Vehicle control method and vehicle

Abstract

This application relates to a vehicle control method and a vehicle, belonging to the field of smart cockpit technology. The method includes: acquiring an environmental data sequence, a state data sequence, and a key data sequence of a digital key; determining the in-vehicle position probability of the digital key based on the environmental data sequence, the state data sequence, and the key data sequence; determining the actual position of the digital key based on the in-vehicle position probability; and realizing vehicle control based on the actual position. This method avoids the problem of misjudging the digital key position caused by a single Bluetooth signal and simple threshold rules, improves the accuracy of digital key position determination, ensures stable operation of vehicle control functions, and optimizes the user's driving experience.

Description

Technical Field

[0001] This application relates to the field of intelligent cockpit technology, and in particular to a vehicle control method and a vehicle. Background Technology

[0002] With the continuous evolution of vehicle intelligence technology, traditional vehicle keys (such as mechanical keys and remote keys) are gradually being replaced by digital keys, and vehicle control depends on the location of the digital key (such as the digital key being inside the vehicle or the digital key being outside the vehicle).

[0003] Currently, vehicle digital controllers collect signals from digital keys (e.g., digital keys integrated into smartphone Bluetooth modules) and process them according to simple signal strength threshold rules to determine the location of the digital key, and then control the vehicle based on the location of the digital key.

[0004] However, Bluetooth signals are susceptible to interference from environmental and hardware factors, causing inaccuracies in the position of the digital key, which is determined solely by simple rules. This leads to deviations in vehicle control. Specifically, when the digital key is actually located inside the vehicle, it may be misinterpreted as being located outside, causing abnormal vehicle locking or the inability to start the powertrain. Conversely, when the digital key is actually located outside the vehicle, it may be misinterpreted as being inside, preventing the vehicle's locking function from triggering properly. This, in turn, affects the user experience. Summary of the Invention

[0005] This application provides a vehicle control method and vehicle, addressing the technical problems of Bluetooth signals being susceptible to environmental and hardware interference, and the tendency to make errors in determining the digital key position using simple rules, thus causing abnormal vehicle control. It avoids the defects of the key misjudging the vehicle's interior from the exterior, leading to abnormal locking and inability to start the engine, and the key misjudging the vehicle's exterior from the interior, causing locking failure. It stabilizes the vehicle control logic and optimizes the user experience. The specific technical solution is as follows: In a first aspect of this application, a vehicle control method is provided, the method comprising: Acquire the vehicle's environmental data sequence, status data sequence, and digital key key data sequence; Based on the environmental data sequence, the state data sequence, and the key data sequence, determine the probability of the digital key's position inside the vehicle; Based on the probability of the location inside the vehicle, the actual location of the digital key is determined; and based on the actual location, the vehicle is controlled.

[0006] In an optional implementation, determining the probability of the digital key's in-vehicle location based on the environmental data sequence, the state data sequence, and the key data sequence includes: The environmental data sequence, the state data sequence, and the key data sequence are aligned and concatenated according to time sequence to obtain a multimodal feature vector sequence; Local features are extracted from the multimodal feature vector sequence to obtain a feature map sequence; Extract the effective features from the feature map sequence to obtain the initial compressed feature sequence; Based on the temporal sequence of the initial compressed feature sequence and the sequence changes corresponding to the temporal sequence, a global temporal feature sequence is obtained; The global temporal feature sequence is fused with feature information to obtain a deep feature vector; The probability of the vehicle's interior location corresponding to the digital key is calculated based on the deep feature vector.

[0007] In an optional implementation, determining the actual location of the digital key based on the in-vehicle location probability includes: Obtain the set of location probability thresholds; The probability of the in-vehicle location is compared with the set of location probability thresholds to obtain the probability comparison result; Based on the probability comparison results, the actual location of the digital key is obtained.

[0008] In an optional implementation, obtaining the location probability threshold set includes: Acquire historical sample data, which includes location ground truth labels; Determine the predicted probability of the in-vehicle position corresponding to the historical sample data; Based on the predicted in-vehicle location probabilities corresponding to all the historical sample data and the true location labels, a performance evaluation curve is generated. Based on the performance evaluation curve, a first position probability threshold and a second position probability threshold are determined; the first position probability threshold and the second position probability threshold are combined to obtain the position probability threshold set.

