping encoded vector and the device state structured mapping encoded vector are fused to obtain an environmental parameter-device state structured mapping joint encoded vector. It should be understood that there is a complex interaction relationship between environmental parameters and device states, and considering either factor alone cannot comprehensively represent the true working conditions faced by the system. For example, a high-temperature environment may cause a decrease in the heat dissipation efficiency of an in-vehicle electronic tag reader, and the actual degree of this effect depends on the heat dissipation design state of the device itself; similarly, the low battery condition of a mobile terminal will limit the Bluetooth scanning frequency, but this limiting effect may not be significant in an environment with weak electromagnetic interference. If the feature vectors of environmental parameters and device states are processed in isolation, the model will not be able to capture such cross-domain coupling relationships, resulting in deviations in the decision-making basis. Based on this, in the technical solution of this application, the environmental parameter structured mapping encoded vector and the device state structured mapping encoded vector are fused to obtain an environmental parameter-device state structured mapping joint encoded vector, thereby providing a global decision-making basis for the optimal selection of communication types.

[0038] In step S55, the one-hot embedded encoded vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device state structured mapping joint encoded vector are respectively subjected to communication adaptability deep neural encoding and decoding to obtain a first adaptability, a second adaptability, and a third adaptability. Specifically, Figure 3 FIG. is a flowchart of step S55 in the shared electric vehicle return control method according to an embodiment of the present application. As Figure 3 shown, the step S55 includes: S551, respectively performing chained local interaction communication adaptability deep neural encoding on the one-hot embedded encoded vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device state structured mapping joint encoded vector to obtain first to third communication type multi-dimensional feature interaction response encoded vectors; S552, based on the first to third communication type multi-dimensional feature interaction response encoded vectors, obtaining the first adaptability, the second adaptability, and the third adaptability.

[0039] In step S551, the one-hot embedded encoded vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device state structured mapping joint encoded vector are respectively subjected to chained local interaction communication adaptability deep neural encoding to obtain first to third communication type multi-dimensional feature interaction response encoded vectors. Specifically, Figure 4 FIG. is a flowchart of step S551 in the shared electric vehicle return control method according to an embodiment of the present application. As Figure 4As shown, the step S551 includes: S551-1, performing one-dimensional convolutional local implicit feature encoding on the one-hot embedded encoding vector of the first communication type signal quality index and the joint encoding vector of the environmental parameter-device state structured mapping respectively to obtain a set of one-hot embedded encoding vectors of the local implicit features of the first communication type signal quality index and a set of joint encoding vectors of the local implicit features of the environmental parameter-device state structured mapping; S551-2, performing monomer feature interaction on each corresponding one-hot embedded encoding vector of the local implicit features of the first communication type signal quality index and the joint encoding vector of the local implicit features of the environmental parameter-device state structured mapping in the set of one-hot embedded encoding vectors of the local implicit features of the first communication type signal quality index and the set of joint encoding vectors of the local implicit features of the environmental parameter-device state structured mapping respectively to obtain a set of local implicit interaction response encoding vectors of the multi-dimensional features of the first communication type; S551-3, performing modulation chain reasoning based on attention weight interaction on the set of local implicit interaction response encoding vectors of the multi-dimensional features of the first communication type to obtain a multi-dimensional feature interaction response encoding vector of the first communication type.

[0040] It should be understood that there is no simple global linear relationship between the signal quality index (such as Bluetooth signal strength, networking device density) and the joint features of the environment and devices (such as temperature and humidity, device health), but there are a large number of local dynamic coupling effects. For example, when the electromagnetic interference intensity is in a specific range, the attenuation degree of the Bluetooth signal and the device antenna gain index may form a non-linear synergistic effect, but this local correlation is easily diluted by the global feature interaction model. In addition, the reliability of different communication technologies shows different sensitivity characteristics under different combinations of environmental devices - the sensitivity of the networking signal to network latency increases with the increase of temperature at low device density, while the reading stability of the electronic tag may have a strong correlation with the heat dissipation design parameters in a high-temperature and high-humidity environment. If the traditional fully connected interaction method is adopted, it is difficult for the model to focus on these key local correlations, resulting in insufficient fine-grained decision-making basis. Therefore, in the technical solution of this application, the one-hot embedded encoding vectors of the first to third communication type signal quality indexes and the joint encoding vector of the environmental parameter-device state structured mapping are respectively subjected to chain local interaction communication adaptation depth neural encoding to obtain the multi-dimensional feature interaction response encoding vectors of the first to third communication types. In this way, through the depth neural encoding, the multi-dimensional feature interaction response encoding vectors of the first to third communication types can be obtained. These vectors contain the information of the interaction between the communication type signal quality, environmental parameters and device states, and are a more advanced and comprehensive expression of the original data.

