Cooperative control method and device for low-voltage power consumption of whole vehicle and electronic equipment

Abstract

The invention provides a cooperative control method and device for low-voltage power consumption of a whole vehicle and electronic equipment, and relates to the technical field of automotive electronics, and the method comprises the steps: obtaining multi-modal original data (including vehicle state information, environment information and user behavior information), and carrying out the preprocessing; performing feature extraction and fusion processing on the preprocessed multi-modal original data to generate a high-level semantic event used for representing the vehicle leaving intention of the user or the parking scene; matching a global power consumption management strategy of a high-level semantic event through a hierarchical finite-state machine model; and according to the global power consumption management strategy, executing hierarchical power supply control on the low-voltage electricity load of the whole vehicle, and in the hierarchical power supply control process, dynamically adjusting control parameters in the global power consumption management strategy based on a comparison result of the actual power consumption monitoring data and a preset baseline model. Depth power shortage of the storage battery caused by redundant power consumption is effectively avoided, the risks that the vehicle cannot be started and equipment is abnormal are reduced, and the reliability of the whole vehicle and the user experience are improved.

Description

Technical Field

[0001] This application relates to the field of automotive electronics technology, and in particular to a collaborative control method, device, and electronic device for low-voltage power consumption of a vehicle. Background Technology

[0002] With the development of automotive intelligence, the number of low-voltage electrical devices installed in vehicles is increasing, leading to significant accumulation of standby power consumption during static parking, which can easily cause battery depletion. Currently, power management mainly relies on each ECU independently executing sleep and wake-up based on local signals (such as CAN bus activity or timers), lacking a unified coordination mechanism. Furthermore, control strategies often depend on single trigger conditions, such as door locking or voltage thresholds, resulting in limited information integration dimensions and fixed strategy parameters, making dynamic adjustments impossible based on ambient temperature, user behavior patterns, and battery health. Information silos exist between systems, hindering global optimization. In summary, existing technologies easily lead to problems such as vehicle startup failure and device malfunctions, severely impacting user experience and vehicle reliability. Summary of the Invention

[0003] The purpose of this application is to provide a collaborative control method, device, and electronic device for low-voltage power consumption of a vehicle, so as to alleviate the above-mentioned technical problems existing in the prior art.

[0004] In a first aspect, the present invention provides a coordinated control method for low-voltage power consumption of a vehicle, comprising: Acquire multimodal raw data and preprocess it; the multimodal raw data includes vehicle status information, environmental information, and user behavior information. Feature extraction and fusion processing are performed on the preprocessed multimodal raw data to generate high-level semantic events that characterize the user's intention to leave the vehicle or the parking scenario. A global power management strategy is adopted to match high-level semantic events using a hierarchical finite state machine model. According to the global power consumption management strategy, the vehicle's low-voltage electrical loads are subjected to hierarchical power control. During the hierarchical power control process, the control parameters in the global power consumption management strategy are dynamically adjusted based on the comparison results between the actual power consumption monitoring data and the preset baseline model.

[0005] In an optional implementation, the multimodal raw data is acquired and preprocessed, including: The vehicle status information transmitted on the vehicle bus is collected. The vehicle status information includes ignition gear position, opening and closing and locking status of doors / windows / trunk, real-time voltage, current and temperature of the battery, and battery health status. The system acquires ambient temperature and light intensity data collected by onboard sensors, as well as location and weather information pushed from the cloud via the onboard communication module, to obtain environmental information. The signal strength change trend via smart key or mobile phone Bluetooth key and the occupant presence status information collected by in-vehicle camera are used as user behavior information. The system performs filtering, time synchronization alignment, and standardized format conversion on vehicle status information, environmental information, and user behavior information.

[0006] In an optional implementation, feature extraction and fusion processing are performed on the preprocessed multimodal raw data to generate high-level semantic events characterizing a user's intention to leave the vehicle or a parking scenario, including: The first key features are extracted from the vehicle status information, including the duration of the ignition switch being in the off state, the duration of all door locks, the battery charge change trend, and the battery health status. The second key feature is extracted from user behavior information, including the process data of the smart key signal continuously weakening until it disappears, the state of no face detected in the car, and the typical duration in historical parking habits. The third key feature is extracted from the environmental information, such as whether the current environment is in extremely cold conditions, whether the light level is at nighttime levels, and whether the vehicle is located in a commonly used parking area. Multi-source information fusion processing is performed on the first key feature, the second key feature, and the third key feature to obtain the fusion result; Based on the fusion results, it is determined whether the user's intention to leave the vehicle is higher than a preset probability threshold. If it is, a corresponding high-level semantic event is generated.

[0007] In an optional implementation, a global power management strategy for matching high-level semantic events using a hierarchical finite state machine model includes: Construct a hierarchical state structure that includes multiple top-level operation modes and corresponding sub-states under each mode. The top-level operation modes include at least normal operation, shallow hibernation, deep energy saving, and emergency power supply. Based on the vehicle state, user behavior, or environmental conditions represented by high-level semantic events, determine whether the preset state transition conditions are met and trigger the corresponding target mode switch. After entering the target mode, the system queries the pre-configured adaptive power management strategy library and selects the matching power management strategy based on the current battery state of charge, ambient temperature range, and estimated shutdown cycle.

[0008] In an optional implementation, hierarchical power control is performed on the low-voltage electrical loads of the vehicle according to a global power consumption management strategy, including: A control instruction set is generated based on the power management strategy. The control instruction set includes the target load identifier, operation action type, delay execution time and wake-up parameter settings, and is sent to the distributed execution unit to coordinate the power state adjustment of the vehicle's low-voltage load. Based on the control instruction set, the low-voltage electrical load of the vehicle is divided into multiple power management levels, and each power management level corresponds to different power consumption levels and functional availability. For loads belonging to the first power management level, maintain normal power supply and allow all functions to be executed; For loads belonging to the second power management level, some of their functional performance is limited, including reducing the brightness of the display unit, turning off the audio output unit, or reducing the operating frequency of the computing unit; For loads belonging to the third power management level, the main power supply is turned off, and only the wake-up monitoring circuit remains active and listens for specific communication messages. For loads belonging to the fourth power management level, a periodic wake-up mechanism is set up to check for wake-up requests at set intervals. For loads that fall under the fifth power management level, disconnect their physical electrical connection to the power network.

[0009] In an optional implementation, based on the comparison results between actual power consumption monitoring data and a preset baseline model, the control parameters in the global power consumption management strategy are dynamically adjusted, including: The battery output current is continuously collected by the current monitoring device from the time the control command set takes effect until the vehicle wake-up signal is received. The actual power consumption monitoring data corresponding to the collection period is calculated. The actual power consumption monitoring data includes the average static current and the cumulative power consumption. Construct a preset baseline model, and determine the energy-saving gain based on the comparison results between the actual power consumption monitoring data and the preset baseline model; The wake-up response time is determined based on the time required for the vehicle to return to full functionality after receiving the wake-up signal. The energy-saving gain is used as a positive adjustment factor, and the wake-up response time exceeding the preset threshold is used as a negative penalty factor to generate a comprehensive feedback signal. Based on the comprehensive feedback signal, the current vehicle operating status, and the control actions already executed, the control parameters in the global power management strategy are iteratively updated using a closed-loop optimization algorithm to optimize the global power management strategy in subsequent cycles.

[0010] In an optional implementation, it further includes: When a high-level semantic event identifies that the ambient temperature is below a set threshold, the battery health status is declining, or the expected shutdown time is long, before entering the deep energy-saving mode, it will retain power to some heating circuits or increase the maintenance level of key modules to prevent the battery from over-discharging or malfunctioning. During strategy execution, continuously monitor for instructions from the remote service platform, anti-theft alarm trigger signals, or wake-up requests from users approaching the vehicle with keys; Upon detecting a valid wake-up signal, the system controls each electrical load to restore power sequentially, so that the vehicle can complete the wake-up process within a specified time.