[0009] In an optional implementation, generating a performance evaluation curve based on the predicted in-vehicle position probabilities corresponding to all the historical sample data and the ground truth labels of the positions includes: Multiple candidate probability thresholds are set. For any one of the candidate probability thresholds, the predicted probability of the in-vehicle position corresponding to each of the historical sample data is compared with the candidate probability threshold to obtain the predicted position determination result corresponding to each of the historical sample data. The predicted location determination results and location truth labels corresponding to each of the historical sample data are compared, and the number of true positives and false positives in the historical sample data are statistically obtained. The true positive rate and false positive rate under the candidate probability threshold are calculated based on the number of true positives and the number of false positives. Using the false positive rate as the horizontal axis and the true positive rate as the vertical axis, the coordinate points corresponding to the true positive rate and false positive rate obtained under all candidate probability thresholds are connected sequentially to generate the performance evaluation curve.

[0010] In an optional implementation, determining the first location probability threshold and the second location probability threshold based on the performance evaluation curve includes: Obtain the business requirement type, and determine the upper limit of the false positive rate and the lower limit of the true positive rate based on the business requirement type; On the performance evaluation curve, select the node with the highest true positive rate and a false positive rate that does not exceed the upper limit of the false positive rate, and determine the candidate probability threshold corresponding to the node as the first position probability threshold. On the performance evaluation curve, select the node with the lowest false positive rate and the true positive rate that is not lower than the lower limit of the true positive rate. Then, determine the candidate probability threshold corresponding to the node as the second position probability threshold.

[0011] In one optional implementation, the acquisition of business requirement types includes: Obtain vehicle application scenarios; Based on the preset mapping relationship between vehicle application scenarios and business requirement types, the business requirement type corresponding to the vehicle application scenario is determined.

[0012] In an optional implementation, the set of location probability thresholds includes a first location probability threshold and a second location probability threshold; The step of obtaining the actual location of the digital key based on the probability comparison result includes: When the probability of the location inside the vehicle is greater than or equal to the first location probability threshold, the actual location of the digital key is determined to be inside the vehicle. When the probability of the position inside the vehicle is less than or equal to the second position probability threshold, the actual position of the digital key is determined to be outside the vehicle. When the probability of the position inside the vehicle is less than the first position probability threshold but greater than the second position probability threshold, the actual position of the digital key is determined to be a hesitant position.

[0013] In an optional implementation, the step of controlling the vehicle based on the actual location includes: When it is determined that the actual location of the digital key is inside the vehicle, the user permissions are set to the first permission set, and the vehicle is controlled according to the first permission set; When it is determined that the actual location of the digital key is outside the vehicle, the user permissions are set to the second permission set, and the vehicle is controlled according to the second permission set; When the actual location of the digital key is determined to be a hesitant location, a digital key location confirmation prompt instruction is sent to the user to request the user to confirm whether the digital key is currently inside or outside the vehicle; in response to the user confirming that the digital key is inside the vehicle, the user's permissions are set to a first permission set, and the vehicle is controlled according to the first permission set; or, in response to the user confirming that the digital key is outside the vehicle, the user's permissions are set to a second permission set, and the vehicle is controlled according to the second permission set.

[0014] In a second aspect of this application, a vehicle control device is also provided, the device comprising: The data acquisition module is used to acquire the vehicle's environmental data sequence, status data sequence, and digital key key data sequence; The probability output module is used to determine the probability of the digital key's position inside the vehicle based on the environmental data sequence, the state data sequence, and the key data sequence. The vehicle control module is used to determine the actual location of the digital key based on the probability of the location inside the vehicle; and to implement vehicle control based on the actual location.

[0015] In a third aspect of the embodiments of this application, a vehicle is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements any of the vehicle control methods described in the first aspect above.

[0016] In a fourth aspect of the embodiments of this application, a storage medium is also provided, the storage medium storing instructions that, when run on a computer, cause the computer to execute any of the vehicle control methods described in the first aspect above.

[0017] In a fifth aspect of the embodiments of this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform any of the vehicle control methods described above.

[0018] The technical solution provided in this application embodiment acquires a vehicle's environmental data sequence, a state data sequence, and a key data sequence of a digital key; determines the vehicle's in-vehicle position probability of the digital key based on the environmental data sequence, the state data sequence, and the key data sequence; determines the actual position of the digital key based on the in-vehicle position probability; and realizes vehicle control based on the actual position.

