Control method, system and apparatus for in-vehicle systems in a vehicle

Abstract

The embodiment of the application provides a kind of control method, system and device of vehicle-mounted system in vehicle, the method comprises: the multi-dimensional operation data of vehicle is identified and is handled, obtains the target travel data of vehicle, wherein target travel data is used to indicate the travel data of vehicle in historical time period Frequency is higher than the frequency threshold value;Call prediction model, analyze target travel data, obtain the route prediction result of vehicle, wherein prediction model is obtained by training machine learning model using target travel data sample of vehicle sample;Based on route prediction result, determine the control strategy of vehicle, wherein control strategy is used to indicate the rule of at least one vehicle-mounted system in vehicle is controlled;According to control strategy, vehicle-mounted system is controlled.The application solves the technical problem that the effectiveness of the control of vehicle-mounted system in vehicle is low.

Description

Technical Field

[0001] This application relates to the field of vehicle control, and more specifically, to a control method, system, and apparatus for an on-board system in a vehicle. Background Technology

[0002] Currently, the control of in-vehicle systems is often based on fixed rules or real-time feedback from a single trip, failing to deeply integrate the long-term personalized driving habits of passengers (e.g., drivers) and the micro-road condition characteristics of high-frequency routes. Furthermore, the control of in-vehicle systems only adjusts after the trip begins, unable to perform pre-trip energy planning and pre-adjustment of the in-vehicle system based on high-probability travel plans. Therefore, the technical problem of low effectiveness in controlling in-vehicle systems remains.

[0003] There is currently no good solution to the above problems. Summary of the Invention

[0004] This application provides a method, system, and apparatus for controlling an on-board system in a vehicle, to at least solve the technical problem of low effectiveness in controlling the on-board system in a vehicle.

[0005] According to one aspect of the embodiments of this application, a control method for an in-vehicle system in a vehicle is provided. The method may include: identifying and processing multi-dimensional operating data of the vehicle to obtain target travel data of the vehicle, wherein the target travel data represents travel data of the vehicle that occurs more frequently than a frequency threshold within a historical time period; calling a prediction model to analyze the target travel data to obtain a route prediction result of the vehicle, wherein the prediction model is obtained by training a machine learning model using target travel data samples of vehicle samples, the machine learning model is used to learn the spatiotemporal travel patterns of the driver and passengers, and based on the learning results, adjusting the vehicle's thermal management system before departure and controlling the vehicle's power system during vehicle operation; determining a control strategy for the vehicle based on the route prediction result, wherein the control strategy represents the rules for controlling at least one in-vehicle system in the vehicle; and controlling the in-vehicle system according to the control strategy.

[0006] Furthermore, the prediction model is invoked to analyze the target travel data and obtain the vehicle's route prediction result, including: in response to the vehicle meeting the triggering condition, the prediction model is invoked to analyze the target travel data and obtain the probability of the vehicle's travel route occurring; based on the probability of occurrence, the route prediction result is determined.

[0007] Furthermore, the method also includes: updating the prediction model to obtain an updated prediction model; and analyzing the target travel data according to the updated prediction model and the occurrence probability determined by the updated prediction model to obtain route prediction results.

[0008] Furthermore, the multidimensional operational data of the vehicle is identified and processed to obtain the vehicle's target travel data, including: preprocessing the multidimensional operational data to obtain the vehicle's structured travel events, wherein the travel events are used to represent the spatiotemporal characteristics of the vehicle within a single travel cycle; and identifying and processing the travel events to obtain the target travel data.

[0009] Furthermore, the multidimensional operational data is preprocessed to obtain structured travel events for vehicles, including: cleaning the multidimensional operational data to obtain cleaned multidimensional operational data; fuzzifying the cleaned multidimensional operational data to obtain fuzzified multidimensional operational data; and formatting the fuzzified multidimensional operational data to obtain structured travel events.

[0010] Furthermore, the travel events are identified and processed to obtain target travel data, including: identifying and processing the travel events to obtain multiple parking spots; grouping the parking spots according to their spatial density to obtain multiple parking spot clusters; and analyzing the multiple parking spot clusters to obtain target travel data.

[0011] Furthermore, multiple parking spot clusters are analyzed to obtain target travel data, including: analyzing multiple parking spot clusters to obtain the parking duration, frequency of occurrence, and time characteristics of vehicles in each parking spot cluster; labeling multiple parking spot clusters based on parking duration, frequency of occurrence, and time characteristics to obtain labeling results; and identifying parking spot clusters with a frequency higher than the frequency threshold in the labeling results as target travel data.

[0012] Furthermore, based on parking duration, frequency of occurrence, and time characteristics, multiple parking spot clusters are labeled to obtain labeling results, including: determining the type of each parking spot in multiple parking spot clusters based on parking duration, frequency of occurrence, and time characteristics; and labeling multiple parking spot clusters according to type to obtain labeling results.

[0013] Furthermore, based on the route prediction results, the vehicle control strategy is determined, including: based on the route prediction results and multi-dimensional operating data, determining at least one of the following strategies for the vehicle: adjustment strategy, switching strategy, and allocation strategy, wherein the adjustment strategy is used to represent the rules for adjusting the on-board system before the vehicle travels, the switching strategy is used to represent the rules for switching the operating mode of the on-board system during vehicle travel, and the allocation strategy is used to represent the rules for allocating energy to the on-board system during vehicle travel or charging.

[0014] Furthermore, according to the control strategy, the vehicle system is controlled, including: parsing the control strategy to obtain control commands for the vehicle system; and sending the control commands to the vehicle system to control the vehicle system.

[0015] Furthermore, the control strategy is parsed to obtain control commands for the vehicle system, including: parsing the control strategy to obtain the control parameters corresponding to the control strategy; converting the control parameters to obtain converted control parameters, wherein the communication protocol format of the converted control parameters is consistent with the communication protocol format of the vehicle system; and generating control commands based on the converted control parameters.

[0016] Furthermore, the vehicle system includes an air conditioning system. Control commands are sent to the vehicle system to control it, including: using the vehicle's target bus to send a start command to the air conditioning system to start the air conditioning system.

[0017] Furthermore, the vehicle system includes a battery management system, which sends control commands to the vehicle system to control the vehicle system, including sending a battery heating start command to the battery management system to heat the battery to a target temperature.

[0018] Furthermore, the vehicle system includes a power system, which sends control commands to the vehicle system to control the vehicle system, including: in response to detecting a power switching command sent to the power system, switching the operating mode of the power system from an initial mode to a target mode, and controlling the power system in the target mode, wherein the energy-saving effectiveness of the target mode is higher than that of the initial mode.

[0019] Furthermore, the method also includes: in response to the vehicle not being connected to a charging device, or the battery charge in the vehicle being below a charge threshold, controlling the vehicle's cabin system to heat the vehicle's seats.

[0020] Furthermore, the method also includes: generating vehicle prompt information based on route prediction results and control strategies, wherein the prompt information is used to display the vehicle's predicted trip and / or the vehicle's energy-saving results to at least one driver or passenger in the vehicle.

[0021] According to another aspect of the embodiments of this application, a control system for an in-vehicle system in a vehicle is also provided, comprising: a data processing module, used to identify and process multi-dimensional operating data of the vehicle to obtain target travel data of the vehicle, wherein the target travel data represents travel data of the vehicle that occurs more frequently than a frequency threshold within a historical time period; a prediction module, used to call a prediction model to analyze the target travel data and obtain a route prediction result of the vehicle, wherein the prediction model is obtained by training a machine learning model using target travel data samples of vehicle samples, the machine learning model is used to learn the spatiotemporal travel patterns of the driver and passengers, and based on the learning results, adjust the vehicle's thermal management system before departure and control the vehicle's power system during vehicle operation; a decision module, used to determine the vehicle's control strategy based on the route prediction result, wherein the control strategy represents the rules for controlling at least one in-vehicle system in the vehicle; and an execution module, used to control the in-vehicle system according to the control strategy.

[0022] Furthermore, the control system also includes an interaction and feedback module, used to generate vehicle prompts based on route prediction results and control strategies, and to process the prompts, wherein the prompts are used to display the vehicle's predicted trip and / or the vehicle's energy-saving results to at least one passenger in the vehicle.

[0023] According to another aspect of the embodiments of this application, a control device for an in-vehicle system in a vehicle is also provided, comprising: an identification unit, configured to identify and process multi-dimensional operating data of the vehicle to obtain target travel data of the vehicle, wherein the target travel data represents travel data of the vehicle that occurs more frequently than a frequency threshold within a historical time period; an analysis unit, configured to invoke a prediction model to analyze the target travel data and obtain a route prediction result of the vehicle, wherein the prediction model is obtained by training a machine learning model using target travel data samples of vehicle samples, the machine learning model is used to learn the spatiotemporal travel patterns of the driver and passengers, and based on the learning results, adjust the vehicle's thermal management system before departure and control the vehicle's power system during vehicle operation; a determination unit, configured to determine a control strategy of the vehicle based on the route prediction result, wherein the control strategy represents the rules for controlling at least one in-vehicle system in the vehicle; and a control unit, configured to control the in-vehicle system according to the control strategy.

[0024] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0025] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0026] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0027] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the methods in various embodiments of this application.

[0028] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0029] In this embodiment, travel data with a frequency exceeding a frequency threshold within a historical time period is selected from the vehicle's multidimensional operational data as target travel data, thereby accurately identifying frequently used routes and travel patterns. Subsequently, using the target travel data samples from the vehicle sample, a prediction model trained on a machine learning model is used to perform in-depth analysis of the target travel data, generating route prediction results for the vehicle. Based on these prediction results, targeted control strategies are formulated and executed by the onboard system. This overcomes the shortcomings of related technologies that employ general fixed energy-saving modes or simple real-time feedback, lacking the ability to learn from long-term user habits. It achieves a shift from passive response to proactive, forward-looking control, solving the technical problem of low control effectiveness of onboard systems and improving the technical effect of controlling onboard systems in vehicles. Attached Figure Description

[0030] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0031] Figure 1 This is a flowchart of a control method for an on-board system in a vehicle according to an embodiment of this application;

[0032] Figure 2 This is a schematic diagram of a common route analysis and adaptive energy-saving optimization system based on vehicle-side machine learning according to an embodiment of this application;

[0033] Figure 3This is a flowchart of a common route analysis and adaptive energy-saving optimization method based on vehicle-side machine learning according to an embodiment of this application;

[0034] Figure 4 This is a flowchart of a vehicle-side machine learning method for recognizing common routes according to an embodiment of this application;

[0035] Figure 5 This is a schematic diagram of a control system for an in-vehicle system according to an embodiment of this application;

[0036] Figure 6 This is a schematic diagram of a control device for an in-vehicle system according to an embodiment of this application. Detailed Implementation

[0037] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0038] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0039] According to an embodiment of this application, an embodiment of a control method for an in-vehicle system in a vehicle is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0040] This embodiment provides a control method for an on-board system in a vehicle. Figure 1 This is a flowchart of a control method for an on-board system in a vehicle according to an embodiment of this application, such as... Figure 1 As shown, the process includes the following steps.