[0041] Specifically, in the embodiment of the present application, the step S551-1 includes: performing one-dimensional convolutional local implicit feature encoding on the one-hot embedding encoding vector of the first communication type signal quality index and the joint encoding vector of the environmental parameter-device state structured mapping respectively to obtain a set of one-hot embedding encoding vectors of the local implicit features of the first communication type signal quality index and a set of joint encoding vectors of the local implicit features of the environmental parameter-device state structured mapping, which can be expressed by the following formula: ; where is the one-hot embedding encoding vector of the first communication type signal quality index, is the joint encoding vector of the environmental parameter-device state structured mapping, is the one-dimensional convolutional local implicit feature encoding, is the length of the one-dimensional convolutional kernel, and are respectively the 1st, 2nd, th, and th one-hot embedding encoding vectors of the local implicit features of the first communication type signal quality index in the set of one-hot embedding encoding vectors of the local implicit features of the first communication type signal quality index, and are respectively the 1st, 2nd, th, and th joint encoding vectors of the local implicit features of the environmental parameter-device state structured mapping in the set of joint encoding vectors of the local implicit features of the environmental parameter-device state structured mapping, is and the number of vectors in, and and have the same length.

[0042] It should be understood that due to the non-linear synergistic effect of signal attenuation and antenna gain, it is difficult for traditional fully connected networks to capture such fine-grained patterns. However, one-dimensional convolution can efficiently extract local serialized patterns (such as the interval segment of signal strength mutation or the periodic fluctuation of temperature and humidity) in the one-hot embedding encoding vector of the first communication type signal quality index and the joint encoding vector of the environmental parameter-device state structured mapping through the sliding window mechanism of parameter sharing. Specifically, small-size convolutional kernels focus on fine-grained features (such as instantaneous signal jitter), and large-size convolutional kernels capture macroscopic trends (such as temperature cumulative effect), so as to provide a structured set of local implicit features for subsequent interaction modeling.

[0043] Specifically, in the embodiments of the present application, the step S551-2 includes: performing single-feature interaction on each corresponding first communication type signal quality indicator local implicit feature one-hot embedding coding vector and environment parameter-device status structured mapping local implicit feature joint coding vector in the set of the first communication type signal quality indicator local implicit feature one-hot embedding coding vectors and the set of the environment parameter-device status structured mapping local implicit feature joint coding vectors to obtain a set of first communication type multi-dimensional feature local implicit interaction response coding vectors, which can be expressed by the following formula: ; where is dot product by position, is addition by position, is subtraction by position, is concatenation operation, is the th communication type multi-dimensional implicit feature interaction response weight matrix in the set of communication type multi-dimensional implicit feature interaction response weight matrices, is the th communication type multi-dimensional implicit feature interaction response bias vector in the set of communication type multi-dimensional implicit feature interaction response bias vectors, is the th first communication type multi-dimensional feature local implicit interaction response coding vector in the set of first communication type multi-dimensional feature local implicit interaction response coding vectors.

[0044] It should be understood that since the reliability of Bluetooth signals exhibits differential sensitivity characteristics under different environmental combinations (such as increased network latency sensitivity at high temperatures), interaction needs to be constructed at the local implicit feature level. Performing single-feature interaction on each group of the local implicit features of Bluetooth signals (such as signal strength fluctuation patterns) and the local features of environmental parameters (such as temperature and humidity segments related to heat dissipation design) can effectively capture local non-linear coupling (such as the multiplicative relationship between signal attenuation and heat dissipation parameters in the high-temperature range). That is, this step avoids the noise interference of global interaction and focuses on key local associations (such as the co-variation of Bluetooth signals and device health at specific temperature and humidity thresholds), generating a set of first communication type multi-dimensional feature local implicit interaction response coding vectors.