[0011] Secondly, the present invention provides a coordinated control device for low-voltage power consumption of a vehicle, comprising: The data acquisition module is used to acquire multimodal raw data and preprocess the multimodal raw data; the multimodal raw data includes vehicle status information, environmental information, and user behavior information. The feature extraction and information fusion module is used to extract and fuse features from the preprocessed multimodal raw data to generate high-level semantic events that characterize the user's intention to leave the vehicle or the parking scenario. The strategy decision module is used to match the global power management strategy of high-level semantic events through a hierarchical finite state machine model. The load control and strategy optimization module is used to perform hierarchical power control on the low-voltage electrical loads of the vehicle according to the global power management strategy, and dynamically adjust the control parameters in the global power management strategy based on the comparison results of actual power consumption monitoring data and preset baseline models during the hierarchical power control process.

[0012] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the coordinated control method for low-voltage power consumption of a vehicle according to any of the foregoing embodiments.

[0013] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are invoked and executed by a processor, the computer-executable instructions cause the processor to implement the coordinated control method for low-voltage power consumption of the whole vehicle according to any of the foregoing embodiments.

[0014] The collaborative control method, device, and electronic equipment for low-voltage power consumption in vehicles provided in this application acquire and integrate multimodal information such as vehicle status, environment, and user behavior to accurately identify the user's intention to leave the vehicle and parking scenarios, overcoming the limitations of traditional single-signal judgment. A hierarchical finite state machine model is used to dynamically match the optimal power consumption strategy, improving the systematic nature and adaptability of decision-making. Hierarchical power control, while ensuring the normal operation of safety-related loads (such as the braking system), implements refined power-off management for non-critical loads, thereby significantly reducing static current. By combining actual power consumption monitoring with baseline model comparison, closed-loop optimization of control parameters enables the system to have self-learning capabilities. Overall, this effectively avoids deep battery depletion caused by redundant power consumption, reducing the risk of vehicle starting failure and equipment malfunction, and improving vehicle reliability and user experience. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a collaborative control method for low-voltage power consumption in a vehicle, provided as an embodiment of this application; Figure 2 A core flowchart of collaborative power management decision-making is provided in an embodiment of this application; Figure 3 A detailed flowchart of a collaborative control method for low-voltage power consumption in a vehicle, provided in an embodiment of this application; Figure 4 A structural diagram of a collaborative control device for low-voltage power consumption in a vehicle, provided in an embodiment of this application; Figure 5 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0019] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0020] This application provides a collaborative control method for low-voltage power consumption in a vehicle. See [link to relevant documentation]. Figure 1 As shown, the method mainly includes the following steps: S110: Acquire multimodal raw data and preprocess the multimodal raw data; wherein, the multimodal raw data includes vehicle status information, environmental information and user behavior information.

[0021] In one implementation, the acquired multimodal raw data refers to a dataset containing vehicle status information, environmental information, and user behavior information from different systems and exhibiting heterogeneous characteristics. Vehicle status information includes ignition gear position (e.g., OFF / ACC / ON), door opening / closing and locking status, real-time battery voltage, current, temperature, and health status, as well as parking duration predicted based on historical trajectories. Environmental information may include external temperature, light intensity, geographical location, and weather conditions obtained from the cloud, such as heavy rain or extreme cold warnings. User behavior information may include the signal strength trends of the smart key or mobile phone Bluetooth key, whether the in-vehicle camera recognizes occupant faces, and user driving habits learned through cloud computing, such as frequently used departure times or feature preference settings.

[0022] Preprocessing the aforementioned multimodal raw data can include filtering and denoising continuous signals, judging the state stability of event signals (e.g., a door lock is considered valid only if it remains closed for more than 5 seconds), and aligning data with asynchronous timestamps, thereby ensuring the accuracy and consistency of data in subsequent processing.

[0023] S120 performs feature extraction and fusion processing on the preprocessed multimodal raw data to generate high-level semantic events that characterize the user's intention to leave the vehicle or the parking scenario.

[0024] The feature extraction referred to in this embodiment involves extracting key indicators from the aforementioned multimodal raw data to determine the user's intention to leave the vehicle or the parking scenario. For example, the mean and slope of the battery voltage are calculated using a sliding window to determine the power decay trend; the Kalman filter algorithm is used to smooth the fluctuations in the key signal strength, and its decay rate is calculated to determine whether the user is moving away from or approaching the vehicle; persistent confirmation of the door closing and locking actions is performed to avoid momentary false triggering.

[0025] During the fusion process, multiple independent pieces of evidence can be combined into a unified judgment based on fusion algorithms such as DS evidence theory or fuzzy logic. For example, when the driver's door is closed, the key signal disappears, the GPS location is stable, and there is no facial recognition result inside the vehicle, the support corresponding to "the user has left the vehicle" can be matched to each piece of evidence according to a pre-configured basic probability allocation table, and then the final confidence score can be synthesized through DS combination rules. If the confidence score of "the user has been away from the vehicle for a long time" after fusion reaches a preset threshold or higher, a corresponding high-level semantic event is generated.

[0026] S130 uses a hierarchical finite state machine model to match the global power management strategy of high-level semantic events.

[0027] The aforementioned Hierarchical Finite State Machine (HFSM) model divides the overall operating state of the system into several main states. Each main state further defines multiple sub-states and corresponding state transition conditions. By receiving the aforementioned high-level semantic events as input, state transitions are triggered. Based on the current state, the corresponding power management strategy template is called to map complex vehicle usage scenarios to specific control commands, thereby matching the corresponding global power management strategy.

[0028] For example, the top-level state can be divided into driving, short-term parking, and long-term parking, and the control granularity can be further subdivided within each main state. When the system receives a high-level semantic event that the user has left the vehicle for an extended period of time (98% confidence), it enters the long-term parking main state and selects the optimal sub-strategy based on parameters such as the current battery SOH and ambient temperature.

[0029] The global power management strategy has multiple power management modes pre-stored in the strategy library. For example, it can enable a deep power saving strategy when the power is sufficient and the environment is mild, and retain more wake-up capabilities when the temperature is low or the battery is aging.

[0030] S140 performs hierarchical power control on the low-voltage electrical loads of the vehicle according to the global power management strategy, and dynamically adjusts the control parameters in the global power management strategy based on the comparison results of actual power consumption monitoring data and preset baseline model during the hierarchical power control process.

[0031] In tiered power control, differentiated power management and control can be implemented based on the functional importance and usage frequency of different loads. Control commands are generated by the central coordinating controller and then sent to each regional controller for execution via the CAN bus.

[0032] Meanwhile, the system continuously collects actual power consumption monitoring data through high-precision current sensors, such as real-time acquisition of the total battery current, calculation of average static current and cumulative power consumption, and comparison with the baseline model without optimization strategy. If the actual energy-saving effect is better than expected, the current parameter combination is recorded; if the wake-up delay exceeds the allowable range, a functional penalty term is introduced. The optimization strategy parameters are updated through model predictive control or reinforcement learning mechanisms.