[0041] Step S102: Identify and process the multi-dimensional operation data of the vehicle to obtain the vehicle's target travel data.

[0042] In the technical solution provided by step S102 of this application, the target travel data can be used to represent travel data in which the frequency of a vehicle is higher than a frequency threshold within a historical time period.

[0043] In this embodiment, the multidimensional operational data may include Global Positioning System (GPS) coordinates, timestamps, vehicle speed, acceleration, battery status, etc., for illustrative purposes only and without specific limitations. The raw multidimensional operational data can be cleaned to remove noise and outliers. Subsequently, the geographic location information in the multidimensional operational data is anonymized and obfuscated. For example, precise GPS coordinates can be converted into a geofenced area with a radius of 200 meters to protect user privacy. Next, continuous multidimensional vehicle operational data can be segmented into independent travel events. Each travel event can include structured features such as departure time, departure location, arrival location, travel route, and trip duration.

[0044] Optionally, after obtaining travel events, vehicle-side computing power can be utilized while the vehicle is idle or in a low-power state to employ unsupervised learning algorithms, such as the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, to perform cluster analysis on parking points (e.g., engine shutdown points) in historical travel events. Based on the spatial density of parking points, the clustering algorithm can automatically identify frequently occurring parking areas and mark them as internal logical identifiers (e.g., "home," "workplace," "frequently visited shopping areas," etc.).

[0045] Optionally, based on the identified high-frequency parking areas, i.e., the marked frequent stops, the statistical characteristics of travel events can be further analyzed. These statistical characteristics can include the frequency of occurrence at each frequent stop, the distribution of stay duration, and travel preferences during specific time periods (e.g., weekday mornings, weekend afternoons). Travel patterns that occur more frequently than a preset frequency threshold within a historical time period can be filtered out (e.g., a route appearing more than 80% of the time in the past month). Finally, the results obtained from the above statistics and filtering are integrated to obtain the target travel data.

[0046] Optionally, the aforementioned target travel data not only includes frequently occurring physical location labels, but also implicitly contains a probability model of departing from one location and arriving at another location at a specific time, that is, establishing a mapping relationship of "departure location-time-destination".

[0047] In this embodiment, the above steps automatically mine and extract long-term, stable, and high-frequency travel patterns of car owners from massive, chaotic, and multi-dimensional operational data. By identifying travel data where the frequency of a vehicle exceeds a frequency threshold within a historical time period, it is possible to distinguish between occasional and habitual travel, thereby establishing a personalized travel profile unique to the car owner. This provides an accurate and reliable input basis for subsequent route prediction and control strategy formulation, solving the problem that related technologies cannot actively learn users' long-term habits.

[0048] Step S104: Call the prediction model to analyze the target travel data and obtain the vehicle route prediction results.

[0049] In the technical solution provided in step S104 of this application, the prediction model can be obtained by training a machine learning model using target travel data samples of vehicle samples. The machine learning model is used to learn the spatiotemporal travel patterns of the driving and riding objects. Based on the learning results, the vehicle's thermal management system is adjusted before departure, and the vehicle's power system is controlled during vehicle operation.

[0050] In this embodiment, when the vehicle is powered on, unlocked, or detected as about to depart, a prediction model can be invoked to analyze the target travel data and obtain the vehicle's route prediction result.

[0051] Optionally, the predictive model can employ lightweight machine learning algorithms, such as Long Short-Term Memory (LSTM), Bayesian networks, or classification tree models, to infer from the input data. The predictive model can look for historical high-frequency patterns based on the current departure location and time. For example, if the current time is 7:30 AM on a weekday, and the location is "home," the predictive model, through historical data statistics, can calculate that the probability of going to "workplace" is 92%, the probability of going to "gym" is 5%, and the probability of other locations is 3%.

[0052] Optionally, in addition to the destination, the predictive model can also predict parameters such as the expected travel time, the expected travel route (e.g., a sequence of road segments ABC), and potential congestion. These parameters can be estimated based on the average performance of the same route, time period, and weather conditions in the past.

[0053] Optionally, the prediction model can output a list containing multiple possible routes and their corresponding probabilities, and select the route with the highest probability as the final route prediction result.

[0054] Optionally, the prediction model is not static. After each trip, the prediction results can be compared with the actual driving results (actual destination, actual time, actual energy consumption). If there is a discrepancy, the actual data can be used as a new sample to incrementally train or fine-tune the prediction model in the vehicle backend, so that the prediction model can gradually adapt to changes in the driver's habits (such as route changes and adjustments to rest).

[0055] In this embodiment of the application, the predictive model is invoked to analyze the target travel data, which can predict the user's travel intention in advance, thereby changing vehicle control from passive response to active prediction, and providing a time window and decision basis for the subsequent formulation of targeted control strategies.

[0056] Step S106: Determine the vehicle control strategy based on the route prediction results.

[0057] In the technical solution provided in step S106 of this application, the control strategy can be used to represent the rules for controlling at least one on-board system in the vehicle.

[0058] In this embodiment, the route prediction results can be mapped to the real-time vehicle status. For example, if the route prediction result is "Departing from home to the company at 7:30 am on a weekday, the entire journey is 15 kilometers, including long downhill sections, with an ambient temperature of 5°C," and the vehicle is currently plugged in for charging with a battery temperature of 10°C, this can be used to identify a typical winter commuting scenario.

[0059] Optionally, based on the predicted departure time and ambient temperature, an energy consumption model can be calculated to ensure the vehicle's cabin reaches a comfortable temperature. If the ambient temperature indicates that the current environment is winter, the control strategy could be to activate a Positive Temperature Coefficient (PTC) heater or heat pump at a specific time window before departure (e.g., 15 minutes in advance) to preheat the cabin to the user-set 22°C using grid power, and control the airflow and circulation mode.

[0060] Optionally, for long-distance journeys in low-temperature environments, it can be determined that the battery's internal resistance is relatively large and affects efficiency. Therefore, the control strategy can be to use the charging current to gently heat the battery before departure, raising the battery temperature to the lower limit of the optimal operating range (e.g., 15°C) to reduce internal resistance, improve discharge efficiency, and protect battery life.

[0061] Optionally, during the driving phase, an economy route mode command can be generated. This command may include torque limiting, enhanced regenerative braking, and road condition prediction control. Torque limiting can restrict the maximum output torque of the motor, making acceleration smoother. Enhanced regenerative braking can increase the intensity of regenerative braking energy recovery. Road condition prediction control can reduce drive torque in advance for known long downhill sections in the predicted path, utilizing inertial coasting and maximizing energy recovery. In addition, for known congested road sections, the power output characteristics can be adjusted in advance.

[0062] Optionally, if multiple control strategies conflict, for example, if the vehicle's predicted range is short and the vehicle's state of charge (SOC) is sufficient, there is no need to preheat the vehicle. Instead, a preferred control strategy can be selected based on preset priorities (e.g., comfort priority or energy saving priority) to ultimately determine a better, executable control strategy.

[0063] In this embodiment, based on the route prediction results, a targeted vehicle control strategy is determined, thereby improving the vehicle's energy efficiency and driving comfort.

[0064] Step S108: Control the vehicle system according to the control strategy.

[0065] In the technical solution provided by step S108 of this application, the abstract control strategy can be transformed into specific hardware control instructions and sent to the corresponding vehicle system for execution.

[0066] In this embodiment, the control strategy can be parsed to obtain control commands for different vehicle systems. Simultaneously, these control commands are converted into standard communication protocol formats recognizable by the controllers of each vehicle system, such as Controller Area Network (CAN) bus messages and vehicle Ethernet signals, thereby ensuring accurate transmission of the control commands.

[0067] Optionally, the in-vehicle system can be an air conditioning system, a battery management system, or a power system; these are just examples and no specific limitations are made here.

[0068] Optionally, when the pre-adjustment time window set in the control strategy is reached, the vehicle controller can send a control command (e.g., a start command) to the air conditioning system controller via the CAN bus. This control command may include the target temperature (e.g., 22°C), airflow, circulation mode (e.g., internal / external circulation), and heater / compressor power settings. The air conditioning system can operate according to the set parameters and provide real-time feedback on cabin temperature, set temperature, and other status information to the vehicle controller for dynamic fine-tuning.

[0069] Optionally, the vehicle controller can send control commands (such as a battery heating start command) to the battery management system. For winter preheating, the battery management system can control the battery heating film or liquid thermal system to operate, using the charging pile power or vehicle power supply to heat the battery to the target temperature (such as 15°C); for summer cooling, the battery cooling system is activated.

[0070] Optionally, when the vehicle starts driving and the GPS trajectory matches the route prediction result more than a threshold (e.g., 95%), the vehicle controller sends a control command (e.g., a power switching command) to the power system to switch the power system from the initial mode (e.g., standard mode) to the target mode (e.g., economic route mode).

[0071] Optionally, during the execution of the control strategy, the current execution status (e.g., "Air conditioning is preheating", "Battery is heating", "Switched to energy-saving mode") can be displayed to the user through the vehicle's infotainment screen or an application (APP). Simultaneously, user interaction data such as whether they manually intervene (e.g., manually turning off the air conditioning or adjusting the temperature) can be recorded. This interaction data can serve as feedback information for optimizing the predictive model.

[0072] In the embodiments of this application, by precisely executing the control commands of the air conditioning system, battery management system and power system, forward-looking energy management and driving optimization can be achieved, ultimately achieving the technical goals of reducing energy consumption, increasing range and improving comfort.

[0073] Through steps S102 to S108, travel data with a frequency exceeding a frequency threshold within a historical time period is selected from the vehicle's multi-dimensional operational data as target travel data, thereby accurately identifying frequently used routes and travel patterns. Then, using the target travel data samples from the vehicle samples, a prediction model trained on a machine learning model is used to perform in-depth analysis of the target travel data, generating route prediction results for the vehicle. Based on these prediction results, targeted control strategies are formulated and executed by the onboard system. This overcomes the shortcomings of related technologies that employ general fixed energy-saving modes or simple real-time feedback, lacking the ability to learn from long-term user habits. It achieves the goal of shifting from passive response to proactive, forward-looking control, solving the technical problem of low control effectiveness of onboard systems in vehicles, and realizing the technical effect of improving the control effectiveness of onboard systems in vehicles.

[0074] The above-mentioned method of this application will be further described below.

[0075] As an optional implementation, step S104 involves calling a prediction model to analyze the target travel data and obtain the vehicle's route prediction result, including: in response to the vehicle meeting the triggering condition, calling a prediction model to analyze the target travel data and obtain the probability of the vehicle's travel route occurring; and determining the route prediction result based on the probability of occurrence.

[0076] In this embodiment, the prediction model can be activated when a vehicle is detected to meet preset triggering conditions. These triggering conditions may include, but are not limited to, time-triggered, location-triggered, state-triggered, and combined triggering.

