Method for virtually accessing user equipment into vehicle-mounted system in cross-system manner and navigating

Abstract

The invention provides a method for accessing user equipment to a vehicle-mounted system in a cross-system virtual manner and navigating, which comprises the following steps of: in a cross-system virtual access process, converting an operation request and a data transmission instruction of the user equipment into a format which can be identified by the vehicle-mounted system through a virtualization middleware technology, and performing encryption and integrity verification on transmission content; the compatibility and security of communication are ensured; a heartbeat detection mechanism between a vehicle-mounted system and user equipment is established, network connection quality is evaluated in real time by periodically sending probe packets and monitoring response time, and if overtime or too high packet loss rate is found, a standby link is automatically switched to ensure seamless connection and service continuity; historical load data of a vehicle-mounted system are analyzed through a machine learning algorithm, a load prediction model is constructed, and computing resources and network bandwidths are dynamically scheduled and reserved in advance according to spatial and temporal distribution characteristics of navigation requests, so that resource bottlenecks and service congestion are avoided.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for virtually connecting a user device to a vehicle-mounted system across systems and navigating the vehicle. Background Art

[0002] In the scenario where the user device virtually accesses the vehicle system across systems and uses the navigation function, there is a key technical contradiction: how to dynamically adjust and accurately match the computing resources and network bandwidth of the vehicle system according to the real-time behavior and differentiated needs of different users. The root cause of this contradiction lies in the heterogeneity between the user device and the vehicle system in terms of software and hardware architecture, operating system, network protocol, etc., and the usage habits, operating behaviors, and data transmission requirements of different users are also different and change at any time. The vehicle navigation system must take into account the real-time, smoothness and security of the service, and must achieve real-time perception, intelligent prediction and dynamic adaptation of user behavior and resource requirements, and perform continuous strategy optimization and algorithm upgrades in complex and changing application scenarios. This places extremely high demands on cross-system heterogeneous environment adaptation, user behavior data collection and analysis, resource scheduling strategy formulation and execution, and system compatibility and coordination of various components. At the same time, it is also necessary to fully consider the particularity of the vehicle scenario, accurately balance the relationship between navigation performance and driving safety, provide different levels of service quality assurance, and properly handle sudden extreme situations. These bring many technical challenges to the implementation of cross-system virtual access to vehicle navigation, which requires in-depth exploration and continuous innovation. Summary of the invention

[0003] The present invention provides a method for virtually accessing a user device to a vehicle-mounted system across systems and navigating the vehicle, which mainly includes: By collecting data indicators such as the CPU occupancy rate, memory usage, and network throughput of user devices, combined with historical operation records and application usage habits of users, and using machine learning algorithms such as support vector machines or random forests, we can model and predict user behavior patterns and obtain the user's real-time demand feature vector; Based on the user's real-time demand feature vector, a heuristic optimization algorithm such as a genetic algorithm is used to comprehensively consider factors such as the vehicle system's hardware and software performance, network conditions, and load balancing to dynamically generate the optimal computing resource and network bandwidth allocation strategy and determine the priority and resource occupancy ratio of different user requests; During the cross-system virtual access process, the user device's operation requests and data transmission instructions are converted into a format recognizable by the vehicle system through virtualization middleware technology, and the transmission content is encrypted and integrity checked to ensure the compatibility and security of communication; In view of the high real-time requirements of navigation services, edge computing technology is used to deploy navigation data cache and computing modules in the local nodes of the vehicle system, providing low-latency navigation path planning and voice guidance services through fast matching of real-time traffic information and user location; Establish a heartbeat detection mechanism between the vehicle system and the user device. By sending detection packets regularly and monitoring the response time, the network connection quality is evaluated in real time. If a timeout or high packet loss rate is found, it will automatically switch to the backup link to ensure seamless connection and service continuity. In the resource management module of the vehicle system, a dynamic adjustment mechanism for task priority is set up to adjust the computing resources and network bandwidth usage ratio of different tasks in real time according to the urgency of user requests and the quality feedback of navigation services, so as to ensure the priority response and processing of key services; The historical load data of the vehicle system is analyzed through machine learning algorithms to build a load prediction model. According to the spatiotemporal distribution characteristics of navigation requests, computing resources and network bandwidth are dynamically scheduled and reserved in advance to avoid resource bottlenecks and service congestion.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for virtually accessing a user device to a vehicle system across systems and navigating, which solves the problem of difficult service quality assurance caused by factors such as diversified user needs, limited vehicle system resources, and unstable network connection. The present invention first models the user behavior pattern through a machine learning algorithm to obtain the user's real-time demand feature vector. Then, combined with factors such as network conditions, vehicle system hardware and software performance, and load balancing, a genetic algorithm is used to dynamically generate the optimal computing resources and network bandwidth allocation strategy, and by setting a task priority dynamic adjustment mechanism, the resource allocation is adjusted in real time according to the urgency of the user's request and the navigation service quality feedback to ensure the priority of key services. At the same time, in view of the real-time requirements of navigation services, edge computing technology is adopted, and navigation data and computing modules are deployed locally to achieve low-latency navigation. In addition, cross-system virtual access is achieved through virtualization middleware technology to ensure communication compatibility and security. Finally, a heartbeat detection mechanism and a load prediction model are established to ensure the stability of network connection and the effectiveness of resource allocation. The present invention finally realizes the dynamic optimization allocation of vehicle system resources, improves the user service experience, and ensures the real-time and stability of services in key application scenarios such as navigation services. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flow chart of a method for virtually connecting a user device to a vehicle-mounted system across systems and navigating the vehicle.

[0006] Figure 2A schematic diagram of a method for virtually connecting a user device to a vehicle-mounted system across systems and navigating the vehicle in accordance with the present invention.