[0045] Specifically, Figure 5 is the flowchart of step S551-3 in the shared electric vehicle return control method according to the embodiments of the present application. As Figure 5As shown, the step S551-3 includes: S551-31, determining the chain inference attention weights of each first communication type multi-dimensional feature local implicit interaction response coding vector based on the feature distribution characteristics of each first communication type multi-dimensional feature local implicit interaction response coding vector in the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain a set of first communication type multi-dimensional feature chain inference attention weights; S551-32, based on the set of first communication type multi-dimensional feature chain inference attention weights, performing weighted modulation on the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain a set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors; S551-33, inputting the set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors into a chain inference engine based on a forward LSTM model to obtain the first communication type multi-dimensional feature interaction response coding vector.

[0046] More specifically, in the embodiment of the present application, the step S551-31 includes: determining the chain inference attention weights of each first communication type multi-dimensional feature local implicit interaction response coding vector based on the feature distribution characteristics of each first communication type multi-dimensional feature local implicit interaction response coding vector in the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain a set of first communication type multi-dimensional feature chain inference attention weights, which can be represented by the following formula: ; where is the th eigenvalue in is the square of the Euclidean norm of the calculation vector, is the number of eigenvalues in is a function, is the th first communication type multi-dimensional feature chain inference attention weight in the set of first communication type multi-dimensional feature chain inference attention weights.

[0047] It should be understood that there are important differences in the interaction between Bluetooth signals and the environment (for example, the interaction under high temperature and high humidity has more decision-making value than at normal temperature), and attention weights need to be dynamically allocated based on the feature distribution characteristics. For example, when electromagnetic interference is strong, the feature distribution sparsity of the signal-antenna gain interaction response is higher, and a greater weight needs to be given; while the uniform distribution interaction in a stable environment can reduce the weight. That is, this step can screen out local interaction patterns that have a significant impact on Bluetooth communication adaptability (such as the signal mutation interval caused by poor heat dissipation), and thus provide a basis for subsequent modulation.

[0048] More specifically, in the embodiment of the present application, the step S551-32 includes: performing a non-linear perturbation compensation based on chain reasoning drive on the set of multi-dimensional feature chain reasoning attention weights of the first communication type to obtain a set of multi-dimensional feature chain reasoning compensation attention weights of the first communication type, and this process can be expressed by the formula: ; where is the th multi-dimensional feature chain reasoning linear transformation interaction energy intensity in the set of multi-dimensional feature chain reasoning linear transformation interaction energy intensities of the first communication type, is the th multi-dimensional feature chain reasoning linear transformation difference energy intensity in the set of multi-dimensional feature chain reasoning linear transformation difference energy intensities of the first communication type, is the th multi-dimensional feature chain reasoning non-linear perturbation energy intensity in the set of multi-dimensional feature chain reasoning non-linear perturbation energy intensities of the first communication type, is the logarithmic function value with the natural constant as the base, is the th multi-dimensional feature chain reasoning periodic dynamic compensation factor in the set of multi-dimensional feature chain reasoning periodic dynamic compensation factors of the first communication type, is the th multi-dimensional feature chain reasoning periodic auxiliary phase factor in the set of multi-dimensional feature chain reasoning periodic auxiliary phase factors of the first communication type, and are respectively and corresponding dynamic weight coefficients, is the th multi-dimensional feature chain reasoning compensation attention weight in the set of multi-dimensional feature chain reasoning compensation attention weights of the first communication type.

[0049] Using the set of multi-dimensional feature chain reasoning compensation attention weights of the first communication type, perform weighted modulation on the set of multi-dimensional feature local implicit interaction response encoding vectors of the first communication type to obtain the set of multi-dimensional feature local implicit interaction response modulation encoding vectors, and this process can be expressed by the formula: ; where and are respectively the 1st, 2nd, the th and the a first communication type multi-dimensional feature local implicit interaction response modulation coding vector, is a set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors.