[0077] Optionally, time-triggered events can be triggered if the current time falls within a user's historical high-frequency departure time window (e.g., 7:00-8:00 AM on a weekday). Location-triggered events can be triggered if the vehicle's current location is within an identified high-frequency, frequently visited location (e.g., within a geofence of "home" or "workplace"). Status-triggered events can be triggered if the vehicle switches from "off / sleep" to "powered / started," or if actions such as unlocking the vehicle or plugging in the charging gun are detected. Combined triggers can be triggered if the location is within geofence 1, the time is a weekday morning, and the vehicle is powered on.

[0078] Optionally, if the vehicle meets the triggering conditions, a predictive model can be invoked to analyze the target travel data. For example, the predictive model takes the current time, vehicle location (departure location label), vehicle status (e.g., whether a gun is plugged in), and user historical behavior characteristics as input to the target travel data. Through internal algorithms (e.g., Bayesian probability calculation, LSTM sequence prediction), it calculates the probability of traveling to various potential destinations (e.g., "workplace," "frequently visited supermarket," "home"). For instance, the output might be: a 92% probability of traveling to the "workplace," a 6% probability of traveling to the "gym," and a 2% probability of other destinations.

[0079] Optionally, the output set of occurrence probabilities can be sorted and filtered. For example, the trip with the highest occurrence probability that exceeds a preset confidence threshold (e.g., 80%) can be selected as the final route prediction result. If the highest occurrence probability does not exceed the confidence threshold, it can be determined as an unknown trip or no prediction can be performed to conserve computing power or avoid misjudgment. The determined route prediction result includes not only the destination but also the expected path, expected duration, and road condition characteristics corresponding to that destination.

[0080] In this embodiment, by setting explicit triggering conditions, the prediction model runs at the necessary and most likely effective times, avoiding invalid calculations for non-travel periods or irrelevant scenarios. Simultaneously, the probability-based decision-making mechanism provides quantitative evidence for the prediction results, improving the reliability and controllability of the predictions and offering a reliable foundation for the formulation of subsequent control strategies.

[0081] As an optional implementation, the method further includes: updating the prediction model to obtain an updated prediction model; and analyzing the target travel data according to the updated prediction model and the occurrence probability determined by the updated prediction model to obtain route prediction results.

[0082] In this embodiment, after each trip, the actual operating data of the trip can be automatically collected, including: the actual destination reached, the actual driving route, the actual trip duration, the actual energy consumption, and whether the user has manually intervened in the control strategy (e.g., canceling the prediction or modifying the temperature setting).

[0083] Optionally, by comparing the above-mentioned actual operational data with the route prediction results, new training sample pairs containing "input features (departure location, time, etc.)" and "real labels (actual destination, path, etc.)" can be constructed.

[0084] Optionally, based on newly added training sample pairs, the prediction model can be updated at fixed time intervals (e.g., weekly) or at fixed mileage intervals (e.g., every 500 kilometers). For example, when a significant change in user habits is detected (e.g., multiple consecutive deviations from the predicted destination, or the emergence of new high-frequency locations), an incremental update is immediately triggered; offline training is performed using idle computing power during vehicle charging or long-term parking to obtain an updated prediction model. Then, based on the updated prediction model and the probability of occurrence determined by the updated prediction model, the target travel data is analyzed to obtain route prediction results that better reflect the user's current real-life habits.

[0085] In this embodiment, the above steps address the issue of dynamically changing user travel habits. By establishing a closed loop of "prediction-execution-feedback-update," the system can continuously adapt to changes in user behavior patterns over time (e.g., moving, changing jobs, seasonal travel differences), ensuring the prediction model maintains high accuracy over the long term and preventing it from failing due to outdated data.

[0086] As an optional implementation, step S102 involves identifying and processing the multidimensional operating data of the vehicle to obtain the vehicle's target travel data, including: preprocessing the multidimensional operating data to obtain structured travel events of the vehicle, wherein the travel events are used to represent the spatiotemporal characteristics of the vehicle within a single travel cycle; and identifying and processing the travel events to obtain the target travel data.

[0087] In this embodiment, the collected multidimensional operational data, such as high-frequency sampled GPS coordinates, CAN bus data such as vehicle speed / acceleration, timestamps, and environmental sensor data, can be preprocessed, for example, by cleaning and standardization.

[0088] Preprocessing multidimensional operational data can remove outliers caused by GPS signal drift, minor displacement fluctuations when the vehicle is stationary, and invalid data generated by sensor malfunctions. Data from different sampling frequencies (e.g., 1Hz GPS and 10Hz vehicle status data) can be time-aligned to maintain consistency in spatiotemporal characteristics. Precise latitude and longitude coordinates can be blurred, for example, mapping them to a geofenced area of ​​a specific radius (e.g., 200 meters) or converting them into relative location identifiers, reducing the possibility of raw trajectory data leaking user privacy.

[0089] Optionally, based on the preprocessed multidimensional operational data, driving cycles can be segmented by detecting vehicle status (e.g., ignition / shutdown signals, vehicle speed returning to zero for a duration exceeding a threshold). Each complete driving cycle can be defined as a travel event.

[0090] Optionally, for each travel event, key spatiotemporal features can be extracted to form a structured record. These key spatiotemporal features may include: departure time, arrival time, departure point geofence identifier (ID), destination geofence ID, travel trajectory point sequence (compressed), total mileage, average speed, maximum speed, energy consumption data, etc. Subsequently, these key spatiotemporal features can be encapsulated into a unified data object format to form a structured travel event, transforming unstructured or semi-structured raw data into logical units that are easily understood and processed by machines.

[0091] Optionally, after obtaining the structured travel events of the vehicle, the travel events can be identified and processed to obtain target travel data, which represents the user's stable and predictable travel habits, rather than occasional temporary travel.

[0092] In this embodiment of the application, data noise is removed and privacy is protected through preprocessing and identification processing, and representative travel patterns of users are extracted, providing high-quality, high signal-to-noise ratio data input for subsequent prediction model training.

[0093] As an optional implementation method, the multidimensional operational data is preprocessed to obtain structured travel events for vehicles, including: cleaning the multidimensional operational data to obtain cleaned multidimensional operational data; fuzzifying the cleaned multidimensional operational data to obtain fuzzified multidimensional operational data; and formatting the fuzzified multidimensional operational data to obtain structured travel events.

[0094] In this embodiment, the collected multidimensional operational data, such as GPS latitude and longitude, timestamps, vehicle speed, acceleration, and battery SOC, can be cleaned, for example, by removing outliers, filtering noise, and filling in missing values.

[0095] Optionally, outlier removal can utilize statistical methods (e.g., the 3σ principle) or physical constraints to eliminate abnormal data that clearly violates physical laws. For example, removing data where the vehicle speed exceeds its physical limits, GPS coordinates jump instantaneously to points thousands of miles away, or timestamps are out of order. Noise filtering can correct GPS drift data in a stationary state (vehicle off but coordinates slightly changed) by setting a stationary threshold (e.g., vehicle speed below 1 km / h for more than T seconds) to revert to the same coordinate point; for sensor signal jitter, smoothing can be achieved using moving average filtering or median filtering. Missing value imputation can fill in a small number of missing data points caused by signal loss using interpolation of preceding and following data or prediction based on surrounding points, thus ensuring the integrity of the data sequence.

[0096] Optionally, based on the cleaned multidimensional operational data, obfuscation processing can be performed, such as privacy-de-identifying information containing geographic locations. This de-identification processing can include geofencing mapping, precision reduction, and de-identification.

[0097] Optionally, geofencing mapping can map precise GPS latitude and longitude coordinates to predefined geofence areas. For example, high-frequency locations such as residential areas, office areas, and shopping malls can be divided into specific geofence IDs. Precision reduction can involve reducing the precision of trajectory points in non-core areas, such as reducing coordinate precision from meters to hundreds of meters. De-identification can involve removing or encrypting unique identifiers directly related to user identity, enabling the data to be used for internal vehicle logic analysis and reducing the possibility of reverse tracing back to a specific individual.

[0098] Optionally, the data, after cleaning and fuzzing, can be logically segmented and reorganized based on vehicle status changes (start / stop) to form structured travel events. For example, trip segmentation can be performed by detecting vehicle ignition and shutdown signals, defining the continuous data between start and stop as a driving cycle. Then, feature encapsulation can be performed. For instance, key features can be extracted for each driving cycle to construct a standardized data object. This object can include: time features (departure timestamp, arrival timestamp, trip duration, day of the week); spatial features (departure geofence ID, arrival geofence ID, feature point sequence of the driving path); and state features (average vehicle speed, maximum vehicle speed, total mileage, battery SOC change, average ambient temperature).

[0099] Optionally, the multidimensional operational data after fuzzification can be formatted, for example, by encapsulating standardized data objects into data records in a unified format according to a predefined pattern to form structured travel events.

[0100] In this embodiment, data cleaning reduces noise and errors in raw sensor data, improving the authenticity of the analysis. Obfuscation allows for the utilization of location data while strictly adhering to privacy regulations, reducing user concerns about location disclosure. Formatting transforms unstructured or semi-structured raw signals into structured data with clear semantics, providing standardized input for subsequent clustering analysis, pattern recognition, and machine learning model training.

[0101] As an optional implementation method, the travel event is identified and processed to obtain target travel data, including: identifying and processing the travel event to obtain multiple parking points; grouping the parking points according to their spatial density to obtain multiple parking point clusters; and analyzing the multiple parking point clusters to obtain target travel data.

[0102] In this embodiment, structured travel events can be traversed to identify the end point of each trip (i.e., the point where the vehicle is turned off or stationary for a long time).

[0103] Optionally, when a vehicle speed drops to 0 and the duration exceeds a preset threshold (e.g., 15 minutes), or when an engine shutdown signal is detected, the current geofence ID (or blurred coordinates) and timestamp can be recorded as a parking point.

[0104] Optionally, after extracting the spatial features (geofence ID or fuzzy coordinates) and temporal features (specific time and day of the week) of the parking spot, unsupervised learning algorithms, such as the DBSCAN algorithm, can be used to cluster the spatial locations of historical parking spots.

[0105] Optionally, the DBSCAN algorithm can group parking spots based on their spatial density. Parking spots that are spatially close and have a high density will be grouped into the same cluster. For example, parking spots in residential areas during weekday nights can be grouped into one cluster; parking spots in office areas during weekday daytimes can be grouped into another cluster.

[0106] Optionally, sparsely distributed parking spots that are not connected to high-density areas (such as remote scenic spots visited occasionally or temporary parking spots) are identified as noise points and excluded, thereby obtaining multiple parking spot clusters, each cluster representing a physical spatial clustering area.

[0107] Optionally, semantic analysis and statistics can be performed on each cluster generated by the clustering to transform each cluster into target travel data with business meaning.