[0033] This embodiment achieves multi-faceted perception of vehicle status, environmental conditions, and user behavior through the acquisition and preprocessing of multimodal raw data, improving the accuracy and completeness of input information. High-confidence high-level semantic events are generated based on feature extraction and information fusion processing, effectively overcoming the risk of misjudgment caused by single signal judgment. A hierarchical finite state machine model is used for structured decision-making in complex parking scenarios, ensuring the rationality and systematic nature of power consumption strategy matching. Combined with a hierarchical power control mechanism, refined power-off management of unnecessary loads is achieved while ensuring the normal operation of critical functions. Strategy parameters are dynamically optimized by comparing actual power consumption with a baseline model, forming a closed-loop self-learning capability. The overall solution significantly reduces the vehicle's static current, extends battery life, reduces start-up failures and equipment malfunctions caused by low battery, and improves vehicle reliability and user experience.

[0034] For ease of understanding, the following provides a detailed description of the collaborative control method for low-voltage power consumption of the whole vehicle provided in the embodiments of this application.

[0035] In one embodiment, the acquisition of multimodal raw data and the preprocessing of the multimodal raw data described above may include the following steps 1.1 to 1.4: Step 1.1: Collect vehicle status information transmitted on the vehicle bus. The vehicle status information includes ignition gear position, opening / closing and locking status of doors / windows / trunk, real-time voltage, current and temperature of the battery, and battery health status.

[0036] In practical applications, status messages broadcast by nodes such as the engine control unit, body controller, and battery management system can be periodically acquired via CAN or LIN bus networks to collect vehicle status information. Specifically, the ignition position is used to determine whether the vehicle is running, accessory power is supplied, or is off; the opening / closing and locking status of doors / windows / trunk is one of the important bases for determining whether the user has left the vehicle; the real-time voltage, current, and temperature of the battery are mainly used to monitor the immediate operating status of the low-voltage power supply system; and the battery health status is a battery aging index calculated based on historical charge-discharge cycles and internal resistance changes, usually expressed as a percentage, which directly affects the conservatism of the power consumption management strategy.

[0037] Step 1.2: Obtain the ambient temperature and light intensity collected by the vehicle-mounted sensors, and obtain the geographical location information and weather information pushed from the cloud through the vehicle-mounted communication module to obtain environmental information.

[0038] Ambient temperature can be collected by temperature sensors located at different positions on the vehicle body. Extremely cold or hot environments will significantly affect the available capacity and self-discharge rate of the battery. Light intensity can be obtained by light sensors to identify day-night changes or parking location (such as underground garages). Geographic location information can be obtained by receiving satellite signals through the GNSS module. Combined with electronic maps, it can be determined whether the vehicle is parked in a home, office, or other permanent location. Weather information can be accessed through the vehicle communication module to connect to the vehicle network platform and obtain real-time weather warnings (such as heavy rain, heavy snow, and extreme low temperatures) pushed by meteorological services.

[0039] Step 1.3: Obtain the signal strength change trend via the smart key or mobile phone Bluetooth key and the occupant presence status information collected by the in-vehicle camera as user behavior information.

[0040] The aforementioned user behavior information serves as clues to whether a user is currently near the vehicle or still using it. By analyzing the signal strength trends of the smart key or mobile Bluetooth key, the change in radio frequency signal reception strength over time can be measured to determine whether the user is moving away from the vehicle or remaining nearby.

[0041] For example, a continuously weakening signal indicates that the user has left, while a stable, high-strength signal suggests they are still near the vehicle. This data is processed using a Kalman filter to improve the stability of distance estimation. Occupant presence information is generated using cameras installed in the cabin, in conjunction with facial recognition or human detection algorithms, to confirm whether anyone remains inside the vehicle. If no face is recognized, combining this information with other signals can increase the confidence that the user has left the vehicle; if a face is recognized, the system delays entering deep sleep mode to ensure occupant safety and functional availability.

[0042] Step 1.4 involves filtering, aligning, and standardizing the vehicle status information, environmental information, and user behavior information.

[0043] The above filtering process is mainly used to eliminate random noise or transient interference in sensor signals, such as smoothing continuous quantities like voltage and temperature using moving average or Kalman filtering algorithms.

[0044] Time synchronization alignment refers to unifying data from different sources with different sampling frequencies and timestamps under the same time reference, thereby ensuring that the signals have consistency in the time dimension during subsequent fusion processes and avoiding misjudgment due to delays or misalignments.

[0045] Standardized format conversion encodes various heterogeneous data (such as Boolean values, numerical values, and event flags) into a structured message format (such as JSON or a specific CAN frame format), thereby facilitating parsing and processing by the central collaborative controller.

[0046] The above implementation method can improve the coverage of key factors affecting low-voltage power consumption management by systematically collecting vehicle status information, environmental information and user behavior information, and effectively improve the quality and reliability of input data by filtering, aligning and format standardizing multi-source data.

[0047] Furthermore, feature extraction and fusion processing are performed on the preprocessed multimodal raw data to generate high-level semantic events that characterize the user's intention to leave the vehicle or the parking scenario. In specific implementation, this may include the following steps 2.1 to 2.5: Step 2.1: Extract the duration of the ignition switch being in the off state, the duration of all door locks, the battery charge change trend, and the battery health status from the vehicle status information as the first key features.

[0048] The duration the ignition switch remains in the off state refers to the time after the vehicle's ignition position is switched from ON or ACC to OFF. This characteristic indicates whether the vehicle has finished operating and entered a stationary state, thus determining whether the user is preparing to leave the vehicle. When this time gradually increases and there is no indication of restarting, the system can preliminarily infer that the vehicle is about to enter the parking stage.

[0049] The duration of all door locks refers to the continuous length of time that the driver's side, front passenger side, rear doors, and trunk are all closed and locked. This feature confirms that the user has completed the actions of getting out of the vehicle and locking it, and is usually strongly correlated with leaving the vehicle. To avoid misjudgments due to momentary accidental touches, it can be preset to only consider a signal as valid after this state has been stably maintained for more than a certain duration (e.g., 5 seconds).

[0050] By using a sliding window approach to perform statistical calculations on real-time battery voltage and current data, the mean, variance, and slope of change are extracted to determine the battery's charge change trend, which is then used to analyze the battery's current discharge rate and energy decay direction.

[0051] Battery health status refers to the degree of battery performance degradation assessed based on battery internal resistance, charge-discharge cycle count, and historical aging data. It is usually expressed as a percentage. Batteries with poor health status are more prone to depletion in low-temperature environments.

[0052] Step 2.2: Extract the process data of the smart key signal continuously weakening until it disappears, the state of no face detected in the car, and the typical duration in historical parking habits from the user behavior information as the second key features.

[0053] The data on the process of the smart key signal continuously weakening until it disappears refers to the monitoring of the radio frequency signal strength received by the vehicle's antenna from the smart key or mobile phone Bluetooth key over time. If the signal strength continuously decreases and eventually falls below the communication threshold, it indicates that the user is moving away from the vehicle. During feature extraction, this process data can be processed using a Kalman filter algorithm to effectively smooth signal fluctuations, improve the stability of distance estimation, and, combined with the rate of change, determine the user's direction of movement, thus enhancing the accuracy of recognizing the intention to leave the vehicle.

[0054] The "no face detected inside the vehicle" status refers to the use of cameras installed in the cabin, combined with facial recognition or human presence detection algorithms, to determine whether there are occupants inside the vehicle. If multiple image analyses fail to identify a face, it serves as one piece of evidence that the user has left the vehicle. Due to potential obstructions or insufficient lighting, using this feature alone has low confidence and requires integration with other information for assessment.

[0055] Typical durations of historical parking habits can be learned through cloud-based big data analysis to identify long-term parking behavior patterns formed by users. For example, vehicles are typically parked for about 8 hours after get off work each day, and for no more than 4 hours after a short trip on weekends.

[0056] Step 2.3: Extract the following from the environmental information as the third key features: whether the current environment is in extremely cold conditions, whether the light level is at night, and whether the vehicle is located in a commonly used parking area.

