Vehicle function automatic wake-up strategy formulation method based on user usage data
By collecting multi-source data to construct user behavior logs and time-series feature datasets, and using a long short-term memory network model to predict vehicle function wake-up strategies, the limitations of traditional wake-up mechanisms are overcome, enabling personalized automatic wake-up of vehicle functions, improving user experience and battery efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUZHOU ANBOSI AUTOMOTIVE TECHNOLOGY CO LTD
- Filing Date
- 2026-03-06
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional vehicle function wake-up mechanisms rely on manual operation or fixed-time triggering, which leads to inconvenience in human-computer interaction and an inability to adapt to different users' usage habits and scenario needs, resulting in resource waste or delayed function response.
By collecting multi-source data, user behavior logs and time-series feature datasets are constructed. Long short-term memory network models are used to predict the time of boarding and the set of target functions. Based on the priority of battery energy consumption, vehicle functions are automatically woken up to achieve personalized function wake-up strategies.
It enables proactive pre-activation of vehicle functions, enhances the initiative and comfort of the smart cockpit, ensures battery energy efficiency and safety, and improves user experience.
Smart Images
Figure CN122166013A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent vehicle control technology, specifically a method for formulating automatic vehicle function wake-up strategies based on user usage data. Background Technology
[0002] With the rapid development of intelligent connected vehicle technology, in-vehicle systems are typically equipped with multiple functional modules, such as navigation, entertainment, air conditioning, seat adjustment, driver assistance, and remote control. While these functions enhance the user experience, they also bring higher energy consumption and system complexity. Traditional vehicle function wake-up mechanisms mostly rely on manual user operation or are triggered by fixed times or events. However, this approach has significant limitations: on the one hand, frequent manual operation reduces the convenience of human-machine interaction; on the other hand, fixed wake-up logic cannot adapt to different users' usage habits and scenario needs, easily leading to resource waste or delayed function response.
[0003] The problem we need to solve is how to personalize and adaptively adjust vehicle functions based on users' historical usage data and formulate automatic wake-up strategies for vehicle functions. To this end, we now provide a method for formulating automatic wake-up strategies for vehicle functions based on user usage data. Summary of the Invention
[0004] The purpose of this invention is to provide a method for formulating an automatic vehicle function wake-up strategy based on user usage data.
[0005] The objective of this invention can be achieved through the following technical solution: a method for formulating an automatic vehicle function wake-up strategy based on user usage data, comprising the following steps:
[0006] Step S1: Collect raw multi-source data and process it to obtain user behavior logs;
[0007] Step S2: Extract features from the user behavior logs to construct a time-series feature dataset;
[0008] Step S3: Input the time-series feature dataset into the constructed user behavior prediction model, and output the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions;
[0009] Step S4: Based on the output of the user behavior prediction model, obtain the wake-up functions and corresponding wake-up times, execute the wake-up of the wake-up functions and load the initial parameters.
[0010] Furthermore, the original multi-source data includes user usage data, vehicle operation data, and environmental perception data;
[0011] The user usage data includes timestamps of the user's activation and deactivation of the target function, and setting parameters;
[0012] The vehicle operation data includes door opening / closing status, vehicle speed, current remaining battery power, and vehicle operation status.
[0013] The environmental sensing data includes ambient temperature, ambient humidity, and light intensity;
[0014] Obtain basic vehicle parameters, including total vehicle battery capacity, wake-up energy consumption and minimum ready time for various vehicle functions;
[0015] Set a time window, which consists of a timestamp corresponding to the current time and a timestamp corresponding to the previous time with an interval of T from the current time.
[0016] The acquired raw multi-source data is integrated and recorded in a unified format of "timestamp-user usage data-vehicle operation data-environmental perception data", and sorted in ascending order according to the collection timestamp to form a structured user behavior log.