[0108] For example, if parking spots in a cluster are mainly concentrated between 10:00 PM and 7:00 AM the next day, and are distributed on weekends and weekdays, they are identified as "Home"; if parking spots in a cluster are mainly concentrated between 9:00 AM and 6:00 PM, and are distributed on weekdays, they are identified as "Workplace". Other clusters can be labeled as "Frequently Visited Location A", "Frequently Visited Location B", etc.

[0109] In this embodiment, cluster analysis automatically identifies core nodes in a user's life (home, workplace, frequently visited locations) and establishes a probabilistic mapping relationship of "location-time-destination" through statistical analysis over time. This provides high signal-to-noise ratio and high semantic value input data for subsequent trip prediction, avoiding noise interference and privacy risks associated with directly processing raw coordinates.

[0110] As an optional implementation method, multiple parking spot clusters are analyzed to obtain target travel data, including: analyzing multiple parking spot clusters to obtain the parking duration, frequency of occurrence, and time characteristics of vehicles in multiple parking spot clusters at each parking spot; marking multiple parking spot clusters based on parking duration, frequency of occurrence, and time characteristics to obtain marking results; and identifying parking spot clusters with a frequency higher than a frequency threshold in the marking results as target travel data.

[0111] In this embodiment, for each generated parking spot cluster, the historical parking records contained therein can be analyzed in depth to extract key statistical features such as parking duration, frequency of occurrence, and time characteristics.

[0112] Optionally, parking duration statistics can calculate the distribution of dwell time at each parking spot within the cluster, including average dwell time, maximum / minimum dwell time, and standard deviation of dwell time. For example, the cluster corresponding to "home" has a longer nighttime dwell time (e.g., 8-10 hours) and a shorter daytime dwell time; while "office" has a relatively fixed long daytime dwell time.

[0113] Optionally, frequency statistics can be used to count the total number of times the cluster appears as a origin or destination within a specific time window (e.g., the past 30 days, the past 90 days), as well as the frequency of occurrence per unit time. Locations that appear frequently are more likely to represent a user's core social circle.

[0114] Optionally, temporal feature analysis can analyze the specific time period patterns of parking spots within the cluster, which may include: intraday distribution, such as whether it is concentrated during the daytime (9:00-18:00) or at night (22:00-7:00), or during morning and evening rush hours; weekly distribution, such as whether it occurs only on weekdays (Monday to Friday), or on weekends, or occurs irregularly; and periodicity, such as whether there are obvious seasonal or monthly periodic patterns.

[0115] Optionally, using a pre-defined rule engine or lightweight classification model, combined with the aforementioned statistical features, semantic labels can be assigned to each cluster. If a cluster's average nighttime stay is >6 hours and its frequency of occurrence is high on both weekday nights and weekends, it can be labeled "Home". If a cluster's average daytime stay is >6 hours and it only appears during weekday days, it can be labeled "Workplace". If a cluster has a short stay (e.g., 0.5-1 hour), a high frequency of occurrence but scattered throughout the day, or is concentrated on weekends, it can be labeled "Frequently Visited Location A" or "Frequently Visited Location B" (e.g., supermarket, gym, friend's house). If a cluster has an extremely short and irregular stay, it can be labeled "Temporary Parking" or left unlabeled. After obtaining the labeling results, parking spot clusters with a frequency higher than a frequency threshold in the labeling results can be identified as target travel data.

[0116] In this embodiment, multi-dimensional feature analysis (parking duration, frequency of occurrence, and time characteristics) can accurately identify key nodes in a user's life (home, office, etc.) and assign them semantic labels. Simultaneously, frequency threshold filtering eliminates sporadic, low-value, or noisy data, ensuring the data input to the prediction model is highly stable and representative, thereby improving the accuracy and reliability of subsequent trip predictions.

[0117] As an optional implementation method, multiple parking spot clusters are marked based on parking duration, frequency of occurrence, and time characteristics to obtain marking results, including: determining the type of each parking spot in multiple parking spot clusters based on parking duration, frequency of occurrence, and time characteristics; and marking multiple parking spot clusters according to type to obtain marking results.

[0118] In this embodiment, statistical features of each parking spot cluster can be extracted to construct a multi-dimensional feature vector, including: parking duration (e.g., average parking duration), frequency of occurrence, and time features.

[0119] Optionally, the average parking duration is the average and distribution range of the dwell time at parking spots within each parking spot cluster. The frequency of occurrence is the number of times a parking spot cluster appears within a specific time window (e.g., the past 30 days), as well as the temporal density (e.g., the probability of occurrence per hour / day). Temporal characteristics may include the main active time periods (e.g., 22:00-07:00 at night, 09:00-18:00 during the day), weekday distribution (weekdays vs. weekends), and seasonal patterns.

[0120] Optionally, based on the above statistical characteristics, the type of parking spot can be determined according to preset expert rules or a lightweight classification model. For example, if a parking spot has significant characteristics of long nighttime stays (e.g., the duration of a single nighttime stay is usually greater than 8 hours), appears frequently on both weekday nights and weekend nights, and is active mainly from late night to the next morning, the spot can be labeled as "home".

[0121] Optionally, it exhibits significant daytime long-stay characteristics (e.g., single daytime stays between 6 and 9 hours). It appears frequently during weekday days, but with extremely low or zero frequency on weekends and holidays. Active time is mainly concentrated between 9 AM and 6 PM on weekdays, which can confirm the labeling result as a "workplace".

[0122] Optionally, locations with shorter stays (e.g., between 0.5 and 2 hours) that align with shopping, dining, or social activities can be identified. Locations with high frequency of occurrence, but with a more dispersed time distribution, or concentrated during specific non-working hours (e.g., weekend afternoons), or those without a fixed diurnal pattern or exhibiting specific social / entertainment time characteristics, can be identified as "frequently visited locations."

[0123] In this embodiment, by using rules based on duration, frequency, and time characteristics, core scenarios in a user's life (such as residence, workplace, and leisure location) can be automatically identified and assigned clear semantic tags. This reduces reliance on precise geographical location and protects user privacy.

[0124] As an optional implementation, step S106, based on the route prediction results, determines the vehicle control strategy, including: based on the route prediction results and multi-dimensional operating data, determining at least one of the following strategies for the vehicle: adjustment strategy, switching strategy, and allocation strategy, wherein the adjustment strategy is used to represent the rules for adjusting the on-board system before the vehicle travels, the switching strategy is used to represent the rules for switching the operating mode of the on-board system during the vehicle travels, and the allocation strategy is used to represent the rules for allocating energy to the on-board system during the vehicle travels or charging.

[0125] In this embodiment, based on route prediction results (e.g., destination, expected path, expected duration, road condition characteristics) and current multi-dimensional operational data (e.g., battery SOC, ambient temperature, whether the gun is plugged in, user preferences), a comprehensive decision is made to generate a comprehensive control scheme that includes "adjustment strategy", "switching strategy" and "allocation strategy".

[0126] Optionally, the adjustment strategy can be an air conditioning adjustment strategy. If the route prediction result indicates that the user is about to depart (e.g., the predicted departure time is within 15 minutes), and the current vehicle status meets specific conditions (e.g., connected to a charging station, or sufficient SOC, or extreme ambient temperature), then the air conditioning adjustment strategy is activated. For example, based on the target temperature (user preference or default value), the current ambient temperature, and the vehicle's thermal model, the energy required to reach a comfortable temperature and the advance start time are calculated. If the vehicle is connected to a charging station, a strategy of using grid power to preheat the cabin is prioritized to save battery power; if the charging station is not plugged in, a strategy of using the remaining battery power for rapid pre-cooling is adopted.

[0127] Optionally, the adjustment strategy can be a battery thermal management adjustment strategy. If the route prediction results show that the trip is a long distance or a high-speed trip, and the current battery temperature is below the optimal operating range (e.g., below 15°C in winter), a battery thermal management adjustment strategy is generated to heat the battery to the optimal temperature using charging energy or idle time before driving, thereby reducing internal resistance and improving discharge efficiency.

[0128] Optionally, the switching strategy can be a driving mode switching strategy. If the route prediction results show that the road conditions are characterized by a large number of traffic lights, long downhill slopes, and congested sections, rules can be generated to switch to "economy mode" or "customized energy-saving mode". For example, before a known congested section, the motor torque response sensitivity can be reduced in advance to avoid rapid acceleration; before a known downhill section, the energy recovery intensity can be increased in advance.

[0129] Optionally, the allocation strategy can be a dynamic allocation strategy based on the state of energy (SOC). Energy usage priorities can be planned based on the predicted remaining distance of the trip and the current SOC. During charging, if the predicted trip is short and the vehicle is charging, a strategy of "maintaining battery temperature only" or "minor energy replenishment" can be implemented to avoid overcharging and prioritize grid power for air conditioning pre-conditioning.

[0130] Optionally, during driving, a "power conservation strategy" can be implemented if the predicted journey is long. This strategy involves reserving more power in non-critical road sections (e.g., flat roads) for pure electric driving in subsequent critical road sections (e.g., congested urban areas or uphill sections). For plug-in hybrid vehicles, a "forced range extension" or "pure electric priority" strategy can be generated to match the optimal energy efficiency solution for different road segments in the predicted path.

[0131] Optionally, in the energy consumption allocation of accessories, the power budget of accessories such as air conditioning, seat heating, and audio can be dynamically allocated according to the predicted duration of the trip, so that the SOC is within a safe range when the destination is reached.

[0132] Optionally, the aforementioned adjustment, switching, and allocation strategies can be integrated into a complete vehicle control strategy package and distributed to each onboard system for execution via the vehicle controller. If multiple strategies conflict (e.g., battery preheating and air conditioning pre-adjustment operating at high power simultaneously, causing the total power to exceed the charging pile limit), coordination and peak shaving can be performed based on preset priorities (e.g., comfort priority or energy saving priority).

[0133] In this embodiment of the application, by combining the predicted path characteristics with the real-time vehicle status, it is possible to perform fine-grained adjustment, mode switching, and energy distribution of the air conditioning, battery, and power system.

[0134] As an optional implementation, step S108 involves controlling the vehicle system according to a control strategy, including: parsing the control strategy to obtain control commands for the vehicle system; and sending the control commands to the vehicle system to control the vehicle system.

[0135] In this embodiment, the vehicle controller or a dedicated domain controller can receive control strategies generated by the vehicle's energy-saving strategy decision module.

[0136] Optionally, the above control strategy can include a logical description, such as: "Start the air conditioning at 7:15, target temperature 22℃" or "Detected a frequently used route, switch to economy mode". The above control strategy can be converted into control commands that can be recognized by the controllers of each vehicle system, such as standard communication protocol commands (e.g., CAN bus messages, vehicle Ethernet messages). Then, the generated control commands can be sent to the corresponding vehicle systems via the vehicle's internal network (e.g., CAN bus) to control the vehicle systems.

[0137] In the embodiments of this application, through precise control command parsing and distribution, the various dispersed on-board systems of the vehicle can accurately understand and coordinate their execution, thereby transforming the abstract algorithmic advantages into actual energy consumption reduction and experience improvement.