[0057] Whether the current environment is in extremely cold conditions can be determined by checking whether the ambient temperature value, collected by a temperature sensor located on the exterior of the vehicle, is below a set threshold (e.g., -10°C). Under extremely cold conditions, the battery's usable capacity is significantly reduced, and the self-discharge rate increases, affecting its power supply capacity and recovery performance. Therefore, this characteristic will affect whether the system is allowed to enter a deep sleep mode.

[0058] Whether the illumination level is at nighttime levels refers to whether the ambient brightness is below a preset illuminance threshold, as detected by a light sensor. Nighttime usually indicates that the user has completed their daily car use, increasing the likelihood of the vehicle being parked for an extended period. This can serve as an auxiliary basis for judging the parking scenario, especially applicable to areas with limited GNSS signals, such as underground parking garages.

[0059] Whether a vehicle is located in a frequently used parking area can be determined by obtaining the vehicle's geographical location through a GNSS module and combining it with electronic map information to determine whether it is at home, company, or other permanent locations confirmed by historical trajectory.

[0060] Step 2.4: Perform multi-source information fusion processing on the first key feature, the second key feature, and the third key feature to obtain the fusion result.

[0061] In one implementation, when performing multi-source information fusion processing, since the sources of each feature are heterogeneous, have different accuracies, and may conflict, a fusion algorithm based on Dempster-Shafer evidence theory or multi-source fuzzy logic can be used for processing.

[0062] In the implementation based on Dempster-Shafer evidence theory, the recognition framework is first defined, for example, setting a set of propositions for "user's intention to leave the vehicle": {A: Determined to leave the vehicle, B: Determined not to leave the vehicle}. Then, a Basic Probability Assignment (BPA) function is configured for each evidence source, assigning each feature a degree of support for different propositions, as shown in Table 1. For example, "driver's door closed" corresponds to m(A)=0.7, "key signal disappearance" corresponds to m(A)=0.9, and "face detected inside the vehicle" corresponds to m(B)=0.9.

[0063] Table 1. Example of BPA function assignment (determining user's intention to leave the vehicle)

[0064] Then, the basic probabilities of multiple pieces of evidence are synthesized using the DS combination rule to calculate the joint confidence function. For example, combining the two pieces of evidence "the door is closed" (m1) and "the key signal has disappeared" (m2): m(A)=(m1(A)×m2(A)+m1(A) ×m2(θ)+m1(θ) ×m2(A)) / K m(B)=(m1(B) ×m2(B)+m1(B) ×m2(θ)+m1(θ) ×m2(B)) / K Where K is the normalization factor.

[0065] Step 2.5: Based on the fusion results, determine whether the user's intention to leave the vehicle is higher than a preset probability threshold. If it is, generate the corresponding high-level semantic event.

[0066] The above formula yields a fused confidence value, such as Bel(A) = 0.95, meaning the system has a 95% confidence level that "the user has left the vehicle." This high-confidence fused event will be output to the decision-making level.

[0067] The preset probability threshold can be set according to the actual application scenario, for example, set to 0.9. Only when the confidence value exceeds this threshold is it judged as a high-confidence event. Once the condition is met, the system generates a high-level semantic event, such as "User has been away from the vehicle for an extended period of time (confidence 95%)" or "Vehicle enters deep energy-saving preparation state". This event serves as an input signal for the subsequent decision layer, triggering the state transition of the hierarchical finite state machine and initiating the corresponding power management process.

[0068] In practical applications, the system continuously monitors changes in various input features and dynamically updates the fusion results. If new evidence is subsequently input (such as a user return causing the key signal to recover), the confidence level is recalculated, and if necessary, the generated high-level semantic events are revoked to prevent erroneous actions.

[0069] This method provides a reliable basis for subsequent global power consumption strategy decisions by generating high-level semantic events, making the low-voltage power consumption management of the whole vehicle more intelligent and refined. While ensuring user experience, it effectively extends the battery range, reduces the start-up failure problem caused by power depletion, and improves the reliability and intelligence level of the vehicle.

[0070] Furthermore, the aforementioned global power management strategy for matching high-level semantic events using a hierarchical finite state machine model can, in its specific implementation, include the following steps 3.1 to 3.3: Step 3.1: Construct a hierarchical state structure that includes multiple top-level operation modes and corresponding sub-states under each mode. The top-level operation modes include at least normal operation, shallow hibernation, deep energy saving, and emergency power supply.

[0071] In practical implementation, the central collaborative controller internally presets the aforementioned hierarchical state structure and implements the transition control between each state through software logic. See [link to relevant documentation]. Figure 2 The diagram shows the core flowchart of collaborative power consumption management decision-making.

[0072] Normal operating mode refers to the operating phase in which the vehicle is in the ignition-on (ON) or accessory power supply (ACC) state, and all electrical loads operate normally according to design requirements. In this mode, the system does not perform any power cut-off or power reduction operations on non-critical loads, ensuring the complete availability of driving functions, infotainment, and safety systems.

[0073] Shallow sleep mode refers to a transitional state where the vehicle may be restarted shortly after being turned off (e.g., within 5 minutes). In this mode, some non-essential loads (such as the display and speakers) enter a low-power-limited state, but still maintain the ability to wake up quickly, in order to balance energy saving and user experience.

[0074] Deep energy-saving mode refers to a low-power operation phase that the system enters when it recognizes that the user has been away from the vehicle for an extended period and the expected parking period is also long. In this mode, most non-critical loads are powered off or enter a timed wake-up state, significantly reducing the vehicle's static current and thus extending battery life.

[0075] Emergency power supply mode is a protective operating state activated when the battery voltage is detected to be too low, its health status is severely deteriorated, or the ambient temperature is extremely harsh. In this mode, the system actively limits the power supply level of non-core loads, prioritizing the continuous operation of critical functions such as anti-theft alarms and remote communication, and preventing the vehicle from failing to start due to excessive discharge.

[0076] Step 3.2: Based on the vehicle state, user behavior, or environmental conditions represented by the high-level semantic events, determine whether the preset state transition conditions are met, and trigger the corresponding target mode switch.

[0077] The aforementioned high-level semantic events are high-confidence event outputs generated after multi-source information fusion processing, such as "the user has been away from the vehicle for a long time (confidence 98%)", "battery health has decreased (SOH≤70%)", or "entering an extremely cold environment (-20℃)".

[0078] The preset state transition conditions refer to the set of trigger rules set for each state transition. For example, when a high-level semantic event "the user has been away from the vehicle for an extended period of time" is received, and the vehicle is currently in normal operation or shallow sleep mode, if conditions such as "ignition off for more than 10 minutes", "all doors locked", and "key signal disappearing" are met simultaneously, then it is determined that the prerequisite for switching to "deep energy-saving mode" is met.

[0079] For example, when the system detects that "the battery voltage is below 11.8V" or "the ambient temperature is below -15℃ and the SOH is less than 80%", it will immediately trigger a forced switch to "emergency power supply mode" regardless of the current mode, in order to prevent the battery from being deeply discharged.

[0080] In practice, the central coordinating controller receives high-level semantic events in real time and makes a comprehensive judgment based on the current system status and real-time collected data. Once any state transition condition is met, a state switching command is generated to complete a smooth or emergency switch to the target mode, ensuring that power management behavior is highly matched with actual vehicle usage scenarios.

[0081] Step 3.3: After entering the target mode, query the pre-configured adaptive power management strategy library, and select the matching power management strategy based on the current battery state of charge, ambient temperature range and estimated shutdown cycle.