[0007] Figure 3 It is another schematic diagram of a method for virtually connecting a user device to a vehicle-mounted system across systems and navigating the vehicle in accordance with the present invention. DETAILED DESCRIPTION

[0008] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0009] like Figure 1-3 In this embodiment, a method for virtually connecting a user device to a vehicle-mounted system across systems and navigating the vehicle-mounted system may specifically include: S101. By collecting data indicators such as CPU occupancy, memory usage, network throughput, etc. of the user's device, combined with the user's historical operation records and application usage habits, and using machine learning algorithms such as support vector machines or random forests, model and predict user behavior patterns to obtain the user's real-time demand feature vector.

[0010] Obtain performance indicator data such as CPU occupancy, memory usage, and network throughput of user devices, and classify and store different types of performance indicator data. At the same time, by analyzing the user's historical operation records, obtain the user's application usage pattern data such as the frequency and duration of use of different applications. Fusion the obtained performance indicator data and application usage pattern data to generate a user behavior data set. Clean and preprocess the generated user behavior data set to remove noise data and outliers to ensure data quality. Perform feature engineering on the preprocessed user behavior data set to extract key features such as application usage frequency, usage duration, device performance indicators, etc., and perform feature selection and feature dimension reduction to obtain an optimized user behavior feature set. Use the logistic regression algorithm to train the optimized user behavior feature set and establish a user behavior prediction model. Use cross-validation to evaluate the model and calculate the prediction accuracy of the model. If the prediction accuracy is lower than the preset threshold (such as 80%), the decision tree algorithm is used to retrain the model to improve the prediction performance. The real-time collected user device performance indicator data and application usage data are subjected to the same data preprocessing and feature extraction process, and then input into the trained user behavior prediction model to obtain the user's real-time demand prediction results. Based on the user's real-time demand prediction results, combined with the user's historical preference data, a user demand feature vector is generated. The feature vector is transmitted to the recommendation engine based on collaborative filtering, and personalized application recommendations and services are provided to users through user similarity calculation and item similarity calculation. Continuously track user feedback on recommendation results, collect user click, download, use and other behavior data, and use them to optimize the recommendation model and improve the quality of recommendations. At the same time, the user behavior prediction model is regularly updated to adapt to changes in user needs.

[0011] Exemplarily, the CPU occupancy rate, memory usage, network throughput and other performance indicator data of the user device are obtained, and different types of performance indicator data are classified and stored. For example, the CPU occupancy rate can be stored in a time series manner, recorded once every minute; the memory usage can be divided into indicators such as used memory and available memory for storage; the network throughput can be divided into upload speed and download speed for storage. Classification storage facilitates subsequent data analysis and processing. By analyzing the user's historical operation records, the user's application usage pattern data such as the frequency and duration of use of different applications are obtained. For example, user A uses WeChat for 2 hours a day and uses it 10 times; user B uses Douyin for 1 hour a day and uses it 5 times. These data reflect the different application usage habits of users. The obtained performance indicator data and application usage pattern data are fused to generate a user behavior data set. For example, the CPU occupancy rate data, memory usage data, and network throughput data of user A are fused with their WeChat usage duration and frequency data to form a behavior data set of user A. The purpose of data fusion is to integrate data of different dimensions together to more comprehensively understand user behavior. The generated user behavior data set is cleaned and preprocessed to remove noise data and outliers to ensure data quality. For example, if the CPU usage suddenly soars to 100% at a certain moment and lasts for a short time, it can be regarded as noise data and removed. Data cleaning can improve data quality and avoid the impact of outliers on model training. Feature engineering is performed on the preprocessed user behavior data set to extract key features such as application usage frequency, usage time, device performance indicators, etc., and feature selection and feature dimension reduction are performed to obtain an optimized user behavior feature set. For example, the duration and frequency of users using WeChat can be used as features, and the average CPU usage and average network throughput can also be used as features. The purpose of feature engineering is to convert raw data into features that can be understood by the model and select the most effective feature subset. The optimized user behavior feature set is trained using the logistic regression algorithm to establish a user behavior prediction model. Logistic regression is a commonly used classification algorithm that can be used to predict whether a user will perform a certain operation. For example, it can be predicted whether a user will use an application in the next hour based on the user's characteristics. The model is evaluated using cross-validation to calculate the prediction accuracy of the model. Cross-validation can avoid overfitting and more accurately evaluate the generalization ability of the model. For example, the data set is divided into 5 parts, 4 of which are used to train the model each time, and 1 part is used to test the model. Finally, the average of the 5 test results is calculated as the prediction accuracy of the model. If the prediction accuracy is lower than the preset threshold (such as 80%), the decision tree algorithm is used to retrain the model to improve the prediction performance. The decision tree is a nonlinear classification algorithm that may be more suitable for some data sets than logistic regression. Choosing the right algorithm can improve the prediction accuracy of the model.The real-time collected user device performance indicator data and application usage data are subjected to the same data preprocessing and feature extraction process, and then input into the trained user behavior prediction model to obtain the user's real-time demand prediction results. For example, if the user's current CPU occupancy rate is high, the memory usage is also large, and the user is using a video editing application, the model may predict that the user needs more computing resources or storage space. Based on the user's real-time demand prediction results, combined with the user's historical preference data, a user demand feature vector is generated. For example, if the model predicts that the user needs more storage space, and the user has frequently used cloud storage services in the past, "need storage space" and "prefer cloud storage" can be used as part of the feature vector. The feature vector can describe the user's needs more comprehensively. The feature vector is transmitted to the recommendation engine based on collaborative filtering, and personalized application recommendations and services are provided to the user through user similarity calculation and item similarity calculation. For example, if the feature vectors of user A and user B are similar, and user B often uses a certain cloud storage application, the application can be recommended to user A. Collaborative filtering can make recommendations based on user similarity and item similarity. Continuously track user feedback on recommendation results, collect user click, download, use and other behavioral data, and use them to optimize the recommendation model and improve the quality of recommendation. For example, if users frequently click on recommended applications, it can be considered that the recommendation results are relatively accurate, and the model learning can be further strengthened. User feedback can help the model to be continuously optimized and improve the accuracy of recommendations. Regularly update the user behavior prediction model to adapt to changes in user needs. For example, users' application usage habits may change over time, so the model needs to be updated regularly to capture these changes. Regularly updating the model can ensure the effectiveness of the model.