[0050] Specifically, when calculating each first communication type multi-dimensional feature local implicit interaction response coding vector , it is necessary to introduce and the interaction features between them, such as and so on. Its essence corresponds to the spatial dynamic decomposition of different interaction rules, and then a differential spatial constraint association mechanism is formed within the interaction space.

[0051] To improve the rule dynamic correction ability of the first communication type multi-dimensional feature chain inference attention weight, it is necessary to adjust the weight calculation based on the spatial dynamic decomposition of the interaction rules. Specifically: the gradient direction of the linear transformation effect (such as ) is consistent with the spatial interaction direction; the gradient direction of the non-linear perturbation effect (such as ) is orthogonal to the spatial interaction direction.

[0052] For the vector statistics of different action types (such as ), the non-linear perturbation effect will cause a regional periodic dynamic effect in the direction of the linear transformation effect, and the corresponding first communication type multi-dimensional feature chain inference period dynamic compensation factor can be defined as: ; where as a linear localization representation, the energy intensity of the non-linear perturbation effect increases with the growth of the logarithmic dimension of the linear localization representation.

[0053] At the same time, the non-linear perturbation effect will cause the periodic auxiliary phase evolution of the linear localization representation, and its expression is: , finally, the first communication type multi-dimensional feature chain inference attention weight is corrected through weighted sum . This method enhances the relevance of the spatial constraint association mechanism within the interaction space by distinguishing the action differences of different interaction rules in the spatial constraint rule system, thereby improving the calculation accuracy of the first communication type multi-dimensional feature chain inference attention weight.

[0054] It should be understood that by scaling the first communication type multi-dimensional feature local implicit interaction response encoding vector through the first communication type multi-dimensional feature chain inference compensation attention weight, the key local interaction patterns (such as the strong correlation vector of signal - heat dissipation parameters at high temperatures) can be strengthened, and redundant information (such as low-sensitivity interactions in the normal temperature range) can be suppressed. For example, when the Bluetooth signal shows periodic attenuation in the high-temperature range, its corresponding modulation vector will be highlighted, while irrelevant temperature fluctuation segments will be weakened. That is, this step ensures that the chain inference engine preferentially processes local interaction combinations that have a significant impact on the stability of the Bluetooth connection to improve the pertinence of decision-making.

[0055] More specifically, in the embodiment of the present application, the step S551-33 includes: inputting the set of the first communication type multi-dimensional feature local implicit interaction response modulation encoding vectors into a chain inference engine based on a forward LSTM model to obtain the first communication type multi-dimensional feature interaction response encoding vector, which can be represented by the following formula: ; where is the forward LSTM encoding, is the first communication type multi-dimensional feature interaction response encoding vector.

[0056] It should be understood that the interaction between the Bluetooth signal and the environment has temporal dependence (such as the signal deteriorating gradually due to the temperature accumulation effect), and the weighted local interaction response needs to be integrated through the sequence modeling ability of LSTM. For example, LSTM can remember the influence of the previous high temperature on the signal and correlate with the subsequent changes in heat dissipation parameters, and finally output the first communication type multi-dimensional feature interaction response encoding vector. This vector condenses the dynamic response law of the Bluetooth signal strength under different environmental conditions (such as the signal stability change under the chain action of high temperature - high humidity - heat dissipation), and can provide a global basis for communication adaptation decision-making.

[0057] In step S552, based on the first to third communication type multi-dimensional feature interactive response coding vector, the first fitness, the second fitness and the third fitness are obtained. Specifically, in an embodiment of the present application, the step S552 includes: inputting the first to third communication type multi-dimensional feature interactive response coding vectors into a decoder-based fitness analyzer to obtain the first fitness, the second fitness and the third fitness. It should be understood that the first to third communication type multi-dimensional feature interactive response coding vectors contain complex interactive information between each communication type, environmental parameters and device status. This information exists in the form of coding vectors, and it is difficult to intuitively judge the pros and cons of each communication type. The decoder in the fitness analyzer can decode these coding vectors, convert them into a form that is easy to understand and analyze, and extract the key features and performance information about the communication type in the current scenario. By obtaining the first fitness, the second fitness and the third fitness, the applicability of each communication type in the current scenario can be clearly compared.