[0019] By collecting three types of multi-dimensional data—vehicle environment data sequences, vehicle status data sequences, and digital key data sequences—it can cover various factors affecting positioning accuracy, including environmental interference, vehicle operating status, and key hardware working status. Through multi-dimensional data fusion calculation, the probability of the digital key's position inside the vehicle is obtained. A probabilistic quantification method is used to characterize the digital key's positional likelihood, replacing the traditional coarse-grained position determination mode. This avoids positioning deviations caused by interference from a single Bluetooth signal, improving the accuracy and reliability of digital key position determination both inside and outside the vehicle.

[0020] Based on the in-vehicle location probability obtained through multi-source data fusion, the actual location of the digital key is determined, providing an accurate and effective basis for judgment in the vehicle control logic. Based on the determined actual location of the digital key, corresponding vehicle control strategies are executed. When the digital key is inside the vehicle, abnormal vehicle locking is restricted and the powertrain start-up permission is enabled; when the digital key is outside the vehicle, the normal vehicle locking function is executed. This avoids vehicle control malfunctions caused by misjudgment of the digital key's location, ensuring the stable and compliant operation of all vehicle control functions.

[0021] Based on the location determination results of the digital key and standardized vehicle control logic, it avoids abnormal vehicle use scenarios, realizes the function of normal vehicle start-up when getting in the car and normal locking when getting out of the car, adapts to complex vehicle use scenarios such as regular driving and parking, reduces smart key system failures, improves the operational stability and intelligence level of the vehicle intelligent control system, optimizes the user experience of vehicle smart keys, and enhances the product application value of the vehicle intelligent control system. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0025] Figure 1 This is a schematic flowchart illustrating a vehicle control method in an embodiment of this application; Figure 2This is a flowchart illustrating a method for determining the probability of a digital key's position inside a vehicle, as shown in an embodiment of this application. Figure 3 This is a flowchart illustrating a digital key vehicle in-vehicle position probability discrimination model inference method in an embodiment of this application; Figure 4 This is a flowchart illustrating a method for determining the actual location of a digital key in an embodiment of this application; Figure 5 This is a flowchart illustrating a method for obtaining a location probability threshold set in an embodiment of this application; Figure 6 This is a flowchart illustrating a performance evaluation curve generation method in an embodiment of this application; Figure 7 This is a schematic diagram of a performance evaluation curve shown in an embodiment of this application; Figure 8 This is a flowchart illustrating a dual-position probability threshold determination method in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a vehicle control device shown in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a vehicle shown in an embodiment of this application. Detailed Implementation

[0026] 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, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0028] With the rapid development of vehicle intelligence technology, digital keys have gradually replaced traditional mechanical keys and remote keys, becoming the core control medium for functions such as vehicle unlocking, power-on start, and driving control. The intelligent control logic of vehicles is highly dependent on the real-time location status of the digital key. Currently, most in-vehicle digital key location determination schemes rely solely on a single threshold rule based on the strength of the digital key's Bluetooth signal to determine its location. They simply distinguish the digital key's position inside and outside the vehicle through a simple signal comparison, and then execute the corresponding vehicle control strategy.

[0029] During the research and development and deployment testing in complex in-vehicle conditions, researchers in this field have discovered that in-vehicle Bluetooth signals are highly susceptible to interference from multiple factors, including in-vehicle electromagnetic interference, vehicle structure obstruction, external wireless noise, vehicle operating status, and digital key hardware conditions. A single signal threshold determination mode has significant limitations in various scenarios. The coarse-grained determination logic with fixed thresholds cannot adapt to the dynamically changing in-vehicle environment, easily leading to misjudgments such as the digital key being actually inside the vehicle being misjudged as outside, or vice versa. This can result in vehicle control malfunctions such as abnormal locking, failure to start the powertrain, and failure of the parking lock function, severely impacting vehicle control stability and user safety, and significantly reducing the user's smart car experience.

[0030] To address the technical shortcomings of traditional digital key location determination methods, such as weak anti-interference capability, low determination accuracy, and poor scene adaptability, which lead to abnormal vehicle control, this application proposes a vehicle intelligent control scheme based on multi-source data fusion and artificial intelligence model inference technology. This scheme abandons the traditional single Bluetooth signal determination logic and integrates three types of multi-dimensional influencing parameters: the vehicle's environmental data sequence, state data sequence, and the digital key's key data sequence. By constructing a standardized multi-modal feature vector sequence through time alignment and feature splicing, and relying on a multi-level neural network model to complete feature depth mining and probability quantization inference, the scheme accurately outputs the probability of the digital key's position inside the vehicle.