[0012] S102. Based on the user's real-time demand feature vector, a heuristic optimization algorithm such as a genetic algorithm is used to comprehensively consider factors such as the software and hardware performance, network conditions, and load balancing of the vehicle system to dynamically generate the optimal computing resources and network bandwidth allocation strategy and determine the priority and resource occupancy ratio of different user requests.

[0013] Step 1: Obtain the user demand feature vector. Parse the user request content, extract key parameters, quantify the user demand, and form a multi-dimensional feature vector. Each dimension represents a different aspect of the user demand, such as the required computing resources, network bandwidth, latency requirements, etc.

[0014] Step 2: Evaluate the status of the vehicle system. Real-time monitoring of key indicators such as the vehicle system's CPU utilization, memory usage, available bandwidth, and number of connected users, and comprehensively evaluate the system's overall hardware and software performance and network status.

[0015] Step 3: Calculate the system load balancing index. Based on the current system status and user requests, calculate the system load balancing index. This index reflects the allocation of system resources and the degree of load balancing. If the system load balancing index is lower than the preset threshold, execute step 4; if the system load balancing index is higher than the preset threshold, execute step 5.

[0016] Step 4: Dynamically adjust the resource allocation strategy. Use the genetic algorithm to maximize the system load balancing index and search for the optimal computing resource and network bandwidth allocation strategy. The fitness function of the genetic algorithm integrates information such as the system load balancing index, user demand feature vector, and system status. Through the iterative calculation of the genetic algorithm, the optimal resource allocation strategy is obtained.

[0017] Step 5: Determine the priority of user requests. Calculate the priority of each user request based on the user demand feature vector and the current system resource availability. User requests with high priority will be given priority in resource allocation.

[0018] Step 6: Allocate resource proportions. According to the priority of user requests and the optimal resource allocation strategy, determine the computing resources and network bandwidth usage ratio for each user request.

[0019] Step 7: Send resource allocation instructions. Send computing resource and network bandwidth allocation instructions to the vehicle system to achieve dynamic resource allocation and priority processing of user requests.

[0020] For example, obtaining the user demand feature vector is to turn the user's request into data that can be understood and processed by the computer. For example, if a user wants to search for nearby restaurants on the in-vehicle map, this request can be decomposed into several key parameters: search type (restaurant), geographic location (current location), and search range (nearby). Then, these parameters can be quantified into a multidimensional feature vector, for example: [search type = 2 (assuming that the restaurant corresponds to the number 2), longitude = 120.123, latitude = 30.456, search radius = 5 kilometers]. Each dimension represents a different aspect of the user's needs so that the system can analyze and process it. The advantage of this is that the system can understand user needs more accurately and provide services that are more in line with user expectations. Evaluating the status of the in-vehicle system is like a doctor giving a physical examination to a patient. It is necessary to monitor the key indicators of the in-vehicle system in real time, such as CPU utilization, memory usage, available bandwidth, and the number of connected users. For example, the current CPU utilization is 60%, the memory usage is 40%, the available bandwidth is 10Mbps, and the number of connected users is 2. These indicators together reflect the overall health of the system, that is, the system's hardware and software performance and network status. Through these indicators, we can judge whether the system can meet the needs of users. The purpose of calculating the system load balancing index is to measure whether the allocation of system resources is reasonable. This index is like a balance scale, which calculates the load balancing degree of the system according to the current system status and user requests. Assume that the value range of the system load balancing index is 0 to 1. The closer the value is to 1, the more balanced the load is. If the current index is 0.8, it means that the system resource allocation is relatively balanced and user requests can be processed normally; if the index is lower than the preset threshold, such as 0.2, it means that the system load is unbalanced and the resource allocation strategy may need to be adjusted. Dynamically adjusting the resource allocation strategy is to make system resources more effectively utilized. When the system load balancing index is lower than the preset threshold, the resource allocation strategy needs to be adjusted. At this time, the genetic algorithm can be used to find the optimal resource allocation plan. The genetic algorithm is like an intelligent resource dispatcher. It will continuously try different resource allocation plans based on information such as the system load balancing index, user demand feature vector, and system status, and finally find the optimal plan. For example, it may allocate more computing resources to user requests with higher priority, or transfer some user requests to servers with lower load. The purpose of determining the priority of user requests is to give priority to more important user requests when resources are limited. The importance of each user request may be different, so the priority of each user request needs to be calculated based on the user demand feature vector and the current system resource availability. For example, the priority of an emergency call must be higher than that of an ordinary navigation request. Assuming that the priority is represented by 1 to 10, the larger the value, the higher the priority. Then the priority of an emergency call may be 10, while the priority of a navigation request may be 5.The purpose of allocating resource proportions is to reasonably allocate limited resources to different user requests. According to the priority of user requests and the optimal resource allocation strategy, the proportion of computing resources and network bandwidth occupied by each user request can be determined. For example, an emergency call request with a priority of 10 may occupy 80% of computing resources and 90% of available bandwidth, while a navigation request with a priority of 5 may only occupy 10% of computing resources and 5% of available bandwidth. The purpose of issuing resource allocation instructions is to put the resource allocation plan into practice. After determining the resource proportion of each user request, the system will issue computing resource and network bandwidth allocation instructions to the vehicle system, such as allocating more CPU resources to the process of processing emergency call requests and allocating higher network bandwidth to the channel for transmitting emergency call data. In this way, dynamic allocation of resources and priority processing of user requests are achieved.