[0058] In step S56, the communication type corresponding to the largest one among the first adaptability, the second adaptability and the third adaptability is used as the preferred communication type. Accordingly, the adaptability is the comprehensive performance evaluation result of each communication type in the current actual scenario. The higher the adaptability, the more the communication type can meet the business requirements such as shared electric vehicle return verification under the current environment and equipment conditions, such as higher accuracy and stability of data transmission, less interference, etc. Therefore, selecting the communication type with the largest adaptability is based on a comprehensive and objective evaluation, which can maximize the advantages of each communication type.

[0059] In summary, the step S5 is explained clearly. It first converts the signal quality indicators such as Bluetooth signal strength, networking device density and electronic tag reading status into quantifiable and comparable coding vectors, and maps the environmental parameters and device status into structured coding vectors. Then, the characteristics of the communication type signal quality, environment and device status are interactively analyzed through deep neural coding and decoding to explore the potential correlation between the three. Finally, the model outputs the comprehensive fitness score of each communication type and automatically selects the type with the highest score as the basis for verification. In this way, the logical conflict problem when multiple communication types coexist can be effectively solved. By dynamically weighing the real-time changes of environmental interference, device status and signal quality, it ensures that the system gives priority to the communication method with strong anti-interference and high data consistency in the current scenario to complete the vehicle return verification, which not only avoids the lack of adaptability of traditional fixed priority rules in complex environments, but also reduces the misjudgment rate through intelligent decision-making, and improves the fluency of the vehicle return process and the accuracy of operation management.

[0060] In step S6, based on the preferred communication type, it is determined whether to send car return permission information to the mobile terminal. In this way, by selecting the preferred communication type, the accuracy of judging the vehicle position can be maximally ensured. For example, in an area with numerous high-rise buildings, the Bluetooth signal may be blocked and unstable. At this time, if it is determined according to the preferred decision that the electronic tag is a more reliable communication type, and the vehicle position information fed back by the electronic tag is used to judge whether to allow car return, it can effectively avoid misjudgment caused by signal problems and ensure that car return is only allowed when the vehicle is truly in the compliant parking area. Specifically, if the preferred communication type is the first communication type (Bluetooth), the Bluetooth module of the electric vehicle to be returned will actively obtain the Bluetooth broadcast signal in the selected area, continuously detect the intensity value of the signal, and compare it with the preset intensity threshold. When it is detected that the intensity of the Bluetooth broadcast signal is stably greater than the preset value, it indicates that the vehicle is in the parking area where the Bluetooth signal coverage is effective, and the system sends car return permission information to the mobile terminal to allow the user to complete the car return operation. If the preferred communication type is the second communication type (networking), the networking module of the electric vehicle to be returned will search for the networking signal sent by the networking module of at least one returned electric vehicle in the selected area. Once an effective networking signal is detected, the networking module of the electric vehicle to be returned will establish a communication connection with the networking module of the returned electric vehicle, confirm that the vehicle is in the legal parking area through networking interaction, and after completing the networking verification, the system sends car return permission information to the mobile terminal to ensure the legality of the car return operation. If the preferred communication type is the third communication type (radio frequency), the radio frequency module of the electric vehicle to be returned will scan the electronic tag in the selected area. When the radio frequency module successfully reads the preset electronic tag information (such as a specific code or identifier) in this area, it indicates that the vehicle has entered the designated parking area, and the system sends car return permission information to the mobile terminal to confirm that the user can perform car return. Throughout the process, the system strictly conducts detection according to the verification logic corresponding to the preferred communication type, and only when the preset conditions are met in the communication verification link will it send car return permission information to the mobile terminal to ensure the accuracy and standardization of the car return operation and avoid misjudgment or illegal car return situations.