[0082] The adaptive power management strategy library is a set of preset control templates stored in the non-volatile memory of the central coordinating controller. Each template corresponds to the optimal power management scheme under different operating conditions. For example, a standard sleep strategy is enabled in the scenario of "sufficient power + mild environment + short-term parking"; a conservative energy-saving strategy is enabled in the scenario of "low power + low temperature + long-term parking", retaining more wake-up paths. The strategies in the strategy library are dynamically adjustable, and their parameters (such as sleep delay and wake-up threshold) can be self-optimized based on historical data and learning algorithms (such as reinforcement learning).

[0083] In practical implementation, after completing the target mode switch, the system retrieves the optimal strategy template that meets the current multidimensional conditions from the strategy library based on the battery state of charge (the percentage of the current battery's remaining charge relative to its rated capacity, usually provided by the battery management system), the ambient temperature range (the range of external ambient temperature (e.g., normal temperature zone: 0~30℃, low temperature zone: -10~0℃, extreme cold zone: <-10℃)), and the estimated parking cycle (the predicted duration of the vehicle's current stationary time based on factors such as the user's historical parking habits, current time, and geographical location). The system then parameterizes this template into a specific set of control instructions to guide the subsequent hierarchical power control of the load.

[0084] This implementation method can minimize the static power consumption of the vehicle while ensuring the basic functions of the vehicle and the user wake-up experience, effectively extend the battery range, significantly reduce the problem of power loss caused by long-term parking, and improve the reliability and intelligence level of the vehicle's electronic system.

[0085] Furthermore, the aforementioned hierarchical power control of the vehicle's low-voltage electrical loads based on a global power management strategy can, in a specific implementation, generate a control instruction set according to the power management strategy. This control instruction set includes the target load identifier, operation action type, delay execution time, and wake-up parameter settings, and is sent to the distributed execution unit to coordinate the power state adjustment of the vehicle's low-voltage loads. Based on the control instruction set, the vehicle's low-voltage electrical loads are divided into multiple power management levels, each corresponding to different power consumption levels and functional availability. Specifically: For loads belonging to the first power management level, maintain normal power supply and allow all functions to be executed; For loads belonging to the second power management level, some of their functional performance is limited, including reducing the brightness of the display unit, turning off the audio output unit, or reducing the operating frequency of the computing unit; For loads belonging to the third power management level, the main power supply is turned off, and only the wake-up monitoring circuit remains active and listens for specific communication messages. For loads belonging to the fourth power management level, a periodic wake-up mechanism is set up to check for wake-up requests at set intervals. For loads that fall under the fifth power management level, disconnect their physical electrical connection to the power network.

[0086] Based on the control command set generated above, it is distributed via the vehicle bus to the power management chips of each zone controller or the load itself. In one embodiment, the execution unit adopts hierarchical control measures for its assigned load according to the commands, as shown in Table 2.

[0087] Table 2 Load Classification Control Measures

[0088] In Table 2 of the above-mentioned load classification and control measures, L0 level refers to driving safety-related equipment (such as engine ECU and braking system), corresponding to the aforementioned first power management level; L1 level refers to infotainment system, corresponding to the aforementioned second power management level; L2 level refers to T-Box and autonomous driving domain controller, corresponding to the aforementioned third power management level; L3 level refers to body controller and gateway, corresponding to the aforementioned fourth power management level; and L4 level refers to cigarette lighter and USB interface, corresponding to the aforementioned fifth power management level.

[0089] Furthermore, the above-mentioned dynamic adjustment of control parameters in the global power management strategy based on the comparison results between actual power consumption monitoring data and the preset baseline model may include the following steps 4.1 to 4.5 in specific implementation: Step 4.1: Continuously collect the battery output current from the moment the control command set takes effect until the vehicle wake-up signal is received through the current monitoring device, and calculate the actual power consumption monitoring data corresponding to the collection period. The actual power consumption monitoring data includes the average static current and the cumulative power consumption.

[0090] In one implementation, after the system enters low-power management mode, the current monitoring device samples the total output current of the battery at a frequency of not less than 1Hz and uploads the data to the central co-controller. The controller records the complete cycle from the execution of the first control command (e.g., power off the infotainment system) to the first detection of a valid wake-up request (e.g., key proximity triggering keyless entry). Within this cycle, all battery output currents are processed to calculate the average quiescent current I_avg and the cumulative power consumption ΔAh, which serve as the core input data for evaluating the effectiveness of the power management strategy. In the calculation, the average quiescent current I_avg is the average current over a time period T after the strategy is executed; the power consumption ΔAh = ∫I(t)dt (integral over the current).

[0091] Step 4.2: Construct a preset baseline model and determine the energy-saving gain based on the comparison results between the actual power consumption monitoring data and the preset baseline model.

[0092] The preset baseline model refers to the standard power consumption model of the vehicle under the same or similar parking scenarios when the multimodal collaborative management strategy of this invention is not adopted. This model is established based on historical measured data and is used to characterize the average level of static power consumption of the vehicle under normal conditions. The energy-saving gain refers to the difference between the expected power consumption under the baseline model and the current actual power consumption, and is used to quantify the energy-saving effect brought about by the multimodal collaborative management strategy of this invention.

[0093] In practical applications, the central coordinating controller incorporates multiple preset baseline models, corresponding to different ambient temperature ranges (e.g., -20℃~0℃, 0℃~25℃, above 25℃), different battery health states (SOH≥90%, 80%≤SOH<90%, SOH<80%), and different parking durations (short time <1 hour, medium time 1~6 hours, long time >6 hours). When a complete sleep cycle ends, the controller automatically matches the closest baseline model based on the current environmental conditions and parking duration, extracting its corresponding average static current and cumulative power consumption as reference values. Then, it compares the actual measured power consumption data with these reference values ​​to determine the energy-saving gain.

[0094] For example, during an 8-hour overnight parking session, if the baseline model predicts a power consumption of 2.4 Ah, but the actual consumption is only 0.6 Ah, the energy saving gain is 1.8 Ah, indicating that this strategy significantly reduces power consumption.

[0095] Step 4.3: Determine the wake-up response time based on the time required for the vehicle to return to full functionality after receiving the wake-up signal.

[0096] Wake-up response time is an important performance indicator for measuring user experience. It is defined as the time interval from when the system receives a legitimate wake-up source signal (such as a key approach or door handle sensor) to when all critical electrical loads (such as the engine ECU, infotainment system, and instrument panel) are fully restored to normal working state. In other words, wake-up performance (T_wakeup) refers to the time from receiving the wake-up signal to full functionality readiness. This time directly affects the user's perception of vehicle start-up speed and convenience.

[0097] In one implementation, when a user approaches the vehicle with the smart key or presses the unlock button, the wake-up signal is first captured by the Body Controller (BCM) or Gateway (GW) and transmitted to the Central Coordination Controller. The Central Coordination Controller initiates the wake-up process, sending power restoration and initialization commands to loads at L2 and L3 levels. The system synchronously starts a timer to record the time from the moment the signal is received until all critical loads complete their self-tests and enter the L0 full-function state.

[0098] For example, it takes about 300ms for the T-Box to wake up from deep sleep, and about 800ms for the infotainment system to recover from a complete L4 power failure to a displayable screen. The controller ultimately determines the maximum latency path of the entire system as the wake-up response time and incorporates it into the experience evaluation system.

[0099] Step 4.4: Use the energy-saving gain as a positive adjustment factor and the wake-up response time exceeding the preset threshold as a negative penalty factor to generate a comprehensive feedback signal.

[0100] The aforementioned positive adjustment factor indicates that the more energy saved, the greater the positive incentive the system receives. The negative penalty factor indicates that if the wake-up response time exceeds the user's acceptable range (e.g., a preset threshold of 1 second), it is considered a cost affecting the usability of the function and needs to be suppressed. The comprehensive feedback signal is a weighted composite evaluation index that reflects the degree of balance between energy efficiency and user experience achieved by this strategy.