[0021] S103. During the cross-system virtual access process, the user device's operation requests and data transmission instructions are converted into a format recognizable by the vehicle-mounted system through virtualization middleware technology, and the transmission content is encrypted and integrity checked to ensure the compatibility and security of communication.

[0022] According to the pre-established cross-system virtual access protocol, the operation request and data transmission instruction sent by the user device are obtained, parsed and key information is extracted. Through the virtualization middleware technology, the extracted key information is matched and converted with the preset recognizable format of the vehicle system to generate the operation request and data transmission instruction that meet the requirements of the vehicle system. The converted operation request and data transmission instruction are encrypted by the encryption algorithm to ensure the confidentiality and tamper-proof of the transmission content. On the basis of encryption, the check value of the transmission content is calculated by the integrity check algorithm, and it is sent to the vehicle system together with the encrypted transmission content. After receiving the encrypted transmission content and the check value, the vehicle system first decrypts the received content to restore the original operation request and data transmission instruction. The decrypted transmission content is checked for integrity, and the check value is recalculated and compared with the received check value to determine whether the transmission content is complete and correct. If the integrity check passes, the decrypted operation request and data transmission instruction are passed to the corresponding module of the vehicle system for processing and execution; if the integrity check fails, an error prompt is sent to the user device, requiring the operation request and data transmission instruction to be resent.

[0023] Exemplarily, the cross-system virtual access protocol is like a universal translator, which can understand the "language" of different devices and translate them into instructions that the vehicle system can understand. For example, a user uses a music app on a mobile phone to play a song, and the mobile phone sends a play instruction to the vehicle system via Bluetooth. This instruction may be in a specific data format on the mobile phone, and the cross-system virtual access protocol will convert it into an instruction format that the vehicle system can recognize, such as "play song XXX". The process of parsing and extracting key information is like extracting key content from a letter. For example, a user sends a navigation instruction "navigate to the nearest gas station" on a mobile phone, and the virtual access protocol will parse this instruction and extract key information: navigation, destination (nearest gas station). Virtualization middleware technology is like an adapter that can convert information in different formats into a unified format. For example, the navigation instruction format sent by the mobile navigation application may be different from the format required by the vehicle navigation system. The virtualization middleware will convert the mobile navigation instruction into a format that the vehicle navigation system can recognize, such as converting "navigate to the nearest gas station" into a standardized data format containing information such as destination coordinates and navigation path. In order to protect the user's information security, the converted instructions need to be encrypted. For example, the instruction "play song XXX" is encrypted using the AES encryption algorithm and converted into a string of ciphertext to prevent the instruction from being stolen or tampered with during transmission. Integrity verification is like a seal on an envelope, which is used to ensure that the information has not been tampered with during transmission. For example, a check value is calculated for the encrypted instruction "play song XXX", just like stamping an envelope. After receiving the instruction, the vehicle system will recalculate the check value and compare it with the received check value. If the two values ​​are consistent, it means that the instruction has not been tampered with. After the vehicle system receives the encrypted instruction and check value, it first needs to decrypt it, just like opening an envelope, and restore the ciphertext to the original instruction "play song XXX". After decryption, the vehicle system will perform an integrity check, just like checking whether the seal on the envelope is intact. If the check values ​​are consistent, it means that the instruction has not been tampered with during transmission, and you can safely execute this instruction. If the integrity check fails, it means that the instruction may have been tampered with or damaged during transmission. The vehicle system will send an error message to the user device and ask the user to resend the instruction, just like notifying the sender that the letter is damaged and needs to be resent. For example, if the check value calculated by the vehicle system is inconsistent with the received check value, an error message will be sent to the mobile phone, asking the navigation instruction to be resent. Finally, if the integrity check passes, the vehicle system will pass the instruction to the corresponding module for processing and execution. For example, the instruction "Play song XXX" is passed to the multimedia module to start playing the song.

[0024] S104. In view of the high real-time requirements of navigation services, edge computing technology is used to deploy navigation data cache and computing modules in the local nodes of the vehicle system, and provide low-latency navigation path planning and voice guidance services through rapid matching of real-time traffic information and user location.

[0025] The geographic location information and driving direction information of the vehicle system are obtained and used as initial input data; based on the geographic location information of the vehicle system, the road network data around the location is obtained from the navigation data cache of the local node; the real-time traffic condition information of the area, including road congestion, traffic accidents, etc., is obtained from the traffic management center through the wireless communication module; the road network data is integrated with the real-time traffic condition information to construct a dynamic road network model for subsequent path planning; the improved Dijkstra algorithm is used to search the path of the dynamic road network model to obtain multiple alternative paths from the starting point to the destination; the alternative paths are scored and sorted according to user preference settings, and the optimal path is selected by comprehensively considering factors such as path length, estimated travel time, road condition comfort, etc.; the navigation information of the optimal path is converted into voice prompt information, and real-time voice guidance services are provided to users through the vehicle audio system, and the navigation information is dynamically updated according to the driving progress.