[0061] In summary, the shared electric vehicle return control method based on the embodiments of the present application is elucidated. When the positioning signal of the electric vehicle to be returned is not received, it switches to the comprehensive addressing mode. In this mode, the positioning data of the mobile device bound to the electric vehicle to be returned is retrieved as the position reference benchmark point, the selected area is determined in combination with the adjacent parking areas, and then the available communication types in the selected area are obtained. When the first, second, and third communication types are included, the preferred communication type is specified based on the signal quality index, environmental parameters, and device status. Finally, based on this preferred communication type, it is judged whether to send car return permission information to the mobile terminal. In this way, the intelligent level of shared electric vehicle return control can be effectively improved.

[0062] Figure 6The system block diagram of the shared electric vehicle return control system according to an embodiment of the present application is as follows. As Figure 6 shown, the shared electric vehicle return control system 100 according to an embodiment of the present application includes: a signal detection and switching module 110, configured to switch to an integrated addressing mode in response to not receiving a positioning signal of the electric vehicle to be returned; a reference point setting module 120, configured to, in the integrated addressing mode, retrieve positioning data of a mobile device bound to the electric vehicle to be returned and set it as a position reference point; a candidate area generation module 130, configured to obtain a candidate area based on the position reference point and one or more parking areas adjacent to the position reference point; a communication type acquisition module 140, configured to acquire available communication types of the candidate area; a communication adaptation analysis module 150, configured to, when the available communication types of the candidate area include a first communication type, a second communication type, and a third communication type, specify a preferred communication type from the first communication type, the second communication type, and the third communication type based on signal quality indicators, environmental parameters, and device status; a vehicle replacement decision module 160, configured to determine whether to send vehicle replacement permission information to a mobile terminal based on the preferred communication type.

[0063] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above-mentioned shared electric vehicle return control system 100 have been described in detail in the description of the Figures 1 to 5 shared electric vehicle return control method above, and therefore, the repeated description thereof will be omitted.

[0064] In summary, the shared electric vehicle return control system 100 based on an embodiment of the present application is clarified. When it does not receive the positioning signal of the electric vehicle to be returned, it switches to the integrated addressing mode. In this mode, it retrieves the positioning data of the mobile device bound to the electric vehicle to be returned as the position reference point, combines with the adjacent parking areas to determine the candidate area, then obtains the available communication types of the candidate area. When the first, second, and third communication types are included, it specifies the preferred communication type according to the signal quality indicators, environmental parameters, and device status. Finally, based on the preferred communication type, it determines whether to send vehicle replacement permission information to the mobile terminal. In this way, the intelligent level of the shared electric vehicle return control can be effectively improved.

Claims

1. A method for controlling the return of a shared electric vehicle, characterized in that, Including: Switch to the comprehensive addressing mode in response to not receiving the positioning signal of the electric vehicle to be returned; In the comprehensive addressing mode, retrieve the positioning data of the mobile device bound to the electric vehicle to be returned and set it as the position reference benchmark point; Based on the position reference benchmark point and one or more parking areas adjacent to the position reference benchmark point, obtain the candidate areas; Obtain the available communication types of the candidate areas; When the available communication types of the candidate areas include the first communication type, the second communication type, and the third communication type, specify the preferred communication type from the first communication type, the second communication type, and the third communication type based on the signal quality index, environmental parameters, and device status, including: based on the environmental parameters and the device status, perform communication adaptability analysis on the first communication type signal quality index, the second communication type signal quality index, and the third communication type signal quality index in the signal quality index based on the environment-device association characteristics to obtain the preferred communication type, where the first communication type signal quality index includes: the intensity, signal-to-noise ratio, and signal stability of the Bluetooth broadcast signal; the second communication type signal quality index includes: the intensity, connection success rate, data transmission rate, signal delay, and packet loss rate of the networking signal; the third communication type signal quality index includes: the intensity, tag reading success rate, and signal anti-interference ability of the radio frequency signal; the environmental parameters include: weather conditions, electromagnetic interference intensity; the device status includes: mobile terminal battery level, vehicle-mounted sensor health; Based on the preferred communication type, determine whether to send a car change permission message to the mobile terminal.