[0138] As an optional implementation, the control strategy is parsed to obtain control commands for the vehicle system, including: parsing the control strategy to obtain control parameters corresponding to the control strategy; converting the control parameters to obtain converted control parameters, wherein the communication protocol format of the converted control parameters is consistent with the communication protocol format of the vehicle system; and generating control commands based on the converted control parameters.

[0139] In this embodiment, the control strategy can be parsed to obtain the corresponding control parameters. For example, for the air conditioning adjustment strategy, control parameters such as the target cabin temperature (e.g., 22°C), target battery temperature (e.g., 15°C), execution time window (e.g., 15 minutes in advance), and power limit value (e.g., maximum heating power 500W) can be extracted. For the power switching strategy, control parameters such as the target driving mode identifier (e.g., "green route"), torque limit coefficient (e.g., 0.9), energy recovery level (e.g., "high"), and path matching threshold can be extracted. For the energy distribution strategy, control parameters such as target SOC margin, charging current limit, and accessory energy consumption budget threshold can be extracted.

[0140] Optionally, after obtaining the control parameters, the control parameters can be converted so that the communication protocol format of the converted control parameters is consistent with the communication protocol format of the vehicle system, thereby enabling the generation of control commands based on the converted control parameters.

[0141] In the embodiments of this application, the above steps enable complex control strategies to be accurately and unambiguously transmitted to the controllers of various vehicle systems, thereby achieving cross-domain and cross-system collaborative control.

[0142] As an optional implementation, the vehicle system includes an air conditioning system. Control commands are sent to the vehicle system to control the vehicle system, including: using the vehicle's target bus to send a start command to the air conditioning system to start the air conditioning system.

[0143] In this embodiment, the target bus can be a CAN bus, Ethernet, or a Local Interconnect Network (LIN) bus.

[0144] Optionally, depending on the vehicle's network architecture, the target bus connecting the vehicle controller or body domain controller and the air conditioning system controller can be determined. For routine control commands (such as switching, temperature setting, and airflow), a CAN bus can be used as the target bus; if high-definition display or large data volume interaction is involved, an in-vehicle Ethernet or LIN bus (for small actuators) can be used.

[0145] In this embodiment of the application, by utilizing the vehicle's target bus to achieve remote and automated control of the air conditioning system, the adjustment commands can accurately reach the controller of the air conditioning system and drive the air conditioning system to perform the expected thermal management actions, thereby creating a comfortable cabin environment before passengers board the vehicle, while avoiding energy waste during the initial stage of driving.

[0146] As an optional implementation, the vehicle system includes a battery management system that sends control commands to the vehicle system to control the vehicle system, including sending a battery heating start command to the battery management system to heat the battery to a target temperature.

[0147] In this embodiment, the vehicle's control strategy decision module can calculate the target temperature required by the battery (e.g., increasing it from -5°C to 15°C) and the upper limit of heating power (e.g., limiting it to 500W to prevent overcurrent or overheating) based on the predicted trip duration, distance, and current ambient temperature. Then, a battery heating start command can be constructed. This start command can include not only the logic signal to initiate heating but also specific control parameters.

[0148] Optionally, the vehicle controller or domain controller can send the prepared battery heating start command to the vehicle's battery management system via the vehicle's high-speed CAN bus or in-vehicle Ethernet.

[0149] In the embodiments of this application, the above steps can solve the problems of high internal resistance, low discharge efficiency and slow charging speed of batteries in low temperature environments, so that the battery is in the optimal operating temperature range when the vehicle departs, thereby improving power performance, extending driving range and protecting battery life.

[0150] As an optional implementation, the vehicle system includes a power system, and sends control commands to the vehicle system to control the vehicle system, including: in response to detecting a power switching command sent to the power system, switching the operating mode of the power system from an initial mode to a target mode, and controlling the power system in the target mode, wherein the energy-saving effectiveness of the target mode is higher than that of the initial mode.

[0151] In this embodiment, the vehicle controller can collect the vehicle's GPS position, speed, and heading angle in real time and compare them with commonly used routes in the route prediction results. If the spatial overlap between the real-time driving path and the predicted path exceeds a preset threshold (e.g., 90% or 95%), and the vehicle is in motion, it can be determined that a power switching command has been detected.

[0152] Optionally, after detecting a power switching command sent to the power system, the operating mode of the power system can be switched from the initial mode to the target mode to control the power system. The target mode can be the economic route mode, or simply the economic mode.

[0153] In this embodiment of the application, by identifying routes familiar to the user, road conditions (such as traffic light locations, gradient changes, and congestion points) can be known in advance, thereby breaking the limitations of general control strategies and avoiding the sense of sluggishness brought about by a one-size-fits-all economic model.

[0154] As an optional implementation, the method further includes: in response to the vehicle not being connected to a charging device, or the battery charge in the vehicle being below a charge threshold, controlling the vehicle's cabin system to heat the vehicle's seats.

[0155] In this embodiment, if the vehicle is not connected to a charging device, or if the battery charge in the vehicle is below a power threshold, a low-power thermal management switching command can be generated to control the vehicle's cabin system to heat the vehicle's seats.

[0156] Optionally, if the vehicle is not connected to a charging device, regardless of its SOC (State of Charge), it is determined that "pre-regulation using grid power is unavailable." If the vehicle is connected to a charging device, but the current battery SOC is below a preset "charge threshold" (e.g., 30% or 40%, the specific threshold can be adjusted according to battery type and user settings), it is determined that "battery reserves are insufficient, and power energy should be prioritized." Therefore, low-power thermal management switching commands can be generated. For example, a command can be sent to the air conditioning system controller to prohibit the activation of the high-power heating mode of the PTC heater or heat pump system, or to limit the power to a very low maintenance level. PTC heaters have relatively high power (e.g., 5-7kW), and direct battery power will significantly increase energy consumption. Seat / steering wheel heating can be enabled, for example, by sending a command to the seat control module or body domain controller to activate the seat heating function and / or steering wheel heating function.

[0157] In this embodiment of the application, by taking the above steps, the amount of electrical energy drawn from the battery can be minimized while ensuring the core physical comfort (warm hands and feet) of the driver and passengers, thereby ensuring the vehicle's driving range or preventing the battery from being over-discharged.

[0158] As an optional implementation, the method further includes: generating vehicle prompt information based on route prediction results and control strategies, wherein the prompt information is used to display the vehicle's predicted trip and / or the vehicle's energy-saving results to at least one driver or passenger in the vehicle.

[0159] In this embodiment, the prompt information can be sent to the vehicle's central control screen or instrument panel via the vehicle controller.

[0160] For example, when a user is about to leave work, the app can push a notification: "We have detected that you are about to leave work and go home. We have planned an optimal energy-saving route for you. The estimated arrival time is 18:30. The current battery status is good." The app can also broadcast notifications via voice, using the in-car voice assistant to announce key information in voice form, such as, "We have detected your usual route and switched to intelligent energy-saving mode. Please pay attention to the road conditions ahead."

[0161] Optionally, user interaction and feedback collection can also be implemented. For example, "Confirm," "Cancel," or "Modify" buttons can be provided on the vehicle's infotainment screen or instrument panel. For instance, if a user changes their plans and clicks "Cancel," the predetermined control strategy will stop, and this behavior will be fed back to the predictive model to adjust future prediction probabilities. After the trip, users can be invited to provide a simple evaluation of energy efficiency or comfort (e.g., a star rating), serving as implicit feedback data to further optimize the control strategy.

[0162] The vehicle-mounted system control method of this application embodiment selects travel data with a frequency higher than a frequency threshold from the vehicle's multi-dimensional operating data as target travel data, thereby accurately identifying frequently used routes and travel patterns. Then, using the target travel data samples from the vehicle sample, a prediction model trained on a machine learning model is used to perform in-depth analysis of the target travel data, generating route prediction results for the vehicle. Based on these prediction results, a targeted control strategy is formulated and executed by the vehicle-mounted system. This overcomes the shortcomings of related technologies that use general fixed energy-saving modes or simple real-time feedback, lacking the ability to learn from long-term user habits. It achieves a shift from passive response to proactive, forward-looking control, solving the technical problem of low control effectiveness for vehicle-mounted systems and improving the overall effectiveness of vehicle-mounted system control.

[0163] The above technical solutions of the embodiments of this application will be further illustrated below with reference to preferred embodiments.

[0164] Currently, vehicle energy-saving optimization technologies mainly focus on the following aspects.

[0165] Energy-saving strategies based on fixed road conditions: Most new energy vehicles are pre-set with fixed driving modes such as "economy mode" and "sport mode". These modes usually achieve energy saving by limiting the motor output power, adjusting the air conditioning system operating point, or changing the intensity of regenerative braking. However, these strategies are static and general, and do not take into account the specific characteristics of individual driver habits and usual routes.

[0166] Predictive energy-saving control based on navigation information: Some high-end models combine route information (such as gradient, curves, and speed limits) provided by in-vehicle navigation (e.g., GPS) for limited predictive control. For example, it may reduce drive power in advance when a downhill section is predicted, or smoothly decelerate before entering a speed-limited zone to recover more energy. This technology relies on accurate map data and preset routes, and is usually an optimization for a single trip.

[0167] A simple driving behavior scoring system: Existing systems can score behaviors such as rapid acceleration and sudden braking and prompt drivers to improve. This is only a passive feedback mechanism; it does not combine driving behavior with the specific route context, nor does it actively intervene in vehicle control strategies to achieve automated energy saving.

[0168] Therefore, the relevant technologies have the following problems.

[0169] Lack of personalization and adaptability: Most existing solutions are based on real-time data feedback or fixed rules, without actively learning the driver's long-term, personalized travel patterns. They fail to deeply integrate the micro-road conditions of an individual's long-term driving habits with the fixed routes they frequently use (such as commuting routes). Each driver has different operating preferences on the same route, and a general energy-saving strategy can achieve a better match.

[0170] The current approach fails to achieve learning and optimization throughout the entire vehicle lifecycle: existing methods primarily focus on single trips or static analysis, lacking a vehicle-side intelligent agent that continuously learns from the driver's actual behavior and dynamically updates and optimizes the model. Consequently, the vehicle cannot become increasingly adept at understanding the driver and route as the driving time increases, thus failing to provide more precise energy-saving control.

[0171] Limited prediction accuracy and energy-saving potential: Relying solely on navigation predictions or instantaneous information from a single trip limits the ability to plan energy consumption for the entire trip. It fails to utilize the massive amounts of data accumulated from historically frequent and repeated routes (such as precise energy consumption performance under different seasons, time periods, and congestion levels) to build more accurate prediction models, resulting in a bottleneck in energy-saving effects.

[0172] The passivity and lag of prediction and optimization: Optimization based on real-time driving behavior or single navigation is a form of "reactive" control. The system only adjusts after the trip has started, optimizing only the planned route. It cannot perform forward-looking vehicle energy planning and system pre-adjustment based on high-probability travel plans before the vehicle starts or during daily parking.