[0026] Exemplarily, the geographic location information and driving direction information of the vehicle-mounted system are obtained and used as initial input data. For example, the vehicle-mounted system obtains through the GPS module that the current vehicle is located at 31.2297° north latitude and 121.4737° east longitude, and the driving direction is due north. This information is the basis for path planning. The reason for obtaining the driving direction is that it can help the system more accurately determine the road where the vehicle is currently located and the driving intention, and avoid misleading at complex intersections. According to the geographic location information of the vehicle-mounted system, the road network data around the location is obtained from the navigation data cache of the local node. For example, after the system obtains the current location of the vehicle, it will extract all road information within a radius of 1 km from the navigation data cache stored in the vehicle-mounted system with the current location as the center, including road name, road grade, road direction, connection relationship, speed limit, etc. The advantage of doing this is that it can reduce dependence on cloud servers and improve navigation response speed. Assuming that the vehicle is located near Shanghai People's Square, the system will extract information about surrounding roads such as Renmin Avenue and Nanjing West Road. Real-time road conditions in the area are obtained from the traffic management center through the wireless communication module, including road congestion, traffic accidents, etc. For example, through the 4G / 5G network, the vehicle system can receive real-time traffic information released by the traffic management center, such as: the east-west direction of Renmin Avenue is congested, with an average speed of 15 kilometers per hour; a traffic accident occurred on Nanjing West Road, and the estimated processing time is 30 minutes. Obtaining real-time traffic information can help the system avoid congested sections and accident sections. The road network data is integrated with the real-time traffic information to build a dynamic road network model for subsequent path planning. For example, the system combines the road information from the east to the west of Renmin Avenue with the congestion information in the real-time traffic conditions, marks it as congested, and increases the estimated travel time. The accident information on Nanjing West Road will cause the section to be temporarily marked as impassable. In this way, the dynamic road network model can reflect the real-time traffic conditions of the road. The improved Dijkstra algorithm is used to search the path of the dynamic road network model to obtain multiple alternative paths from the starting point to the destination. For example, assuming that the vehicle's destination is Shanghai Railway Station, the system will use the improved Dijkstra algorithm to search for multiple alternative routes based on the dynamic road network model, such as: passing through the Yan'an Elevated Road, passing through the North-South Elevated Road, passing through the ground road, etc. The improved Dijkstra algorithm can integrate real-time traffic information into the path search process to find a more suitable path. The alternative paths are scored and sorted according to user preferences, and the optimal path is selected by comprehensively considering factors such as path length, estimated travel time, and road comfort. For example, if the user sets the preference to "fastest arrival", the system will give priority to the path with the shortest estimated travel time; if the user sets the preference to "avoid highways", the system will give priority to paths that do not include highways.Assuming that the estimated travel time for the route through the Yan'an Elevated Road is 20 minutes, the estimated travel time for the route through the North-South Elevated Road is 25 minutes, and the estimated travel time for the route through the ground road is 30 minutes, the system will select the route through the Yan'an Elevated Road based on the user's "fastest arrival" preference. The navigation information of the optimal route is converted into voice prompt information, and real-time voice guidance services are provided to users through the vehicle audio system, and the navigation information is dynamically updated according to the driving progress. For example, the system will convert the selected route through the Yan'an Elevated Road into a series of voice prompt information, such as: "Turn right 500 meters ahead to enter the Yan'an Elevated Road", "Drive 2 kilometers on the Yan'an Elevated Road", and "Exit from the Yan'an Elevated Road Exit". At the same time, the system will dynamically update the navigation information based on the vehicle's driving progress and real-time road conditions, and re-plan the route when necessary. For example, if there is a sudden congestion on the road ahead, the system will re-plan the route and inform the user through voice prompts.

[0027] S105. Establish a heartbeat detection mechanism between the vehicle system and the user device, and evaluate the network connection quality in real time by sending detection packets at regular intervals and monitoring the response time. If a timeout or high packet loss rate is found, automatically switch to the backup link to ensure seamless connection and service continuity.

[0028] A two-way communication link is established between the vehicle system and the user device, and a heartbeat detection protocol is agreed upon to specify the detection packet format, sending frequency and timeout threshold. The vehicle system periodically sends detection packets to the user device and starts a timer to record the sending time. After receiving the detection packet, the user device immediately replies with a confirmation packet. The vehicle system receives the confirmation packet, calculates the round-trip delay, and updates the average response time. If the vehicle system sends the detection packet and still does not receive the confirmation packet after the preset timeout threshold, it is determined to be packet loss and the packet loss rate is updated. The vehicle system uses a weighted average algorithm to evaluate the current network quality in real time based on the calculated average response time and packet loss rate. If the network quality evaluation result is lower than the preset quality threshold and the duration exceeds the preset tolerance time, the main link is determined to be unavailable and the link switching process is started. When the link is switched, the vehicle system selects the best quality backup link from the backup link list, establishes a new communication connection, and resumes data transmission to ensure uninterrupted service.

[0029] Exemplarily, a two-way communication link is established between the vehicle system and the user device, and a heartbeat detection protocol is agreed upon, the detection packet format is specified as JSON format, the sending frequency is once every 5 seconds, and the timeout threshold is set to 200 milliseconds. The vehicle system periodically sends detection packets to the user device, and starts the timer at the same time to record the sending time point. After receiving the detection packet, the user device immediately replies with a confirmation packet, and the content of the confirmation packet is the receiving timestamp. The vehicle system receives the confirmation packet, calculates the round-trip delay RTT, and updates the average response time of the last 10 RTTs. If the vehicle system does not receive the confirmation packet after more than 200 milliseconds after sending the detection packet, it is determined to be packet loss, and the packet loss rate of the last 100 packets is updated. The vehicle system uses a weighted average algorithm with weights of 70% for average response time and 30% for packet loss rate, and calculates the current network quality score in real time. If the network quality score is lower than 8, and the quality evaluation results for three consecutive cycles are all lower than 8, the main link is determined to be unavailable and the link switching process is started. When the link switches, the vehicle system selects the backup link with the highest current network quality score from the three backup links, establishes a new TCP connection through Socket, and resumes data transmission. The data transmission adopts breakpoint resumption mode and continues transmission from the last interruption position to ensure uninterrupted service.

[0030] S106. In the resource management module of the vehicle system, a dynamic adjustment mechanism for task priority is set to adjust the computing resources and network bandwidth usage ratio of different tasks in real time according to the urgency of user requests and the quality feedback of navigation services, so as to ensure priority response and processing of key services.