[0101] The central coordinating controller maps the energy-saving gain to a non-negative value as a positive contribution term, while simultaneously determining whether the wake-up response time exceeds a preset safety threshold (e.g., 1000ms). If it does not exceed the threshold, the penalty factor is zero; if it does, a corresponding negative penalty term is generated based on the excess ratio. The two factors are weighted and summed to generate a comprehensive feedback signal, for example, the comprehensive feedback signal Reward = α × energy-saving gain factor - β × function penalty factor, where α and β are the corresponding weights.

[0102] Step 4.5: Based on the comprehensive feedback signal, the current vehicle operating status, and the control actions already executed, the control parameters in the global power management strategy are iteratively updated using a closed-loop optimization algorithm to optimize the global power management strategy in subsequent cycles.

[0103] When using reinforcement learning for policy fine-tuning, the following approach is adopted: ①State: Battery SOC, ambient temperature, estimated parking time.

[0104] ②Action: Adjust the parameters in the strategy, such as changing the T-Box's sleep inspection frequency from "once every hour" to "once every two hours".

[0105] ③Reward: Calculated based on the effect after the strategy is executed.

[0106] Reward=α×(ΔCurrent_Saved)-β×(Functionality_Penalty) Among them, ΔCurrent_Saved is the current saving value; Functionality_Penalty is the function penalty item (e.g., if the user wakes up the vehicle due to too deep a sleep state and the delay exceeds 500ms, a large penalty will be imposed).

[0107] Through continuous trial and error and feedback, the system learns which "action" (parameter adjustment) yields the greatest long-term "reward" under specific conditions, thus achieving adaptive optimization. For example: {Target:T-Box,Action:Enter_L3_Mode,Delay:600s,Param:Wakeup_Interval=7200s}.

[0108] The above methods improve the intelligence level and long-term effectiveness of the vehicle's low-voltage power management, fundamentally alleviate the risk of battery depletion, extend battery life, and thus enhance users' trust and satisfaction with intelligent electric vehicles.

[0109] Furthermore, the above also includes the following steps 5.1 to 5.3: Step 5.1: When the high-level semantic event identifies that the ambient temperature is lower than the set threshold, the battery health status is declining, or the expected shutdown time is long, before entering the deep energy-saving mode, the power supply of some heating circuits is retained or the maintenance level of key modules is increased to prevent the battery from being over-discharged or malfunctioning.

[0110] In one example, an ambient temperature below a set threshold means the external temperature enters a low-temperature range (e.g., below -10°C). At this temperature, the battery's chemical activity decreases, its usable capacity diminishes, and it is prone to voltage drops. A decline in battery health status means that the State of Harm (SOH) value assessed by the on-board diagnostic system is below a preset threshold (e.g., 80%), indicating severe battery aging and weakened energy storage capacity. A prolonged expected parking time means that based on analysis of the user's historical behavior, the parking is predicted to last for a certain duration (e.g., more than 24 hours). In these high-risk scenarios, implementing a conventional deep energy-saving strategy may deplete the battery or render critical functions unusable. Therefore, the system proactively adjusts its strategy, preserving necessary power supply circuits or upgrading the maintenance level.

[0111] For example, when the central coordinating controller receives a fused high-level semantic event such as "extremely cold environment," "low SOH," or "long-term parking," it does not immediately execute the standard L3 / L4 level shutdown action. Instead, it invokes a preset safety enhancement strategy. For instance, when the ambient temperature is below -15°C, the system can control relays to maintain periodic power supply to the battery heating film or cabin insulation circuit (e.g., powering on for 10 minutes every 2 hours) to prevent electrolyte freezing. For batteries with an SOH below 75%, the system lowers the sleep level of the gateway (GW) and body control module (BCM) from L3 (deep sleep) to L2 (standby) to ensure they are always in a low-latency wake-up state. If the estimated parking time exceeds 48 hours, the system will postpone the power-off time of unnecessary loads or shorten the inspection interval of the T-Box to ensure that the remote monitoring function remains online.

[0112] Step 5.2: During the policy execution process, continuously listen for instructions from the remote service platform, anti-theft alarm trigger signals, or wake-up requests from users approaching the vehicle with keys.

[0113] Remote service platform commands can include commands such as remotely starting the air conditioner, checking the vehicle's location, and OTA upgrade notifications. Anti-theft alarm trigger signals come from security devices such as collision sensors, illegal door opening detection, or in-vehicle motion detection cameras. When a user approaches the vehicle with a key, the system can detect changes in Bluetooth RSSI signal strength or low-frequency LF field induction.

[0114] Step 5.3: Upon detecting a valid wake-up signal, control each electrical load to restore power in sequence so that the vehicle can complete the wake-up process within a specified time.

[0115] Effective wake-up signals can include local operations (such as pulling a car door handle), remote control commands, or remote platform commands. Power is restored sequentially, that is, power is supplied in an orderly manner according to the importance of the load and the startup dependency, rather than powering up the entire system instantaneously, in order to avoid current surges.

[0116] In practical applications, once the central coordinating controller confirms receipt of a valid wake-up signal, it immediately initiates a tiered wake-up sequence. First, it restores power to L0-level loads related to driving safety, such as the engine ECU and braking system. Then, it activates L1-level devices such as the infotainment system and LCD instrument panel, gradually improving their performance. Next, it wakes up L2 / L3-level modules such as the T-Box and autonomous driving domain controller, bringing them out of sleep mode and reconnecting them to the network. Finally, it restores L4-level non-essential circuits, such as USB charging ports and seat heating, as needed. The entire process is coordinated and executed by the area controller, following a predefined power supply timing diagram to ensure that each module has sufficient initialization time and that the total wake-up time meets design requirements. For example, in a typical wake-up, the BCM completes authentication within 200ms, unlocks the doors within 500ms, displays the instrument panel within 800ms, and achieves full functionality readiness within 1000ms.

[0117] In an application example of the collaborative control method for low-voltage power consumption of a vehicle according to an embodiment of this application, assume a vehicle equipped with this system enters a family garage at 10 PM, the user turns off the engine (OFF), closes the door, locks the car, and leaves with the key. See also Figure 3 As shown, the execution method mainly includes the following steps: S1 Information Collection: The system collected the following information: "OFF position", "all doors locked", "key signal strength continuously weakens until disappearing", "GPS location fixed", "ambient temperature 25℃", and "battery voltage 12.5V (100% SOC)".

[0118] S2 Information Fusion: The fusion algorithm integrates the above information to generate a high-confidence semantic event that "the user has left the car and plans to park it for a long time (based on historical habit learning, usually until 7 am tomorrow)".

[0119] S3 Strategy Decision: Based on this event, the central controller enters "deep energy-saving mode." After querying the strategy library, it determines strategy A+ for the scenario of 9 hours of parking, sufficient battery power, and mild environmental conditions. ①T+5 minutes: Downgrade the infotainment unit, LCD instrument cluster, and USB ports to L4 (complete shutdown).

[0120] ②T+10 minutes: Place the T-Box and gateway in L3 (deep sleep, wake up once every 1 hour to check cloud instructions).

[0121] ③ Keep the Body Control Controller (BCM) in L2 (standby) to receive the key wake-up signal.

[0122] ④ Keep the burglar alarm system in L1 (restricted operation).

[0123] S4 Strategy Execution: Commands are sent via the CAN bus. Zone Controller 1 controls the relay to disconnect the power to the infotainment system; Zone Controller 2 adjusts the T-Box's power management mode. The vehicle's static current rapidly decreases from the usual 30mA and stabilizes at 5mA.