2. The shared electric vehicle return control method according to claim 1, wherein, Based on the environmental parameters and the device status, perform communication adaptability analysis on the first communication type signal quality index, the second communication type signal quality index, and the third communication type signal quality index in the signal quality index based on the environment-device association characteristics to obtain the preferred communication type, including: Extract the first communication type signal quality index, the second communication type signal quality index, and the third communication type signal quality index from the signal quality index; Perform one-hot embedding encoding on the first communication type signal quality index, the second communication type signal quality index, and the third communication type signal quality index respectively to obtain the first to third communication type signal quality index one-hot embedding encoding vectors; Perform structured mapping encoding on the environmental parameters and the device status to obtain the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector; Fuse the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector to obtain the environmental parameter-device status structured mapping joint encoding vector; Perform communication adaptability deep neural encoding and decoding on the first to third communication type signal quality index one-hot embedding encoding vectors and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain the first adaptability, the second adaptability, and the third adaptability; Take the communication type corresponding to the largest of the first fitness, the second fitness, and the third fitness as the preferred communication type.

3. The shared electric vehicle return control method according to claim 2, wherein, Perform structured mapping encoding on the environmental parameters and the device status to obtain an environmental parameter structured mapping encoding vector and a device status structured mapping encoding vector, including: performing structured mapping encoding based on a multi-layer perceptron on the environmental parameters and the device status to obtain the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector.

4. The shared electric vehicle return control method according to claim 3, wherein, Perform communication fitness depth neural encoding and decoding on the one-hot embedded encoding vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain a first fitness, a second fitness, and a third fitness, including: Perform communication fitness depth neural encoding with chained local interaction on the one-hot embedded encoding vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain multi-dimensional feature interaction response encoding vectors of the first to third communication types; Based on the multi-dimensional feature interaction response encoding vectors of the first to third communication types, obtain the first fitness, the second fitness, and the third fitness.

5. The shared electric vehicle return control method according to claim 4, wherein Perform communication fitness depth neural encoding with chained local interaction on the one-hot embedded encoding vectors of the signal quality indicators of the first to third communication types and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain multi-dimensional feature interaction response encoding vectors of the first to third communication types, including: Perform one-dimensional convolutional local implicit feature encoding on the one-hot embedded encoding vector of the signal quality indicator of the first communication type and the environmental parameter-device status structured mapping joint encoding vector respectively to obtain a set of one-hot embedded encoding vectors of the local implicit features of the signal quality indicator of the first communication type and a set of joint encoding vectors of the local implicit features of the environmental parameter-device status structured mapping; Perform monomer feature interaction on each group of corresponding one-hot embedded encoding vectors of the local implicit features of the signal quality indicator of the first communication type and joint encoding vectors of the local implicit features of the environmental parameter-device status structured mapping in the set of one-hot embedded encoding vectors of the local implicit features of the signal quality indicator of the first communication type and the set of joint encoding vectors of the local implicit features of the environmental parameter-device status structured mapping to obtain a set of multi-dimensional feature local implicit interaction response encoding vectors of the first communication type; Perform modulation chain reasoning based on attention weight interaction on the set of multi-dimensional feature local implicit interaction response encoding vectors of the first communication type to obtain a multi-dimensional feature interaction response encoding vector of the first communication type.

6. The shared electric vehicle return control method according to claim 5, characterized in that, Perform modulation chain reasoning based on attention weight interaction on the set of multi-dimensional feature local implicit interaction response encoding vectors of the first communication type to obtain a multi-dimensional feature interaction response encoding vector of the first communication type, including: Determine the chained inference attention weights of each first communication type multi-dimensional feature local implicit interaction response coding vector based on the feature distribution characteristics of each first communication type multi-dimensional feature local implicit interaction response coding vector in the set of first communication type multi-dimensional feature local implicit interaction response coding vectors, so as to obtain a set of first communication type multi-dimensional feature chained inference attention weights; Based on the set of first communication type multi-dimensional feature chained inference attention weights, perform weighted modulation on the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain a set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors; Input the set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors into a chained inference engine based on a forward LSTM model to obtain the first communication type multi-dimensional feature interaction response coding vector.