[0031] According to the pre-established user request urgency assessment model, the urgency score of the current user request is obtained. If the urgency score exceeds the preset threshold, the request is marked as a high-priority request; real-time quality feedback data of the navigation service is obtained, and the quality assessment algorithm is used to determine whether the current navigation service quality meets the preset standard. If not, the priority of navigation service-related tasks is dynamically increased; according to the task priority sorting results, a resource allocation algorithm is used to dynamically adjust the proportion of computing resources and network bandwidth occupied by each task to ensure that high-priority tasks obtain more resources and improve the response speed of key services; in the resource allocation process, the resource prediction model is used to estimate the demand for computing resources and network bandwidth for tasks of different priorities in the future, and resources are reserved and scheduled in advance to avoid critical services being delayed due to resource constraints. The system continuously monitors the resource usage and task execution progress of the system. If the response time of a critical service exceeds the preset threshold, the resource emergency scheduling mechanism is triggered to seize some resources from other low-priority tasks to ensure the timely completion of critical services. Task execution logs and performance analysis models are established to regularly evaluate the system's resource management and task scheduling, identify problems such as low resource utilization and slow task response, and automatically adjust resource management strategies and task priority settings to achieve continuous optimization of system performance. User feedback and service quality evaluation data are collected, and the user request urgency assessment model and navigation service quality assessment algorithm are continuously optimized through machine learning algorithms to improve the accuracy of resource management and dynamic adjustment of task priorities, providing users with a more intelligent in-vehicle service experience.

[0032] Exemplarily, the system uses a decision tree-based machine learning model to evaluate the urgency of user requests. When the urgency score exceeds 8, the request is marked as high priority. The navigation service quality evaluation algorithm collects feedback data such as GPS positioning error and path planning time every 5 seconds. When the positioning error is greater than 10 meters or the planning time exceeds 5 seconds, the navigation task priority is increased by 20%. When allocating resources, the minimum heap algorithm is used to sort the task priorities, and 80% of the CPU time and bandwidth are allocated to the top 20% high-priority tasks through the weighted polling algorithm. The system uses an LSTM neural network to predict resource requirements within the next 1 minute. If the estimated resource utilization rate is higher than 90%, the reserved resources are started 5 seconds in advance. When the response time of a critical service exceeds 1 second, resource preemption is triggered, and 10% of resources are recovered from low-priority tasks. The system performs a performance evaluation every 10 minutes, uses an association rule mining algorithm to find tasks with resource utilization rates below 50%, and automatically reduces their priority. Finally, the system uses a collaborative filtering algorithm to analyze user feedback, continuously improve the urgency evaluation and service quality evaluation models, and realize intelligent in-vehicle service management.

[0033] S107. Analyze the historical load data of the vehicle system through machine learning algorithms, build a load prediction model, and dynamically schedule and reserve computing resources and network bandwidth in advance according to the spatiotemporal distribution characteristics of navigation requests to avoid resource bottlenecks and service congestion.

[0034] Obtain historical load data of the vehicle system, including CPU usage, memory occupancy, network traffic and other indicators, and construct a time series data set; use the long short-term memory neural network (LSTM) algorithm to train historical load data and establish a load prediction model to predict the load change trend in the future based on the load situation in the past; obtain the spatiotemporal distribution characteristics of navigation requests, including the time, location, frequency of requests, etc., and use the density clustering algorithm to analyze them to identify high-incidence areas and peak hours; combine the load prediction results with the request distribution characteristics to estimate the resource demand under different spatiotemporal conditions and formulate dynamic resource scheduling strategies; for high-incidence areas and peak hours, reserve sufficient computing resources and network bandwidth in advance to avoid resource bottlenecks; for sparse request areas and valley hours, release idle resources in time to improve utilization efficiency; in the process of resource scheduling, monitor the system load level in real time, and when there is a large deviation between the actual load and the predicted load, trigger the incremental training and parameter update of the prediction model to improve the prediction accuracy; for the execution effect of the scheduling strategy, collect system operation data, including resource utilization, request response time and other indicators, to evaluate the scheduling performance and optimize the scheduling algorithm, and continuously improve the service quality of the system.

[0035] For example, the historical load data of CPU usage, memory occupancy, network traffic and other indicators are collected through the vehicle system, and a time series data set is constructed at intervals of 5 minutes. The LSTM neural network model is built using the TensorFlow framework, and the number of hidden layers is set to 3, the number of hidden units is set to 128, the learning rate is set to 01, and 1000 epochs are trained. The load change trend in the next 30 minutes is predicted based on the load situation in the past hour, and the average prediction error is controlled within 10%. The spatiotemporal distribution data of navigation requests, including latitude and longitude, timestamp, etc., are obtained, and the DBSCAN density clustering algorithm is used for analysis. The radius parameter eps is set to 500 meters, and the minimum number of samples minPts is set to 50. The high-incidence areas of requests (such as commercial centers, transportation hubs) and peak hours (such as morning and evening peaks) are identified. Combining the load prediction results with the request distribution characteristics, heuristic algorithms such as simulated annealing are used to optimize the resource scheduling strategy. In areas with high request frequency and during peak hours, 20% of redundant resources are reserved in advance through the elastic scaling mechanism; in areas with sparse requests and during off-peak hours, resource allocation is dynamically reduced to control the idle rate below 5%. The system load level is monitored in real time. When the deviation between the actual load and the predicted load exceeds 20%, incremental training is triggered to fine-tune the LSTM model parameters using the latest monitoring data. System operation data such as CPU utilization and average response time are collected, key indicators such as resource utilization efficiency and request timeout rate are calculated, scheduling performance is displayed through a visual dashboard, and scheduling algorithms and parameter configurations are continuously optimized to reduce the request timeout rate to below 1%, continuously improving service quality.