[0017] Furthermore, the process of extracting features from the user behavior logs and constructing a time-series feature dataset includes:
[0018] A feature extraction model is constructed, and user behavior logs are input into the feature extraction model to extract features and obtain a time-series feature dataset that reflects vehicle status, environmental conditions and user intent. The time-series feature dataset includes vector features, time features and type features.
[0019] Furthermore, the process of inputting the time-series feature dataset into the constructed user behavior prediction model and outputting the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions includes:
[0020] The time-series feature dataset is input into the trained user behavior prediction model, and the user behavior prediction model outputs the predicted boarding time, the target function set corresponding to the current comprehensive scenario, the initial parameter prediction and the probability of function activation.
[0021] The target function set is sorted from high to low according to the activation probability of the corresponding function, and the sorting result is used as the output of the user behavior prediction model.
[0022] Furthermore, the process of building a user behavior prediction model includes:
[0023] Construct a long short-term memory network model and initialize the parameters of the constructed long short-term memory network model;
[0024] Collect sample data, which includes time-series feature datasets of vehicles under different usage scenarios. Each set of time-series feature datasets corresponds to a labeled comprehensive scenario category and a corresponding set of target function labels.
[0025] The sample data is randomly divided into a training set and a test set according to a preset ratio;
[0026] The constructed long short-term memory network model is trained under supervision using the training set, and the parameters of the long short-term memory network model are adjusted by optimizing the loss function.
[0027] The prediction accuracy of the trained Long Short-Term Memory (LSTM) network model is verified using a test set. If the prediction accuracy of the training results meets the preset threshold, the training of the LSM network model is completed. If the training results do not meet the preset threshold, the relevant parameters are adjusted and the training is repeated until the training results meet the preset threshold or the number of iterations reaches the preset upper limit, thereby completing the training of the LSM network model and obtaining the user behavior prediction model.
[0028] Furthermore, based on the output of the user behavior prediction model, the process of obtaining the wakeable functions and corresponding wake-up times includes:
[0029] The available wake-up energy consumption budget is obtained based on the vehicle's total battery capacity and the current remaining battery power.
[0030] Calculate a comprehensive priority score for each target function;
[0031] Based on the comprehensive priority score of each target function, the target functions are sorted in descending order to obtain a candidate sequence of target functions;
[0032] Initialize the cumulative energy consumption, traverse the candidate sequence of target functions in turn, obtain the executable wake-up subset based on the cumulative energy consumption and the wake-up energy consumption of the target function, and record the target functions in the executable wake-up subset as wakeable functions;
[0033] For wakeable functions, the system calculates the wake-up time of each target function based on the predicted boarding time and minimum ready time output by the user behavior prediction model.
[0034] Furthermore, the process of waking up the wakeable function and loading initial parameters includes:
[0035] Send a wake-up command to the wake-up function and send the initial parameters output by the user behavior prediction model to the corresponding electronic control unit;
[0036] After each electronic control unit completes initialization based on the received initial parameters, it enters a pre-configuration ready state, thereby enabling the automatic activation of relevant wake-up functions when the user uses the vehicle.
[0037] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention acquires original multi-source vehicle data through a data acquisition unit, processes the original multi-source data to generate user behavior logs, and solves the problem of fragmented and difficult-to-use original data through systematic acquisition and structured processing; it constructs a time-series feature dataset containing vector features, time features, and type features from the user behavior logs, providing a high-precision input foundation for subsequent model construction; it inputs the time-series feature dataset into the constructed user behavior prediction model, outputting the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions, realizing intelligent reasoning from historical data to future needs; based on the remaining battery power and the energy consumption characteristics of each target function, it selects an executable wake-up subset to obtain wake-up functions, sends wake-up commands to them, and loads prediction parameters, intelligently scheduling wake-up tasks according to priority and resource constraints while ensuring battery energy efficiency and safety, avoiding blind power consumption and power overload; this method realizes the transformation of vehicle functions from "passive response" to "active pre-activation," and can automatically wake up commonly used functions according to user habits while ensuring battery energy efficiency and safety, significantly improving the initiative, comfort, and user experience of the intelligent cockpit. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0039] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0040] like Figure 1 As shown, the method for formulating an automatic vehicle function wake-up strategy based on user usage data includes the following steps:
[0041] Step S1: Several data acquisition units are deployed on the vehicle. The data acquisition units are used to collect raw multi-source data and process the raw multi-source data to obtain user behavior logs.