7. The shared electric vehicle return control method according to claim 6, characterized in that, Based on the set of first communication type multi-dimensional feature chained inference attention weights, performing weighted modulation on the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain a set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors includes: Perform non-linear perturbation compensation based on chained inference drive on the set of first communication type multi-dimensional feature chained inference attention weights to obtain a set of first communication type multi-dimensional feature chained inference compensation attention weights; Use the set of first communication type multi-dimensional feature chained inference compensation attention weights to perform weighted modulation on the set of first communication type multi-dimensional feature local implicit interaction response coding vectors to obtain the set of first communication type multi-dimensional feature local implicit interaction response modulation coding vectors.

8. The shared electric vehicle return control method according to claim 7, wherein Based on the first to third communication type multi-dimensional feature interaction response coding vectors, obtain the first fitness, the second fitness, and the third fitness, including: inputting the first to third communication type multi-dimensional feature interaction response coding vectors into a fitness analyzer based on a decoder respectively to obtain the first fitness, the second fitness, and the third fitness.

9. A shared electric vehicle return control system, characterized in that, Including: A signal detection and switching module, configured to switch to a comprehensive addressing mode in response to not receiving the positioning signal of the electric vehicle to be returned; A reference point setting module, configured to, in the comprehensive addressing mode, retrieve the positioning data of the mobile device bound to the electric vehicle to be returned and set it as a position reference point; A candidate area generation module, configured to obtain a candidate area based on the position reference point and one or more parking areas adjacent to the position reference point; A communication type acquisition module, configured to acquire the available communication types of the candidate area; A communication adaptation analysis module, which is used to specify a preferred communication type from the first communication type, the second communication type, and the third communication type based on signal quality indicators, environmental parameters, and device status when the available communication types in the to-be-selected area include the first communication type, the second communication type, and the third communication type. Among them, the communication adaptation analysis module is used to: based on the environmental parameters and the device status, perform communication adaptation degree analysis based on environment-device association characteristics on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type in the signal quality indicators to obtain the preferred communication type. Among them, the signal quality indicators of the first communication type include: the intensity, signal-to-noise ratio, and signal stability of the Bluetooth broadcast signal; the signal quality indicators of the second communication type include: the intensity of the networking signal, the connection success rate, the data transmission rate, the signal delay, and the packet loss rate; the signal quality indicators of the third communication type include: the intensity of the radio frequency signal, the tag reading success rate, and the signal anti-interference ability; the environmental parameters include: weather conditions, electromagnetic interference intensity; the device status includes: the battery power of the mobile terminal, the health of the vehicle-mounted sensor; A vehicle replacement decision module, which is used to determine whether to send vehicle replacement permission information to the mobile terminal based on the preferred communication type.

10. The shared electric vehicle return control system according to claim 9, characterized in that, The communication adaptation analysis module includes: A communication type signal quality indicator extraction unit, which is used to extract the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type from the signal quality indicators; A signal quality indicator encoding unit, which is used to perform one-hot embedding encoding on the signal quality indicators of the first communication type, the signal quality indicators of the second communication type, and the signal quality indicators of the third communication type respectively to obtain the one-hot embedding encoding vectors of the first to third communication type signal quality indicators; A structured mapping encoding unit, which is used to perform structured mapping encoding on the environmental parameters and the device status to obtain an environmental parameter structured mapping encoding vector and a device status structured mapping encoding vector; An environment-device joint encoding unit, which is used to fuse the environmental parameter structured mapping encoding vector and the device status structured mapping encoding vector to obtain an environment-parameter-device-status structured mapping joint encoding vector; An adaptation degree calculation unit, which is used to perform communication adaptation degree deep neural encoding and decoding on the one-hot embedding encoding vectors of the first to third communication type signal quality indicators and the environment-parameter-device-status structured mapping joint encoding vector respectively to obtain a first adaptation degree, a second adaptation degree, and a third adaptation degree; A preferred communication type determination unit, which is used to use the communication type corresponding to the maximum value among the first adaptation degree, the second adaptation degree, and the third adaptation degree as the preferred communication type.

Images