[0031] Meanwhile, this application completes adaptive calibration of dual-position probability thresholds based on performance evaluation curves, constructing a three-stage position determination mechanism including in-vehicle determination interval, out-of-vehicle determination interval, and hesitation determination interval, thus eliminating the shortcomings of traditional fixed threshold forced determination. For operating scenarios where the critical probability interval cannot be accurately determined, a hesitation position is set, and position calibration is completed through user active confirmation. Combined with an incremental learning mechanism, the model performance is continuously iterated and optimized to achieve high-precision and high-stability identification of digital key position under complex in-vehicle operating conditions. This matches differentiated vehicle control permission strategies, solves the vehicle function abnormality problem caused by position misjudgment in traditional solutions, balances vehicle safety and user experience, and improves the scenario adaptability and operational stability of the smart key control system.

[0032] Based on the above inventive concept, such as Figure 1 The diagram shown is a schematic representation of the implementation flow of a vehicle control method provided in this application. This method is applied to a vehicle and may specifically include the following steps: S101, acquire the vehicle's environmental data sequence, status data sequence, and digital key key data sequence.

[0033] In this embodiment, environmental data sequences, state data sequences, and key data sequences of the digital key are acquired. For example, the acquired environmental data sequences include various environmental parameters affecting Bluetooth signal transmission, such as in-vehicle electromagnetic interference intensity, environmental obstruction status, and external signal interference parameters over a period of time. The acquired vehicle state data sequences include various vehicle operating state parameters, such as vehicle door opening / closing status, vehicle start / stop status, vehicle power supply status, and in-vehicle Bluetooth hardware operating parameters over a period of time. The acquired key data sequences include four types of key operating condition data: key Bluetooth signal reception strength, signal transmission delay, key hardware operating status, and key real-time communication parameters over a period of time. This provides solid data computation support for accurately determining the actual location of the digital key.

[0034] In this embodiment, specifically, using the unified collection timestamp of the in-vehicle device as the sole benchmark, the original collection time nodes of the vehicle's environmental data, status data, and digital key key data are extracted respectively. Data groups within the same collection time series are selected to eliminate the time difference in multi-source data collection. Based on the collection characteristics of different data dimensions, the electromagnetic interference intensity, environmental obstruction status, and external signal interference parameters in the environmental data sequence; the door opening / closing status, vehicle start / stop status, vehicle power supply status, and in-vehicle Bluetooth hardware operating parameters in the vehicle status data sequence; and the Bluetooth signal reception strength, signal transmission delay, key hardware operating status, and key real-time communication parameters in the key data sequence are mapped and matched one-to-one according to the unified timestamp, ensuring that various types of data correspond to each other under a single time series node.

[0035] In this embodiment, for example, the electromagnetic interference intensity inside the vehicle, the opening and closing status of the vehicle doors, and the Bluetooth signal reception strength of the key are acquired according to a predetermined acquisition period. The acquisition period is based on Tc=T0+(c 1) ΔT (Formula 1) is determined; where Tc is the time of the cth acquisition, T0 is the time of the first acquisition (starting from the vehicle bus wake-up time, it is recommended that T0≥50ms), ΔT is the fixed interval between two adjacent acquisitions, it is recommended that ΔT be 20ms, and the number of consecutive acquisitions should not be less than 10.

[0036] S102, based on the environmental data sequence, the state data sequence, and the key data sequence, determine the probability of the digital key's position inside the vehicle.

[0037] In this embodiment, considering that traditional determination methods rely solely on simple rules to judge Bluetooth signals, which cannot adapt to environmental interference and hardware differences in complex vehicle usage scenarios and suffer from inaccurate location determination, this application determines the probability of the digital key's location inside the vehicle based on environmental data sequences, state data sequences, and key data sequences. By fusing and calculating multi-dimensional data, various interference factors affecting the positioning accuracy of the digital key are taken into account, and the probability of the digital key being inside the vehicle is objectively represented in a probabilistic quantification form. This overcomes the limitations of traditional qualitative determination methods, improves the fault tolerance and adaptability of location determination, and provides solid data computation support for subsequently accurately determining the actual location of the digital key and achieving reliable vehicle control.

[0038] S103 determines the actual location of the digital key based on the probability of its position inside the vehicle; and controls the vehicle based on the actual location.