[0124] S5 Optimization: The system detected that the actual current dropped to 5mA, consistent with the expected target, indicating that the strategy was effective. Simultaneously, the learning module recorded this successful experience for future strategy fine-tuning in similar scenarios.

[0125] In summary, the vehicle low-voltage power consumption collaborative control method proposed in this application achieves globally optimal decision-making by integrating multimodal data such as vehicle status, environmental information, and user behavior, effectively avoiding power consumption redundancy caused by independent control of each ECU. Employing a hierarchical state machine and graded management strategy, it significantly improves energy efficiency while ensuring user experience and vehicle safety. The system can dynamically adjust management strategies based on historical data, user habits, and environmental changes, possessing continuous learning and adaptive capabilities to continuously optimize energy efficiency throughout the vehicle's lifecycle. Furthermore, this method fundamentally avoids battery depletion issues and, through functions such as a "welcome preparation mode," achieves energy-saving goals while also considering user convenience and technological experience.

[0126] Based on the above method embodiments, this application also provides a collaborative control device for low-voltage power consumption of the whole vehicle, see [link to relevant documentation]. Figure 4 As shown, the device mainly includes the following parts: The data acquisition module 410 is used to acquire multimodal raw data and preprocess the multimodal raw data; wherein, the multimodal raw data includes vehicle status information, environmental information and user behavior information; The feature extraction and information fusion module 420 is used to extract and fuse features from the preprocessed multimodal raw data to generate high-level semantic events that characterize the user's intention to leave the vehicle or the parking scenario. The strategy decision module 430 is used to match the global power management strategy of high-level semantic events through a hierarchical finite state machine model. The load control and strategy optimization module 440 is used to perform hierarchical power control on the low-voltage electrical load of the vehicle according to the global power management strategy, and dynamically adjust the control parameters in the global power management strategy based on the comparison results of actual power consumption monitoring data and preset baseline model during the hierarchical power control process.

[0127] In one feasible implementation, the data acquisition module 410 described above is specifically used for: The vehicle status information transmitted on the vehicle bus is collected. The vehicle status information includes ignition gear position, opening and closing and locking status of doors / windows / trunk, real-time voltage, current and temperature of the battery, and battery health status. The system acquires ambient temperature and light intensity data collected by onboard sensors, as well as location and weather information pushed from the cloud via the onboard communication module, to obtain environmental information. The signal strength change trend via smart key or mobile phone Bluetooth key and the occupant presence status information collected by in-vehicle camera are used as user behavior information. The system performs filtering, time synchronization alignment, and standardized format conversion on vehicle status information, environmental information, and user behavior information.

[0128] In one feasible implementation, the feature extraction and information fusion module 420 is specifically used for: The first key features are extracted from the vehicle status information, including the duration of the ignition switch being in the off state, the duration of all door locks, the battery charge change trend, and the battery health status. The second key feature is extracted from user behavior information, including the process data of the smart key signal continuously weakening until it disappears, the state of no face detected in the car, and the typical duration in historical parking habits. The third key feature is extracted from the environmental information, such as whether the current environment is in extremely cold conditions, whether the light level is at nighttime levels, and whether the vehicle is located in a commonly used parking area. Multi-source information fusion processing is performed on the first key feature, the second key feature, and the third key feature to obtain the fusion result; Based on the fusion results, it is determined whether the user's intention to leave the vehicle is higher than a preset probability threshold. If it is, a corresponding high-level semantic event is generated.

[0129] In one feasible implementation, the strategy decision module 430 is specifically used for: Construct a hierarchical state structure that includes multiple top-level operation modes and corresponding sub-states under each mode. The top-level operation modes include at least normal operation, shallow hibernation, deep energy saving, and emergency power supply. Based on the vehicle state, user behavior, or environmental conditions represented by high-level semantic events, determine whether the preset state transition conditions are met and trigger the corresponding target mode switch. After entering the target mode, the system queries the pre-configured adaptive power management strategy library and selects the matching power management strategy based on the current battery state of charge, ambient temperature range, and estimated shutdown cycle.

[0130] In one feasible implementation, the load control and strategy optimization module 440 is specifically used for: A control instruction set is generated based on the power management strategy. The control instruction set includes the target load identifier, operation action type, delay execution time and wake-up parameter settings, and is sent to the distributed execution unit to coordinate the power state adjustment of the vehicle's low-voltage load. Based on the control instruction set, the low-voltage electrical load of the vehicle is divided into multiple power management levels, and each power management level corresponds to different power consumption levels and functional availability. For loads belonging to the first power management level, maintain normal power supply and allow all functions to be executed; For loads belonging to the second power management level, some of their functional performance is limited, including reducing the brightness of the display unit, turning off the audio output unit, or reducing the operating frequency of the computing unit; For loads belonging to the third power management level, the main power supply is turned off, and only the wake-up monitoring circuit remains active and listens for specific communication messages. For loads belonging to the fourth power management level, a periodic wake-up mechanism is set up to check for wake-up requests at set intervals. For loads that fall under the fifth power management level, disconnect their physical electrical connection to the power network.

[0131] In one feasible implementation, the load control and strategy optimization module 440 is specifically used for: The battery output current is continuously collected by the current monitoring device from the time the control command set takes effect until the vehicle wake-up signal is received. The actual power consumption monitoring data corresponding to the collection period is calculated. The actual power consumption monitoring data includes the average static current and the cumulative power consumption. Construct a preset baseline model, and determine the energy-saving gain based on the comparison results between the actual power consumption monitoring data and the preset baseline model; The wake-up response time is determined based on the time required for the vehicle to return to full functionality after receiving the wake-up signal. The energy-saving gain is used as a positive adjustment factor, and the wake-up response time exceeding the preset threshold is used as a negative penalty factor to generate a comprehensive feedback signal. Based on the comprehensive feedback signal, the current vehicle operating status, and the control actions already executed, the control parameters in the global power management strategy are iteratively updated using a closed-loop optimization algorithm to optimize the global power management strategy in subsequent cycles.

[0132] In one feasible implementation, the system further includes a wake-up control module, for: When a high-level semantic event identifies that the ambient temperature is below a set threshold, the battery health status is declining, or the expected shutdown time is long, before entering the deep energy-saving mode, it will retain power to some heating circuits or increase the maintenance level of key modules to prevent the battery from over-discharging or malfunctioning. During strategy execution, continuously monitor for instructions from the remote service platform, anti-theft alarm trigger signals, or wake-up requests from users approaching the vehicle with keys; Upon detecting a valid wake-up signal, the system controls each electrical load to restore power sequentially, so that the vehicle can complete the wake-up process within a specified time.

[0133] The vehicle low-voltage power consumption collaborative control device provided in this application has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the embodiment of the vehicle low-voltage power consumption collaborative control device can be referred to the corresponding content in the aforementioned vehicle low-voltage power consumption collaborative control method embodiment.

[0134] This application also provides an electronic device, such as... Figure 5 The diagram shows the structure of the electronic device 100, which includes a processor 51 and a memory 50. The memory 50 stores computer-executable instructions that can be executed by the processor 51. The processor 51 executes the computer-executable instructions to implement any of the above-mentioned collaborative control methods for low-voltage power consumption of the whole vehicle.

[0135] exist Figure 5 In the illustrated embodiment, the electronic device further includes a bus 52 and a communication interface 53, wherein the processor 51, the communication interface 53, and the memory 50 are connected via the bus 52.

[0136] The memory 50 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 53 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 52 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 52 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0137] The processor 51 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 51 or by instructions in software form. The processor 51 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory. The processor 51 reads the information in the memory and, in conjunction with its hardware, completes the steps of the coordinated control method for low-voltage power consumption of the whole vehicle in the aforementioned embodiment.