[0173] Insufficient utilization of vehicle-side computing power: With the improvement of computing power of in-vehicle chips, vehicles have stronger edge computing capabilities. Existing systems have failed to fully utilize this capability to run complex machine learning models, continuously learn user habits locally, and update and improve personalized optimization strategies. The massive amount of driving data generated by vehicle sensors and controllers has not been systematically used to build owner-specific route and energy consumption models, and the value of the data has not been deeply explored, which is not conducive to data privacy protection.

[0174] Poor balance between user experience and energy saving: Some energy-saving strategies come at the cost of sacrificing power responsiveness or driving comfort, which can easily cause driver discomfort or prompt drivers to actively turn off energy-saving functions. It is difficult for related technologies to find a balance between user driving preferences and maximum energy saving effect based on long-term data analysis.

[0175] To address the aforementioned issues, this application provides a common route analysis and adaptive energy-saving optimization system and method based on vehicle-side machine learning. The core of this system lies in: deploying a lightweight machine learning model on the vehicle side to continuously learn the driver's travel spatiotemporal patterns; and based on the learning results, performing predictive adjustments to the vehicle's thermal management system before departure and adaptive control of the powertrain system during driving.

[0176] Figure 2 This is a schematic diagram of a common route analysis and adaptive energy-saving optimization system based on vehicle-side machine learning according to an embodiment of this application, such as... Figure 2 As shown, the system may include: a data acquisition and preprocessing module 201, a vehicle-side machine learning and route recognition module 202, an energy-saving strategy decision-making module 203, a vehicle control execution module 204, and a human-machine interaction and feedback module 205.

[0177] The data acquisition and preprocessing module 201 continuously collects data from the vehicle's CAN bus, GPS, ambient temperature sensor, etc., including the vehicle's position coordinates, timestamps, driving trajectory, vehicle speed, three-dimensional acceleration, ambient temperature, battery status, air conditioning switch status and set values ​​when the vehicle starts and stops, and performs data cleaning, desensitization and formatting processing.

[0178] The vehicle-side machine learning and route recognition module 202 utilizes local computing power to perform offline analysis of historical travel data when the vehicle is in sleep or low-power mode. For example, feature extraction: key features are extracted from the travel data, such as geofences of departure / arrival points, travel time (weekdays / weekends, hours / minutes), trip duration, and driving routes. Model training and updates: unsupervised learning is used, for example, the DBSCAN algorithm is used to cluster historical parking locations, automatically identifying core permanent locations such as "home" and "workplace". Time series analysis or supervised learning models are used to analyze the frequency and time patterns of departures from each permanent location, establishing a probabilistic prediction model of "departure location-time-destination". Privacy protection: data storage and computation are all completed on the vehicle side. The generated location tags (e.g., "home", "workplace") are only internal logical identifiers and are not uploaded to cloud services or contain precise coordinate information, thus protecting user privacy.

[0179] The energy-saving strategy decision module 203 receives the prediction results and, based on the output of the route recognition module (e.g., there is a 90% probability that the vehicle will depart from home to the company at 7:30 am on a weekday, with a journey of about 15 kilometers and expected moderate road conditions), combined with the current vehicle status (battery SOC, plug status, ambient temperature) and user preset preferences (expected arrival time, comfortable temperature range), formulates specific energy-saving strategies, which include strategy instructions.

[0180] Air conditioning pre-adjustment strategy: If the vehicle is connected to a charging station, the optimal cabin preheating / precooling start time and power curve are calculated within a certain time window before the predicted departure time. The cabin temperature is adjusted to a comfortable range using grid power, thereby minimizing or turning off the air conditioning compressor at the beginning of the journey.

[0181] Power battery pre-preparation strategy: Optimize battery thermal management based on predicted trip distance and road conditions. For example, when a long trip is predicted in winter, the battery is gently heated in advance while it is charging, so that it is in the optimal operating temperature range when departing, reducing internal resistance and heating energy consumption during driving.

[0182] The vehicle control execution module 204 is responsible for translating the strategies generated by the decision-making module into specific execution instructions and sending them to the corresponding controllers. For the air conditioning system: it automatically starts the air conditioning at the predicted time, controlling the compressor, PTC heater, airflow, and circulation mode according to the strategy. For the battery management system: it controls the operating status and power of the battery heater or cooling system. For the powertrain system: during driving, if the vehicle's GPS trajectory closely matches the predicted "frequently used route," it activates an "economical" power curve and energy recovery strategy customized for that route. For example, it reduces power output in advance on downhill sections of familiar roads to enhance energy recovery; and it implements a more aggressive deceleration and recovery strategy at known traffic light intersections.

[0183] The human-machine interaction and feedback module 205 presents the system's prediction results to the user via the vehicle's infotainment screen or mobile app (e.g., "The system predicts you will arrive at the company at 7:30 and has already started pre-adjusting the air conditioning"), and provides options to confirm, cancel, or modify. Simultaneously, the system records the actual energy-saving effect of each optimization strategy, using this as feedback data to fine-tune the machine learning model and optimization strategy, achieving continuous self-evolution.

[0184] In this embodiment, the data acquisition and preprocessing module 201 is connected to the vehicle's CAN bus, GPS receiver, ambient temperature sensor, and cloud service clock system, traffic information, and weather information. The acquired data includes, but is not limited to: spatiotemporal data: GPS location coordinates and timestamps when the vehicle starts and when it stops. Driving data: driving trajectory (GPS point sequence), vehicle speed, three-axis acceleration (x-axis, y-axis, z-axis), etc. Environmental and vehicle status: ambient temperature, battery SOC, battery temperature, battery discharge status, battery charging status (whether plugged in, fast or slow charging), air conditioning on / off status and set temperature, fan speed mode, etc. User operation data: brake pedal opening, accelerator pedal opening, gear information, preferred temperature, etc. Afterwards, the acquired raw data can be preprocessed: first, data cleaning is performed to remove outliers and noisy data; then, the location information is blurred, for example, converting precise GPS coordinates into a geofenced area (e.g., a circular area with a radius of 200 meters); finally, the data is formatted to form a structured travel event record.

[0185] The vehicle-side machine learning and route recognition module 202 utilizes the idle computing power of the vehicle-side high-computing chip (e.g., domain controller) to call pre-processed historical travel data for offline learning when the vehicle is in a low-power state (e.g., after locking the car) or charging.

[0186] Permanent Location Identification: The DBSCAN algorithm is used to cluster the GPS coordinates of historical parking locations (e.g., engine shutdown points). This algorithm can automatically aggregate dense parking spots into clusters and identify core locations. For example, the algorithm clusters parking spots that frequently appear on weekday nights and weekend nights into one cluster and internally labels them as "home"; parking spots that frequently appear from 9 am to 6 pm on weekdays into another cluster and internally labels them as "workplace". This label is only used for internal system logic and does not output precise coordinates externally.

[0187] Travel Pattern Learning: Based on labeled permanent locations, this module analyzes each "start-stop" trip, extracting feature vectors including: departure location (e.g., "home"), departure time (day of the week, hour), trip duration, destination (e.g., "workplace"), and travel route (path point sequence). Then, a lightweight time series prediction model (e.g., Long Short-Term Memory network LSTM or a simple Bayesian network) or supervised learning model is used to learn the mapping relationship between "(departure location, departure time) -> (destination, estimated trip duration, estimated route)". At the start of the current trip, based on the current time, starting location, and historical trip records, the learned sequence pattern is used to predict the most likely destination (or intermediate stop) and its probability in real time. For example, the model learns that starting from "home" between 7:20 and 7:40 on weekdays, there is a 92% probability that the destination is "workplace," with an estimated trip duration of 25-35 minutes and an estimated route of segment ABC. The model is updated regularly as new data is added to adapt to changes in driver habits.

[0188] The energy-saving strategy decision module 203 is activated when the vehicle is powered on or about to be powered on (for example, when the user wakes up the vehicle via an app or based on the user's usual travel patterns). It makes a decision based on the prediction results of the route recognition module and the current vehicle status.

[0189] Receiving Prediction Results: Assuming the current time is 7:25 AM on Monday, the vehicle is plugged in for charging, and the ambient temperature is 0°C. The energy-saving strategy decision module obtains the prediction results from the route recognition module: There is a 92% probability that the user will depart for the "workplace" within 5 minutes, with an estimated journey of 30 kilometers, including a section of highway. Combining real-time traffic information (e.g., provided by the vehicle's navigation system) or historical average travel time, the estimated arrival time at the predicted destination is estimated.

[0190] The air conditioning pre-adjustment strategy is formulated as follows: The energy-saving strategy decision module calculates the optimal cabin preheating scheme based on the predicted departure time (7:30), the user's preset comfort temperature (22℃), the current ambient temperature (0℃), and the vehicle's charging status. Specifically, by consulting the preset energy consumption model, it determines that to reach 22℃ at departure, the PTC heater / heat pump needs to be started in advance, and the airflow should be controlled at medium speed with internal circulation. Therefore, the decision module generates an instruction: start the heating system at 7:25 and operate the air conditioning heater according to the calculated power curve.

[0191] A pre-trip strategy for the power battery was formulated: Based on the predicted travel distance (30 km) and road conditions (including highways), the decision-making module determined that this was a medium-to-long-distance trip. In the current 0℃ environment, the battery's internal resistance is relatively high. Therefore, the decision-making module generated an instruction: starting at 7:25, using the grid power from the charging station, control the battery management system to activate the battery heater, slowly heating the battery temperature from the current temperature to 15℃ (the lower limit of the battery's optimal operating temperature range), ensuring the battery is in a highly efficient operating state at departure.

[0192] The vehicle control execution module 204, as the command conversion and distribution center, receives strategy commands from the decision module and converts them into control commands for various subsystems.

[0193] For the air conditioning system: at 7:25, a start command is sent to the air conditioning controller via the CAN bus, including parameters such as the target temperature of 22℃, air volume, and circulation mode.

[0194] For the battery management system: at 7:25, a battery heating start command is sent to the BMS, setting the target temperature to 15℃ and limiting the heating power to ensure a gentle and safe heating process.

[0195] For the powertrain: During driving, when the vehicle's GPS trajectory matches the "common route ABC segment" predicted by the route recognition module more than a preset threshold (e.g., 95%), the execution module sends a request to the vehicle control unit (VCU) to switch powertrain modes. The VCU switches the current mode to "economy route mode," which features: limiting the maximum output torque of the drive motor to 90% of the standard mode, setting the energy recovery intensity to the highest level, and further reducing drive torque in advance when a known downhill section is identified in the route, prioritizing the use of coasting energy recovery.