[0036] What is disclosed above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for virtually accessing a user device to a vehicle system across systems and navigating, characterized in that: The method comprises: By collecting data indicators such as the CPU occupancy rate, memory usage, and network throughput of user devices, combined with historical operation records and application usage habits of users, and using machine learning algorithms such as support vector machines or random forests, we can model and predict user behavior patterns and obtain the user's real-time demand feature vector; Based on the user's real-time demand feature vector, a heuristic optimization algorithm such as a genetic algorithm is used to comprehensively consider factors such as the vehicle system's hardware and software performance, network conditions, and load balancing to dynamically generate the optimal computing resource and network bandwidth allocation strategy and determine the priority and resource occupancy ratio of different user requests; During the cross-system virtual access process, the user device's operation requests and data transmission instructions are converted into a format recognizable by the vehicle system through virtualization middleware technology, and the transmission content is encrypted and integrity checked to ensure the compatibility and security of communication; In view of the high real-time requirements of navigation services, edge computing technology is used to deploy navigation data cache and computing modules in the local nodes of the vehicle system, providing low-latency navigation path planning and voice guidance services through fast matching of real-time traffic information and user location; Establish a heartbeat detection mechanism between the vehicle system and the user device. By sending detection packets regularly and monitoring the response time, the network connection quality is evaluated in real time. If a timeout or high packet loss rate is found, it will automatically switch to the backup link to ensure seamless connection and service continuity. In the resource management module of the vehicle system, a dynamic adjustment mechanism for task priority is set up to adjust the computing resources and network bandwidth usage ratio of different tasks in real time according to the urgency of user requests and the quality feedback of navigation services, so as to ensure the priority response and processing of key services; The historical load data of the vehicle system is analyzed through machine learning algorithms to build a load prediction model. According to the spatiotemporal distribution characteristics of navigation requests, computing resources and network bandwidth are dynamically scheduled and reserved in advance to avoid resource bottlenecks and service congestion.

2. According to claim 1, a method for virtually accessing a user device to a vehicle system across systems and navigating, characterized in that: The method collects data indicators such as CPU occupancy rate, memory usage, network throughput, etc. of user devices, combines historical operation records and application usage habits of users, and uses machine learning algorithms such as support vector machines or random forests to model and predict user behavior patterns, and obtains user real-time demand feature vectors, including: Obtain performance indicator data such as CPU occupancy, memory usage, and network throughput of user devices, and classify and store different types of performance indicator data; At the same time, by analyzing the user's historical operation records, the user's application usage pattern data such as the frequency and duration of use of different applications is obtained; Fuse the acquired performance indicator data and application usage pattern data to generate a user behavior data set; Perform data cleaning and preprocessing on the generated user behavior data set to remove noise data and outliers to ensure data quality; Perform feature engineering on the preprocessed user behavior dataset to extract key features such as application usage frequency, usage duration, and device performance indicators, and perform feature selection and feature dimensionality reduction to obtain an optimized user behavior feature set; Use the logistic regression algorithm to train the optimized user behavior feature set and establish a user behavior prediction model; Use cross-validation to evaluate the model and calculate the prediction accuracy of the model; If the prediction accuracy is lower than the preset threshold, the decision tree algorithm is used to retrain the model to improve the prediction performance; The real-time collected user device performance indicator data and application usage data are subjected to the same data preprocessing and feature extraction process, and then input into the trained user behavior prediction model to obtain the user's real-time demand prediction results; Generate user demand feature vector based on user real-time demand prediction results and user historical preference data; The feature vector is transmitted to the recommendation engine based on collaborative filtering, and personalized application recommendations and services are provided to users through user similarity calculation and item similarity calculation. Continuously track user feedback on recommendation results and collect user click, download, and usage behavior data to optimize recommendation models and improve recommendation quality; At the same time, the user behavior prediction model is updated regularly to adapt to changes in user needs.

3. According to claim 1, a method for virtually accessing a user device to a vehicle system across systems and navigating, characterized in that: The method adopts a heuristic optimization algorithm such as a genetic algorithm based on the user's real-time demand feature vector, comprehensively considers factors such as the software and hardware performance, network conditions, and load balancing of the vehicle system, dynamically generates the optimal computing resources and network bandwidth allocation strategy, and determines the priority and resource occupancy ratio of different user requests, including: Obtain user demand feature vector; Parse user request content, extract key parameters, quantify user needs, and form a multi-dimensional feature vector; Each dimension represents a different aspect of user needs.

4. The method for virtually accessing a user device to a vehicle system across systems and navigating according to claim 1, characterized in that: In the cross-system virtual access process, the user device's operation request and data transmission instruction are converted into a format recognizable by the vehicle system through virtualization middleware technology, and the transmission content is encrypted and integrity checked to ensure the compatibility and security of communication, including: According to the pre-established cross-system virtual access protocol, the operation request and data transmission instruction issued by the user equipment are obtained, and the key information is extracted by parsing the request and data transmission instruction. Through virtualized middleware technology, the extracted key information is matched and converted with the preset format recognizable by the vehicle system, generating operation requests and data transmission instructions that meet the requirements of the vehicle system; The converted operation requests and data transmission instructions are encrypted using encryption algorithms to ensure the confidentiality and tamper-proof nature of the transmission content. On the basis of encryption processing, the check value of the transmission content is calculated by using the integrity check algorithm, and it is sent to the vehicle system together with the encrypted transmission content; After receiving the encrypted transmission content and check value, the vehicle system first decrypts the received content to restore the original operation request and data transmission instruction; Perform integrity check on the decrypted transmission content, and determine whether the transmission content is complete and correct by recalculating the check value and comparing it with the received check value; If the integrity check passes, the decrypted operation request and data transmission instruction are passed to the corresponding module of the vehicle system for processing and execution; If the integrity check fails, an error prompt is sent to the user equipment, requiring the operation request and data transmission instruction to be resent.