[0042] Step S2: Extract features from the user behavior logs to construct a time-series feature dataset;
[0043] Step S3: Input the time-series feature dataset into the constructed user behavior prediction model, and output the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions;
[0044] Step S4: Based on the output of the user behavior prediction model, obtain the wake-up functions and corresponding wake-up times, execute the wake-up of the wake-up functions and load the initial parameters.
[0045] It should be further explained that, in the specific implementation process, the process of collecting raw multi-source data and processing the raw multi-source data to obtain user behavior logs includes:
[0046] The original multi-source data includes user usage data, vehicle operation data, and environmental perception data;
[0047] The user usage data includes timestamps of the user's activation and deactivation of the target function, and setting parameters;
[0048] The vehicle operation data includes door opening / closing status, vehicle speed, current remaining battery power, and vehicle operation status.
[0049] The environmental sensing data includes ambient temperature, ambient humidity, and light intensity;
[0050] Obtain basic vehicle parameters, including the total battery capacity, wake-up energy consumption and minimum ready time of various vehicle functions;
[0051] The data acquisition unit synchronously marks a unified timestamp when acquiring the above-mentioned raw multi-source data and performs periodic sampling;
[0052] Set a time window, which consists of a timestamp corresponding to the current time and a timestamp corresponding to the previous time with an interval of T from the current time.
[0053] The acquired raw multi-source data is integrated and recorded in a unified format of "timestamp-user usage data-vehicle operation data-environmental perception data", and sorted in ascending order according to the collection timestamp to form a structured user behavior log.
[0054] It should be further explained that, in the specific implementation process, the process of extracting features from the user behavior logs and constructing a time-series feature dataset includes:
[0055] A feature extraction model is constructed, and user behavior logs are input into the feature extraction model to extract features and obtain a time-series feature dataset that reflects vehicle status, environmental conditions and user intent. The time-series feature dataset includes vector features, time features and type features.
[0056] in:
[0057] Vector features: Vector features are numerical feature vectors obtained directly from user behavior log records or through simple calculation. They represent quantifiable states at a certain moment. For example, environmental state vectors include outside temperature, inside temperature, ambient humidity, and light intensity; vehicle state vectors include door opening / closing status and vehicle operating status (off, starting, driving); user operation vectors can be reflected as function setting parameter values in the previous operation record, such as the previously set air conditioning temperature or seat heating level.
[0058] Time features: Time features represent the change pattern features extracted by time-series modeling of dynamic parameters in user behavior logs within each time window. For example, the trend of ambient temperature change within the time window, the frequency of a user activating a certain function while using the vehicle, and the time interval since the most recent activation.
[0059] Type features: Type features refer to discrete category attributes obtained by semantic induction of user behavior logs, which are used to describe comprehensive car use scenarios. For example, type features may include travel type (commuting, weekend leisure travel); environment type (cold winter morning, hot summer afternoon, rainy night, sunny day); user status type (single driver, passenger status).
[0060] It should be further explained that, in the specific implementation process, the process of inputting the time-series feature dataset into the constructed user behavior prediction model and outputting the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions includes:
[0061] The time-series feature dataset is input into the trained user behavior prediction model, and the user behavior prediction model outputs the predicted boarding time, the target function set corresponding to the current comprehensive scenario, the initial parameter prediction and the probability of function activation.
[0062] The target function set is sorted from high to low according to the activation probability of the corresponding function, and the sorting result is used as the output of the user behavior prediction model.