[0196] In the vehicle control execution module, the implementation of the "economic route mode" can include, in addition to limiting the maximum torque and adjusting the energy recovery intensity, actively adjusting the excitation current of the motor to optimize its operation in the high-efficiency range; turning off the air conditioning compressor in advance and only maintaining the blower operation when it is predicted that the vehicle will enter a congested section, so as to reduce non-driving energy consumption; or for plug-in hybrid electric vehicles, when the predicted end of the route is a congested urban section, the vehicle can be forced to enter the range-extending mode in the early part of the trip to reserve more power for the battery so that it can achieve pure electric driving in the later congested section, further improving fuel economy.

[0197] The human-machine interaction and feedback module 205 interacts with the user through the vehicle's central control screen and a mobile app. When the system makes a prediction, for example, at 7:10, a pop-up window will appear on the vehicle screen / mobile app prompting: "The system predicts you will be heading to the company at 7:30. Battery preheating and air conditioning pre-adjustment have been initiated. Do you need to modify them? [Confirm] [Cancel]." The user can click "Cancel" to stop all pre-adjustment actions. After the trip, the module will calculate and display the estimated benefits of this energy-saving strategy, for example, "This trip saved approximately 5 kilometers of range in electricity through intelligent pre-adjustment." The user's feedback (e.g., "Confirm" or "Cancel") and the actual energy-saving data (e.g., the actual electricity consumption compared to the historical average consumption when the strategy was not used) will serve as feedback data to fine-tune the model parameters of the vehicle-side machine learning and route recognition module and the energy-saving strategy decision-making module in subsequent low-power states, forming a self-evolving closed loop. Implicit learning can be achieved through subsequent user actions. For example, if the system predicts and performs pre-adjustment of the air conditioning, but the user manually turns it off immediately after getting into the car, the system should learn this preference and reduce the probability of performing pre-adjustment in similar scenarios in the future.

[0198] Figure 3 This is a flowchart of a common route analysis and adaptive energy-saving optimization method based on vehicle-side machine learning according to an embodiment of this application, such as... Figure 3 As shown, the method may include the following steps.

[0199] Step S301: Data collection.

[0200] In this embodiment, vehicle travel and status data are continuously collected and stored locally.

[0201] Step S302: Perform data preprocessing.

[0202] In this embodiment, when the vehicle is idle, a vehicle-side machine learning model is used to analyze historical data, identify frequently used routes and their departure time patterns, and establish a travel prediction model.

[0203] Step S303: Train the travel pattern model.

[0204] In this embodiment, a travel pattern model can be trained.

[0205] Step S304, permanent location clustering identification.

[0206] In this embodiment, vehicle status, time, and location can be monitored in real time. When preset trigger conditions are met (e.g., arriving at the predicted departure time window, vehicle unlocking, or changes in charging status), the strategy decision-making process is initiated. The trip prediction model is invoked, and the frequent stops are clustered and identified based on the current context.

[0207] Step S305, travel intention prediction.

[0208] In this embodiment, travel intention prediction can be performed. High-probability trips are identified. If a high-probability trip is predicted, an integrated energy-saving strategy for that trip is generated, including battery and powertrain optimization schemes. The energy-saving strategy is then executed. If it's a pre-departure strategy, relevant systems (e.g., air conditioning system, battery thermal management system) are automatically controlled; if it's a strategy during travel, the motor torque output and energy recovery intensity are dynamically adjusted. Air conditioning pre-adjustment strategies can control the air conditioning system. Battery pre-preparation strategies can control battery thermal management. The powertrain's economic mode can control the energy recovery intensity and drive mode.

[0209] Step S306: Human-computer interaction and feedback present the prediction results, and record user feedback and energy-saving feedback to optimize the model.

[0210] In this embodiment, after the trip is completed, the actual energy-saving effect and impact on user comfort of the optimization strategy can be evaluated, and the evaluation results can be used as feedback data to update the local machine learning model and strategy parameters.

[0211] Figure 4 This is a flowchart of a vehicle-side machine learning method for recognizing common routes according to an embodiment of this application, such as... Figure 4 As shown, the method may include the following steps.

[0212] Step S401, raw data acquisition.

[0213] In this embodiment, raw data such as GPS coordinates, timestamps, driving trajectory point sequences, and vehicle speed are continuously collected when the vehicle starts / stops.

[0214] Step S402, data preprocessing.

[0215] In this embodiment, parking spot data, travel trajectory data, and time / environment data can be preprocessed. For example, the raw data can be cleaned (outliers and noise are removed), desensitized (location coordinates are blurred and converted into geofenced areas), and formatted to form structured travel event records.

[0216] Step S403, cluster analysis.

[0217] In this embodiment, the coordinates of all shutdown points can be extracted from the preprocessed data, and DBSCAN (a density-based spatial clustering algorithm) can be used to perform cluster analysis on the historical parking points. For example, points with high density are clustered together; noise points are excluded; each cluster represents a permanent location.

[0218] Step S404: Permanent location identification and tag generation.

[0219] In this embodiment, statistical analysis can be performed on each cluster generated by clustering. For example, the location type can be automatically determined based on the distribution of dwell time (e.g., nighttime, weekday daytime); internal logical identifiers can be generated, including: "home", "workplace", "frequently visited location A", "frequently visited location B", etc.; this label is only an internal logical identifier, does not contain precise coordinate information, and is not uploaded to the cloud.

[0220] Step S405, travel event construction.

[0221] In this embodiment, a series of "start-stop" events can be defined as a travel event, and a feature vector can be extracted for each travel event. For example, the departure location (labeled permanent location); departure time (day of the week, hour, minute); arrival location (labeled permanent location); trip duration; and travel route (path features after GPS point sequence compression).

[0222] Step S406: Training the travel pattern model.

[0223] In this embodiment, a probabilistic prediction model can be trained based on a historical travel event dataset. For example, the input features are: (departure location, departure time); the output prediction is: (destination, estimated trip duration, estimated route); and time series analysis, Bayesian networks, or lightweight LSTM models can be used.

[0224] Step S407, Model Output and Application.

[0225] In this embodiment, after the model is trained, it can be called by the energy-saving strategy decision module to output the travel intention prediction results in real time based on the current time and vehicle status, including: predicted departure time; predicted destination; estimated travel time; and estimated travel route.

[0226] In this embodiment, the machine learning model update cycle can be set to automatically trigger retraining once a week or every 500 kilometers. The predicted departure time window can be set to ±30 minutes. A trip matching threshold is set so that a customized power strategy is enabled when the overlap between the real-time driving path and the predicted route exceeds 80%. Users can manually turn this function off or on and can personalize settings such as the pre-adjusted temperature and the earliest start time.

[0227] The following uses a winter commuting scenario as an example to further illustrate the embodiments of this application.

[0228] After several weeks of learning, the vehicle has identified User A's workday pattern: from Monday to Friday, between 7:20 and 7:40 in the morning, he / she leaves home (geofence 1) and has an 85% probability of going to work (geofence 2).

[0229] The vehicle was connected to a home charging station at 6:00 AM on Tuesday. The ambient temperature was -5°C.

[0230] At 6:30 AM, the strategy decision-making module was triggered on a scheduled basis. The module detected that the vehicle was currently located at fence 1, it was a weekday, the probability of moving to fence 2 at 7:30 AM was predicted to be 85%, the current battery SOC was 90%, the vehicle was charging, and the ambient temperature was low.

[0231] Module Decision: Activate the "Comprehensive Pre-Departure Warming Strategy." For example, the battery preheating sub-strategy: Immediately start the battery heater at low power, aiming to raise the battery temperature from -5°C to above 10°C (lower limit of the optimal operating range) before 7:25. The cabin preheating sub-strategy: Calculations show that cabin preheating needs to begin at 7:00. At 7:00, the air conditioning system automatically starts, setting the temperature to 22°C, with automatic fan speed, and using the charging station's power to operate the PTC heater. By 7:25, the cabin temperature has reached 20°C.

[0232] User A boarded the vehicle at 7:28. The interior was warm and comfortable, and the battery was ready. The vehicle departed.

[0233] During the journey, the vehicle's GPS track matched the frequently used "home-to-work" route. The powertrain activated a "winter economy mode" customized for this route: the accelerator pedal response was slightly slower, and the motor torque output was smoother; at a known long downhill section in the middle of the route, the system reduced power output in advance and automatically adjusted the energy recovery level to "strong". The air conditioning maintained the temperature at low power throughout the journey.

[0234] Upon arrival at the company, the system assessed that compared to previous trips during the same period when this feature was not enabled, the battery SOC remained 8% higher during this trip, and the user did not perform any manual operations.

[0235] The following example, taking a summer after-get off work shopping trip as an example, will further illustrate the embodiments of this application.

[0236] The vehicle identified that user B frequently visits a certain supermarket after get off work every Wednesday (geofence 3).

[0237] On a Wednesday afternoon at 5:00 PM, the vehicle was in the parking lot of the office (geofence 2), without the battery plugged in, the ambient temperature was 35°C, and the battery SOC was 70%.

[0238] User B remotely unlocks the vehicle via a mobile app. This action triggers a policy decision.

[0239] Module prediction: Currently located at fence 2, Wednesday evening, the probability of heading to fence 3 is 60%, and the probability of heading to fence 4 is 35%. Since no guns are inserted and SOC is sufficient, the module decides to implement the "rapid pre-driving cooling strategy".

[0240] Within five minutes of User B walking towards the vehicle, the air conditioning system automatically started at medium power to pre-cool the car. By the time User B got in, the interior was no longer extremely stuffy.

[0241] After the vehicle leaves the parking lot, the navigation destination is set to Supermarket 3. The system confirms that the trip matches the prediction and activates the "Summer Urban Economy" power strategy. Before approaching known congested sections, the system suggests that the user activate a higher level of intelligent cruise control for further energy saving (human-machine interaction).

[0242] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0243] According to an embodiment of this application, a control system embodiment for an on-board system in a vehicle is provided. Figure 5 This is a schematic diagram of a control system for an in-vehicle system according to an embodiment of this application, such as... Figure 5 As shown, the control system 500 of the vehicle-mounted system may include: a data processing module 502, a prediction module 504, a decision-making module 506, and an execution module 508.

[0244] The data processing module 502 is used to identify and process the multi-dimensional operating data of the vehicle to obtain the vehicle's target travel data, which represents travel data that occurs more frequently than a frequency threshold within a historical time period. The prediction module 504 is used to call the prediction model to analyze the target travel data and obtain the vehicle's route prediction result. The prediction model is obtained by training a machine learning model using target travel data samples from vehicle samples. The machine learning model is used to learn the spatiotemporal patterns of the travel of the driver and passengers. Based on the learning results, the vehicle's thermal management system is adjusted before departure, and the vehicle's power system is controlled during vehicle operation. The decision module 506 is used to determine the vehicle's control strategy based on the route prediction result. The control strategy represents the rules for controlling at least one onboard system in the vehicle. The execution module 508 is used to control the onboard system according to the control strategy.

[0245] Furthermore, the control system also includes an interaction and feedback module, used to generate vehicle prompts based on route prediction results and control strategies, and to process the prompts, wherein the prompts are used to display the vehicle's predicted trip and / or the vehicle's energy-saving results to at least one passenger in the vehicle.