5. The method for virtually accessing a user device to a vehicle system across systems and navigating according to claim 1, characterized in that: In view of the high real-time requirements of navigation services, edge computing technology is used to deploy navigation data cache and computing modules in the local nodes of the vehicle system, and provide low-latency navigation path planning and voice guidance services through fast matching of real-time traffic information and user location, including: Obtaining geographic location information and driving direction information of the vehicle system as initial input data; According to the geographic location information of the vehicle system, the road network data around the location is obtained from the navigation data cache of the local node; Obtain real-time traffic information in the area from the traffic management center through the wireless communication module, including road congestion and traffic accidents; Integrate road network data with real-time traffic information to build a dynamic road network model for subsequent route planning; The improved Dijkstra algorithm is used to search the path of the dynamic road network model and obtain multiple alternative paths from the starting point to the destination; Score and sort the alternative routes based on user preferences, and select the best route by comprehensively considering factors such as route length, estimated travel time, and road comfort; The navigation information of the optimal route is converted into voice prompt information, providing users with real-time voice guidance services through the vehicle audio system, and dynamically updating the navigation information according to the driving progress.

6. The method for virtually accessing a user device to a vehicle system across systems and navigating according to claim 1, characterized in that: The heartbeat detection mechanism between the vehicle system and the user device is established to evaluate the network connection quality in real time by sending detection packets regularly and monitoring the response time. If a timeout or high packet loss rate is found, it will automatically switch to the backup link to ensure seamless connection and service continuity, including: A two-way communication link is established between the vehicle system and the user equipment, and a heartbeat detection protocol is agreed upon to specify the detection packet format, sending frequency and timeout threshold; The vehicle-mounted system periodically sends a detection packet to the user device and starts a timer to record the sending time point; After receiving the detection packet, the user device immediately replies with a confirmation packet. The vehicle system receives the confirmation packet, calculates the round-trip delay, and updates the average response time. If the vehicle system has not received a confirmation packet after sending a detection packet and exceeds the preset timeout threshold, it is considered as packet loss and the packet loss rate is updated; The vehicle system uses a weighted average algorithm to evaluate the current network quality in real time based on the calculated average response time and packet loss rate; If the network quality assessment result is lower than the preset quality threshold and the duration exceeds the preset tolerance time, the primary link is determined to be unavailable and the link switching process is initiated; When the link switches, the vehicle system selects the best quality backup link from the backup link list, establishes a new communication connection, and resumes data transmission to ensure uninterrupted service.

7. The method for virtually accessing a user device to a vehicle system across systems and navigating according to claim 1, characterized in that: In the resource management module of the vehicle system, a task priority dynamic adjustment mechanism is set to adjust the computing resources and network bandwidth usage ratio of different tasks in real time according to the urgency of user requests and the quality feedback of navigation services, so as to ensure the priority response and processing of key services, including: According to the pre-established user request urgency evaluation model, the urgency score of the current user request is obtained. If the urgency score exceeds a preset threshold, the request is marked as a high priority request; Obtain real-time quality feedback data of navigation services, and use quality assessment algorithms to determine whether the current navigation service quality meets the preset standards. If not, dynamically increase the priority of navigation service-related tasks; According to the task priority sorting results, a resource allocation algorithm is used to dynamically adjust the proportion of computing resources and network bandwidth occupied by each task, ensuring that high-priority tasks obtain more resources and improving the response speed of key services; In the resource allocation process, the resource prediction model is used to estimate the computing resources and network bandwidth requirements of tasks of different priorities in the future, and resources are reserved and scheduled in advance to avoid delayed response of key services due to insufficient resources. Continuously monitor the system's resource usage and task execution progress. If the response time of a critical service exceeds the preset threshold, the resource emergency scheduling mechanism is triggered to seize some resources from other low-priority tasks to ensure the timely completion of critical services. Establish task execution logs and performance analysis models, regularly evaluate the system's resource management and task scheduling, identify problems such as low resource utilization and slow task response, and automatically adjust resource management strategies and task priority settings to achieve continuous optimization of system performance; Collect user feedback and service quality evaluation data, continuously optimize the user request urgency assessment model and navigation service quality assessment algorithm through machine learning algorithms, improve the accuracy of resource management and dynamic adjustment of task priorities, and provide users with a more intelligent in-vehicle service experience.

8. The method for virtually accessing a user device to a vehicle system across systems and navigating according to claim 1, characterized in that: The historical load data of the vehicle system is analyzed by machine learning algorithms to build a load prediction model, and computing resources and network bandwidth are dynamically scheduled and reserved in advance according to the spatiotemporal distribution characteristics of navigation requests to avoid resource bottlenecks and service congestion, including: Obtain historical load data of the vehicle system, including indicators such as CPU usage, memory usage, and network traffic, and construct a time series data set; The long short-term memory neural network (LSTM) algorithm is used to train historical load data and establish a load forecasting model to predict the load change trend in the future based on the load conditions in the past period of time; Obtain the spatiotemporal distribution characteristics of navigation requests, including the time, location, and frequency of requests, and use density clustering algorithms to analyze them to identify high-frequency request areas and peak hours; Combine the load prediction results with the request distribution characteristics to estimate the resource demand under different time and space conditions and formulate dynamic resource scheduling strategies; For high-request areas and peak hours, reserve sufficient computing resources and network bandwidth in advance to avoid resource bottlenecks; For areas with sparse requests and low-volume periods, idle resources can be released in a timely manner to improve utilization efficiency; During resource scheduling, the system load level is monitored in real time. When there is a large deviation between the actual load and the predicted load, incremental training and parameter updates of the prediction model are triggered to improve prediction accuracy. Based on the execution effect of the scheduling strategy, system operation data is collected, including indicators such as resource utilization and request response time, to evaluate scheduling performance and optimize scheduling algorithms, and continuously improve the service quality of the system.

Images