[0138] This application also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the aforementioned collaborative control method for low-voltage power consumption of the entire vehicle. For specific implementation details, please refer to the foregoing method embodiments, which will not be repeated here.

[0139] The computer program product of the collaborative control method, device and electronic device for low-voltage power consumption of the whole vehicle provided in the embodiments of this application includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0140] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0141] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, 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 a portion 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, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0142] In the description of this application, the terms “first,” “second,” “third,” “fourth,” “fifth,” etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for cooperative control of low-voltage power consumption of a whole vehicle, characterized by, The method comprises the following steps: acquiring multi-modal raw data and pre-processing the multi-modal raw data; wherein the multi-modal raw data comprises vehicle state information, environmental information and user behavior information; performing feature extraction and fusion processing on the pre-processed multi-modal raw data to generate high-level semantic events for representing user intention to leave the vehicle or parking scenarios; matching a global power consumption management strategy of the high-level semantic events through a hierarchical finite state machine model; performing hierarchical power supply control on the vehicle low-voltage electrical load according to the global power consumption management strategy, and dynamically adjusting the control parameters in the global power consumption management strategy based on the comparison result of the actual power consumption monitoring data and the preset baseline model during the hierarchical power supply control.

2. The vehicle low-voltage power consumption cooperative control method according to claim 1, characterized by, acquiring multi-modal raw data and pre-processing the multi-modal raw data, comprising: collecting vehicle state information transmitted on the vehicle bus, the vehicle state information including ignition gear, opening and closing and locking state of doors, windows and trunk, real-time voltage, current and temperature of the battery, and battery health status; obtaining environmental temperature and light intensity collected by vehicle-mounted sensors, and geographical location information and weather information pushed by the cloud through the vehicle-mounted communication module to obtain the environmental information; obtaining the signal strength trend of the intelligent key or the mobile phone Bluetooth key and the occupant presence state information collected by the in-vehicle camera as user behavior information; performing filtering processing, time synchronization alignment and standard format conversion on the vehicle state information, the environmental information and the user behavior information.

3. The method according to claim 1, wherein, performing feature extraction and fusion processing on the pre-processed multi-modal raw data to generate high-level semantic events for representing user intention to leave the vehicle or parking scenarios, comprising: extracting the time length of the ignition switch in the off state, the duration of all door locks, the battery power change trend and the battery health status from the vehicle state information as the first key feature; extracting the process data of the intelligent key signal continuously weakening until disappearing, the state of not detecting a face in the vehicle and the typical time length in the historical parking habit from the user behavior information as the second key feature; extracting whether the current environment is in extremely cold conditions, whether the light is at night level and whether the vehicle is located in the commonly used parking area from the environmental information as the third key feature; performing multi-source information fusion processing on the first key feature, the second key feature and the third key feature to obtain a fusion result; determining whether the user intention to leave the vehicle is higher than a preset probability threshold based on the fusion result, and if so, generating a corresponding high-level semantic event.

4. The vehicle low-voltage power consumption cooperative control method according to claim 1, characterized by, matching a global power consumption management strategy of the high-level semantic events through a hierarchical finite state machine model, comprising: building a hierarchical state structure containing multiple top-level running modes and corresponding sub-states under each mode, wherein the top-level running modes at least include normal operation, shallow hibernation, deep energy saving and emergency power supply; determining whether the preset state transition condition is met according to the vehicle state, user behavior or environmental condition represented by the high-level semantic events, and triggering the switching of the corresponding target mode. After entering the target mode, a pre-configured adaptive power management strategy library is queried, and a matched power management strategy is selected according to the current battery state of charge, the environmental temperature interval and the estimated parking period.

5. The method according to claim 1, wherein, According to the global power management strategy, hierarchical power supply control is performed on the low-voltage electrical load of the vehicle, including: According to the power management strategy, a control instruction set is generated, which includes target load identification, operation action type, delay execution time and wake-up parameter setting, and is sent to a distributed execution unit to coordinate the power supply state adjustment of the low-voltage load of the vehicle; According to the control instruction set, the low-voltage electrical load of the vehicle is divided into multiple power management levels, each power management level corresponds to different power consumption levels and functional availability; For the load belonging to the first power management level, the normal power supply state is maintained, and all functions are allowed to be executed; For the load belonging to the second power management level, the performance of part of its functions is limited, including reducing the brightness of the display unit, turning off the audio output unit or reducing the operating frequency of the operation unit; For the load belonging to the third power management level, the main power supply is turned off, only the wake-up monitoring circuit is kept in an active state, and specific communication messages are monitored; For the load belonging to the fourth power management level, a periodic wake-up mechanism is set to check whether there is a wake-up request at a set interval; For the load belonging to the fifth power management level, the physical electrical connection between it and the power supply network is cut off.

6. The method according to claim 1, wherein, Based on the comparison result of the actual power consumption monitoring data and the preset baseline model, the control parameters in the global power management strategy are dynamically adjusted, including: The current monitoring device continuously collects the battery output current from the time when the control instruction set takes effect to the time before the vehicle wake-up signal is received, and calculates the actual power consumption monitoring data corresponding to the collection period, which includes the average static current and the cumulative power consumption; A preset baseline model is constructed, and the energy saving gain is determined according to the comparison result of the actual power consumption monitoring data and the preset baseline model; The time required from receiving the wake-up signal to the vehicle recovering to the full function state is determined as the wake-up response time; The energy saving gain is used as a positive adjustment factor, and the wake-up response time exceeding the preset threshold is used as a negative penalty factor to generate a comprehensive feedback signal; Based on the comprehensive feedback signal, the current vehicle operating state and the control action performed, the control parameters in the global power management strategy are iteratively updated by using a closed-loop optimization algorithm to optimize the global power management strategy in the subsequent period.

7. The method according to claim 1, wherein, It also includes: When the high-level semantic event identifies that the environmental temperature is lower than the set threshold, the battery health state decreases or the expected parking time is long, before entering the deep energy saving mode, part of the heating circuit is powered or the maintenance level of the key module is increased to prevent the battery from being discharged too much or the function from failing; During the strategy execution process, the wake-up request from the remote service platform instruction, the anti-theft alarm trigger signal or the user carrying the key close to the vehicle is continuously monitored; When a valid wake-up signal is detected, the power supply of each electrical load is restored in sequence so that the vehicle completes the wake-up process within a specified time.

8. A collaborative control device for low-voltage power consumption in a vehicle, characterized in that, It includes: The data acquisition module is configured to acquire multi-modal raw data and pre-process the multi-modal raw data; wherein the multi-modal raw data comprises vehicle state information, environment information, and user behavior information. The feature extraction and information fusion module is configured to perform feature extraction and fusion processing on the pre-processed multi-modal raw data to generate a high-level semantic event representing a user intention to leave the vehicle or a parking scenario. The strategy decision module is configured to match a global power consumption management strategy of the high-level semantic event through a hierarchical finite state machine model. The load control and strategy optimization module is configured to perform hierarchical power supply control on the low-voltage electrical loads of the vehicle according to the global power consumption management strategy, and dynamically adjust the control parameters in the global power consumption management strategy based on a comparison result of actual power consumption monitoring data and a preset baseline model during the hierarchical power supply control.

9. An electronic device, comprising: The computer readable storage medium stores computer executable instructions, and the computer executable instructions, when invoked and executed by the processor, cause the processor to implement the vehicle low-voltage power consumption cooperative control method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions, and the computer executable instructions, when invoked and executed by the processor, cause the processor to implement the vehicle low-voltage power consumption cooperative control method according to any one of claims 1 to 7.

Images