[0246] In this embodiment, the data processing module can be a data acquisition and preprocessing module; the prediction module can be a vehicle-side machine learning and route recognition module; the decision-making module can be an energy-saving strategy decision-making module; the execution module can be a vehicle control execution module; and the interaction and feedback module can be a human-machine interaction and feedback module.

[0247] According to an embodiment of this application, a control device for an in-vehicle system is provided. It should be noted that the control device for the in-vehicle system can be used to execute the above-described control method for the in-vehicle system.

[0248] Figure 6 This is a schematic diagram of a control device for an in-vehicle system according to an embodiment of this application, such as... Figure 6 As shown, the control device 600 of the vehicle's on-board system may include: an identification unit 602, an analysis unit 604, a determination unit 606, and a control unit 608.

[0249] The identification unit 602 is used to identify and process the multi-dimensional operating data of the vehicle to obtain the vehicle's target travel data, wherein the target travel data represents travel data in which the vehicle occurs more frequently than a frequency threshold within a historical time period; the analysis unit 604 is used to call the prediction model to analyze the target travel data and obtain the vehicle's route prediction result, wherein the prediction model is obtained by training a machine learning model using target travel data samples from vehicle samples, and the machine learning model is used to learn the spatiotemporal patterns of the travel of the driving and riding objects, and based on the learning results, to adjust the vehicle's thermal management system before departure and to control the vehicle's power system during vehicle operation; the determination unit 606 is used to determine the vehicle's control strategy based on the route prediction result, wherein the control strategy represents the rules for controlling at least one on-board system in the vehicle; the control unit 608 is used to control the on-board system according to the control strategy.

[0250] In the vehicle-mounted system control device of this embodiment, the identification unit 602 identifies and processes the multi-dimensional operating data of the vehicle to obtain the vehicle's target travel data, which represents travel data that occurs more frequently than a frequency threshold within a historical time period. The analysis unit 604 calls the prediction model to analyze the target travel data and obtain the vehicle's route prediction result. The prediction model is obtained by training a machine learning model using target travel data samples from vehicle samples. The machine learning model is used to learn the spatiotemporal travel patterns of the driver and passengers. Based on the learning results, the vehicle's thermal management system is adjusted before departure, and the vehicle's power system is controlled during vehicle operation. The determination unit 606 determines the vehicle's control strategy based on the route prediction result, whereby the control strategy represents the rules for controlling at least one vehicle-mounted system. The control unit 608 controls the vehicle-mounted system according to the control strategy, thereby solving the technical problem of low control effectiveness of the vehicle-mounted system and achieving the technical effect of improving the control effectiveness of the vehicle-mounted system.

[0251] Embodiments of this application also provide a vehicle, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods described in various embodiments of this application when it runs.

[0252] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0253] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0254] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0255] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0256] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0257] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0258] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0259] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0260] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0261] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A control method for an on-board system in a vehicle, characterized in that, include: The multidimensional operation data of the vehicle is identified and processed to obtain the target travel data of the vehicle, wherein the target travel data is used to represent the travel data of the vehicle that occurs more frequently than a frequency threshold within a historical time period. The prediction model is invoked to analyze the target travel data and obtain the route prediction result of the vehicle. The prediction model is obtained by training a machine learning model using the target travel data sample of the vehicle sample. The machine learning model is used to learn the travel spatiotemporal patterns of the driver and passenger. Based on the learning results, the thermal management system of the vehicle is adjusted before departure, and the power system of the vehicle is controlled during the vehicle's operation. Based on the route prediction results, a control strategy for the vehicle is determined, wherein the control strategy is used to represent the rules for controlling at least one on-board system in the vehicle. The vehicle system is controlled according to the control strategy described above.

2. The method according to claim 1, characterized in that, The prediction model is invoked to analyze the target travel data, and the route prediction results for the vehicle are obtained, including: In response to the vehicle meeting the triggering condition, the prediction model is invoked to analyze the target travel data and obtain the probability of the vehicle's travel route occurring. The route prediction result is determined based on the probability of occurrence.

3. The method according to claim 2, characterized in that, The method further includes: The prediction model is updated to obtain the updated prediction model; The target travel data is analyzed according to the updated prediction model and the occurrence probability determined by the updated prediction model to obtain the route prediction result.

4. The method according to claim 1, characterized in that, The multidimensional operational data of the vehicle is identified and processed to obtain the target travel data of the vehicle, including: The multidimensional operational data is preprocessed to obtain the structured travel events of the vehicle, wherein the travel events are used to represent the spatiotemporal characteristics of the vehicle within a single travel cycle. The travel event is identified and processed to obtain the target travel data.

5. The method according to claim 4, characterized in that, The multidimensional operational data is preprocessed to obtain the vehicle's structured travel events, including: The multidimensional operational data is cleaned to obtain the cleaned multidimensional operational data. The multidimensional operational data after cleaning is fuzzified to obtain the fuzzified multidimensional operational data. The multidimensional operational data after fuzzification is formatted to obtain the structured travel event.

6. The method according to claim 4, characterized in that, The travel event is identified and processed to obtain the target travel data, including: The travel events are identified and processed to obtain multiple parking locations; The parking spots are grouped according to their spatial density to obtain multiple parking spot clusters; The target travel data is obtained by analyzing the multiple parking spot clusters.

7. The method according to claim 6, characterized in that, The target travel data is obtained by analyzing the multiple parking spot clusters, including: By analyzing the multiple parking spot clusters, the parking duration, frequency of occurrence, and time characteristics of the vehicles at each parking spot in the multiple parking spot clusters are obtained; Based on the parking duration, the frequency of occurrence, and the time characteristics, the multiple parking point clusters are marked to obtain the marking results; The parking spot clusters that appear more frequently than the frequency threshold in the labeling results are identified as the target travel data.

8. The method according to claim 7, characterized in that, Based on the parking duration, the frequency of occurrence, and the time characteristics, the multiple parking point clusters are marked to obtain the marking results, including: Based on the parking duration, the frequency of occurrence, and the time characteristics, the type of each parking spot in the plurality of parking spot clusters is determined; According to the type, the multiple parking spot clusters are marked to obtain the marking results.

9. The method according to claim 1, characterized in that, Based on the route prediction results, the control strategy for the vehicle is determined, including: Based on the route prediction results and the multidimensional operational data, at least one of the following strategies for the vehicle is determined: an adjustment strategy, a switching strategy, and an allocation strategy. The adjustment strategy represents the rules for adjusting the on-board system before the vehicle travels. The switching strategy represents the rules for switching the operating mode of the on-board system during the vehicle's operation. The allocation strategy represents the rules for allocating energy to the on-board system during the vehicle's operation or charging process.

10. The method according to claim 1, characterized in that, The vehicle system is controlled according to the control strategy, including: The control strategy is parsed to obtain control commands for the vehicle system; The control command is sent to the vehicle system to control the vehicle system.

11. The method according to claim 10, characterized in that, The control strategy is parsed to obtain control commands for the vehicle system, including: The control strategy is analyzed to obtain the control parameters corresponding to the control strategy; The control parameters are converted to obtain the converted control parameters, wherein the communication protocol format of the converted control parameters is consistent with the communication protocol format of the vehicle system; The control command is generated based on the converted control parameters.

12. The method according to claim 10, characterized in that, The vehicle system includes an air conditioning system. Sending the control commands to the vehicle system to control it includes: Using the vehicle's target bus, a start command is sent to the air conditioning system to start the air conditioning system.

13. The method according to claim 10, characterized in that, The vehicle system includes a battery management system, which sends control commands to the vehicle system to control the vehicle system, including: Send a battery heating start command to the battery management system to heat the battery to the target temperature.

14. The method according to claim 10, characterized in that, The vehicle system includes the power system. Sending control commands to the vehicle system and controlling the vehicle system includes: In response to detecting a power switching command sent to the power system, the operating mode of the power system is switched from an initial mode to a target mode, and the power system is controlled in the target mode, wherein the energy-saving effectiveness of the target mode is higher than that of the initial mode.

15. The method according to claim 1, characterized in that, The method further includes: In response to the vehicle not being connected to a charging device, or the battery charge in the vehicle being below a charge threshold, the vehicle's cabin system is controlled to heat the vehicle's seats.

16. The method according to any one of claims 1 to 15, characterized in that, The method further includes: Based on the route prediction results and the control strategy, a prompt message is generated for the vehicle, wherein the prompt message is used to display the predicted trip of the vehicle and / or the energy-saving results of the vehicle to at least one driver or passenger in the vehicle.

17. A control system for an on-board system in a vehicle, characterized in that, include: The data processing module is used to identify and process the multi-dimensional operation data of the vehicle to obtain the target travel data of the vehicle, wherein the target travel data is used to represent the travel data of the vehicle that occurs more frequently than a frequency threshold within a historical time period. The prediction module is used to call the prediction model to analyze the target travel data and obtain the route prediction result of the vehicle. The prediction model is obtained by training a machine learning model using the target travel data sample of the vehicle sample. The machine learning model is used to learn the travel spatiotemporal patterns of the driver and passenger. Based on the learning results, the thermal management system of the vehicle is adjusted before departure, and the power system of the vehicle is controlled during the vehicle's operation. A decision module is used to determine a control strategy for the vehicle based on the route prediction results, wherein the control strategy represents a rule for controlling at least one onboard system in the vehicle. An execution module is used to control the vehicle system according to the control strategy.

18. The control system according to claim 17, characterized in that, The control system further includes: An interaction and feedback module is used to generate prompt information for the vehicle based on the route prediction results and the control strategy, and to process the prompt information, wherein the prompt information is used to display the vehicle's predicted trip and / or the vehicle's energy-saving results to at least one passenger in the vehicle.

19. A control device for an on-board system in a vehicle, characterized in that, include: The identification unit is used to identify and process the multi-dimensional operation data of the vehicle to obtain the target travel data of the vehicle, wherein the target travel data is used to represent the frequency of the vehicle's occurrence in a historical time period, and the travel data that is higher than the frequency threshold. The analysis unit is used to call the prediction model to analyze the target travel data and obtain the route prediction result of the vehicle. The prediction model is obtained by training a machine learning model using the target travel data sample of the vehicle sample. The machine learning model is used to learn the travel spatiotemporal patterns of the driver and passenger. Based on the learning results, the thermal management system of the vehicle is adjusted before departure, and the power system of the vehicle is controlled during the vehicle's operation. A determining unit is configured to determine a control strategy for the vehicle based on the route prediction results, wherein the control strategy represents a rule for controlling at least one onboard system in the vehicle. The control unit is used to control the vehicle system according to the control strategy.

20. A processor, characterized in that, The processor is used to run a program, wherein the program, when running, performs the method according to any one of claims 1 to 16.

21. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 16.

22. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the storage medium is located to perform the method according to any one of claims 1 to 16.

23. A vehicle, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 16.

Images