[0063] It should be further explained that, in the specific implementation process, the process of building a user behavior prediction model includes:
[0064] Construct a long short-term memory network model and initialize the parameters of the constructed long short-term memory network model;
[0065] Collect sample data, which includes time-series feature datasets of vehicles under different usage scenarios. Each set of time-series feature datasets corresponds to a labeled comprehensive scenario category and a corresponding set of target function labels.
[0066] The sample data is randomly divided into a training set and a test set according to a preset ratio;
[0067] The constructed long short-term memory network model is trained under supervision using the training set, and the parameters of the long short-term memory network model are adjusted by optimizing the loss function.
[0068] The prediction accuracy of the trained Long Short-Term Memory (LSTM) network model is verified using a test set. If the prediction accuracy of the training results meets the preset threshold, the training of the LSM network model is completed. If the training results do not meet the preset threshold, the relevant parameters are adjusted and the training is repeated until the training results meet the preset threshold or the number of iterations reaches the preset upper limit, thereby completing the training of the LSM network model and obtaining the user behavior prediction model.
[0069] It should be further explained that, in the specific implementation process, the process of obtaining the wakeable function and the corresponding wake-up time based on the output of the user behavior prediction model includes:
[0070] Let F be the set of target functions output by the user behavior prediction model, and let each target function be labeled as follows: Where i = 1, 2, ..., n, that is:
[0071] ;
[0072] The available wake-up energy consumption budget is obtained based on the vehicle's total battery capacity and the current remaining battery power, denoted as . ,Right now:
[0073] ;
[0074] in, For safety reasons, This is the current remaining battery level. This refers to the total capacity of the vehicle's battery.
[0075] Calculate a comprehensive priority score for each target function, denoted as . ,Right now:
[0076] ;
[0077] in, This represents the activation probability of the target function labeled i. This represents the importance weight corresponding to the target function labeled i. This indicates the wake-up energy consumption of the target function labeled i. , , For the empirical weighting coefficients, satisfying:
[0078] ,and ;
[0079] Based on the comprehensive priority score of each target function, the target functions are sorted in descending order to obtain a candidate sequence of target functions;
[0080] Initialize the cumulative energy consumption, and iterate through the candidate sequence of target functions. For the current target function, if the following conditions are met: If the relationship is such that the current target function is added to the executable wake-up subset, then update... for ;
[0081] If the conditions are not met, the current target function is skipped, and the executable wake-up subset is finally obtained. The target functions in the executable wake-up subset are recorded as wake-up functions.
[0082] For wakeable functions, the system calculates the wake-up time of each target function based on the predicted boarding time output by the user behavior prediction model and the minimum ready time, denoted as . ,Right now:
[0083] ;
[0084] in, The predicted boarding time is output by the user behavior prediction model. This is the minimum ready time.
[0085] It should be further explained that, in the specific implementation process, the process of waking up the wakeable function and loading initial parameters includes:
[0086] Send a wake-up command to the wake-up function and send the initial parameters output by the user behavior prediction model to the corresponding electronic control unit;
[0087] After each electronic control unit completes initialization based on the received initial parameters, it enters a pre-configuration ready state, thereby enabling the automatic activation of relevant wake-up functions when the user uses the vehicle.
[0088] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications or equivalent substitutions made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A method for formulating an automatic vehicle function wake-up strategy based on user usage data, characterized in that, Includes the following steps: Step S1: Collect raw multi-source data and process it to obtain user behavior logs; Step S2: Extract features from the user behavior logs to construct a time-series feature dataset; Step S3: Input the time-series feature dataset into the constructed user behavior prediction model, and output the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and its initial parameter predictions; Step S4: Based on the output of the user behavior prediction model, obtain the wake-up functions and corresponding wake-up times, execute the wake-up of the wake-up functions and load the initial parameters.
2. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 1, characterized in that, The original multi-source data includes user usage data, vehicle operation data, and environmental perception data; The user usage data includes timestamps of the user's activation and deactivation of the target function, and setting parameters; The vehicle operation data includes door opening / closing status, vehicle speed, current remaining battery power, and vehicle operation status. The environmental sensing data includes ambient temperature, ambient humidity, and light intensity; Obtain basic vehicle parameters, including total vehicle battery capacity, wake-up energy consumption and minimum ready time for various vehicle functions; Set a time window, which consists of a timestamp corresponding to the current time and a timestamp corresponding to the previous time with an interval of T from the current time. The acquired raw multi-source data is integrated and recorded in a unified format of "timestamp-user usage data-vehicle operation data-environmental perception data", and sorted in ascending order according to the collection timestamp to form a structured user behavior log.
3. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 2, characterized in that, The process of extracting features from the user behavior logs and constructing a time-series feature dataset includes: A feature extraction model is constructed, and user behavior logs are input into the feature extraction model to extract features and obtain a time-series feature dataset that reflects vehicle status, environmental conditions and user intent. The time-series feature dataset includes vector features, time features and type features.
4. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 3, characterized in that, The process of inputting the time-series feature dataset into the constructed user behavior prediction model and outputting the predicted boarding time, the target function set corresponding to the current comprehensive scenario, and the prediction of its initial parameters includes: The time-series feature dataset is input into the trained user behavior prediction model, and the user behavior prediction model outputs the predicted boarding time, the target function set corresponding to the current comprehensive scenario, the initial parameter prediction and the probability of function activation. The target function set is sorted from high to low according to the activation probability of the corresponding function, and the sorting result is used as the output of the user behavior prediction model.
5. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 4, characterized in that, The process of building a user behavior prediction model includes: Construct a long short-term memory network model and initialize the parameters of the constructed long short-term memory network model; Collect sample data, which includes time-series feature datasets of vehicles under different usage scenarios. Each set of time-series feature datasets corresponds to a labeled comprehensive scenario category and a corresponding set of target function labels. The sample data is randomly divided into a training set and a test set according to a preset ratio; The constructed long short-term memory network model is trained under supervision using the training set, and the parameters of the long short-term memory network model are adjusted by optimizing the loss function. The prediction accuracy of the trained Long Short-Term Memory (LSTM) network model is verified using a test set. If the prediction accuracy of the training results meets the preset threshold, the training of the LSM network model is completed. If the training results do not meet the preset threshold, the relevant parameters are adjusted and the training is repeated until the training results meet the preset threshold or the number of iterations reaches the preset upper limit, thereby completing the training of the LSM network model and obtaining the user behavior prediction model.
6. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 5, characterized in that, The process of obtaining wakeable functions and corresponding wake-up times based on the output of the user behavior prediction model includes: The available wake-up energy consumption budget is obtained based on the vehicle's total battery capacity and the current remaining battery power. Calculate a comprehensive priority score for each target function; Based on the comprehensive priority score of each target function, the target functions are sorted in descending order to obtain a candidate sequence of target functions; Initialize the cumulative energy consumption, traverse the candidate sequence of target functions in turn, obtain the executable wake-up subset based on the cumulative energy consumption and the wake-up energy consumption of the target function, and record the target functions in the executable wake-up subset as wakeable functions; For wakeable functions, the system calculates the wake-up time of each target function based on the predicted boarding time and minimum ready time output by the user behavior prediction model.
7. The method for formulating an automatic vehicle function wake-up strategy based on user usage data according to claim 6, characterized in that, The process of waking up a wakeable function and loading initial parameters includes: Send a wake-up command to the wake-up function and send the initial parameters output by the user behavior prediction model to the corresponding electronic control unit; After each electronic control unit completes initialization based on the received initial parameters, it enters a pre-configuration ready state, thereby enabling the automatic activation of relevant wake-up functions when the user uses the vehicle.