[0039] In this embodiment, considering the inherent flaw of relying solely on a single signal threshold to determine the digital key's location, which is prone to misjudgment and cannot guarantee the accurate execution of vehicle control logic, the actual location of the digital key is determined based on the in-vehicle position probability. Vehicle control is then implemented according to this actual location. By relying on quantified in-vehicle position probability results for location verification and determination, the actual in-vehicle and out-of-vehicle scenarios of the digital key can be accurately distinguished, avoiding judgment errors caused by interference factors. By matching the corresponding vehicle control logic with the true and accurate key location results, it is ensured that vehicle start-up, power start-stop, and other control functions accurately correspond to the actual state of the digital key, achieving precise and standardized execution of vehicle control.

[0040] This application embodiment collects three types of multi-dimensional data: vehicle environment data sequence, vehicle status data sequence, and digital key data sequence. This covers various factors affecting positioning accuracy, including environmental interference, vehicle operating status, and key hardware working status. By fusing multi-dimensional data, the probability of the digital key's position inside the vehicle is obtained. A probabilistic quantification method is used to characterize the digital key's position probability, replacing the traditional coarse-grained position determination mode. This avoids positioning deviations caused by interference from a single Bluetooth signal, improving the accuracy and reliability of determining the digital key's position inside and outside the vehicle.

[0041] Based on the in-vehicle location probability obtained through multi-source data fusion, the actual location of the digital key is determined, providing an accurate and effective basis for judgment in the vehicle control logic. Based on the determined actual location of the digital key, corresponding vehicle control strategies are executed. When the digital key is inside the vehicle, abnormal vehicle locking is restricted and the powertrain start-up permission is enabled; when the digital key is outside the vehicle, the normal vehicle locking function is executed. This avoids vehicle control malfunctions caused by misjudgment of the digital key's location, ensuring the stable and compliant operation of all vehicle control functions.

[0042] Based on the location determination results of the digital key and standardized vehicle control logic, it avoids abnormal vehicle use scenarios, realizes the function of normal vehicle start-up when getting in the car and normal locking when getting out of the car, adapts to complex vehicle use scenarios such as regular driving and parking, reduces smart key system failures, improves the operational stability and intelligence level of the vehicle intelligent control system, optimizes the user experience of vehicle smart keys, and enhances the product application value of the vehicle intelligent control system.

[0043] In the implementation of this application, if Figure 2 As shown, the specific steps in S102 to "determine the probability of the digital key's position inside the vehicle based on the environmental data sequence, state data sequence, and key data sequence" are as follows: S1021, Align and concatenate the environmental data sequence, state data sequence, and key data sequence according to time sequence to obtain a multimodal feature vector sequence.

[0044] In this embodiment, considering the inconsistencies in the timing of various data acquisitions and the potential for fusion bias due to the independent and scattered data dimensions, which prevents accurate characterization of the real-time location status of the digital key, the environmental data sequence, state data sequence, and key data sequence are aligned and concatenated according to their timing to obtain a multimodal feature vector sequence. By unifying the acquisition time benchmark for various data types, the timing misalignment problem of data from different dimensions is eliminated. Effective feature information from environmental interference parameters, vehicle operating status parameters, and key operating condition parameters is integrated to construct a multimodal feature vector sequence with complete dimensions and unified timing. This achieves standardized fusion of multi-source heterogeneous data, ensuring the data integrity and timing accuracy of subsequent in-vehicle location probability calculations, avoiding judgment errors caused by data dispersion and timing discrepancies, and providing standardized feature data support for the refined and precise quantitative determination of the digital key's location.

[0045] In this embodiment of the application, specifically, a multimodal feature vector sequence at time point t is constructed based on the environmental data sequence, the state data sequence, and the key data sequence. This multimodal feature vector sequence combines the data of the environmental dimension, the vehicle state dimension, and the digital key dimension to form a complete "vehicle and digital key state snapshot", which is provided as input to the digital key in-vehicle position probability discrimination model, so that the digital key in-vehicle position probability discrimination model can eliminate Bluetooth signal interference caused by the environment and hardware, and accurately calculate the in-vehicle position probability of the digital key.

[0046] The multimodal feature vector sequence is shown in formula (2): (Formula 2); Among them, EMI t Block represents the electromagnetic interference intensity inside the vehicle at time t. t RFInt represents the environmental occlusion state at time t. tDoor represents the external signal interference parameter at time t. t Eng represents the open / closed state of the vehicle door at time t. t Vbat represents the vehicle's start / stop status at time t. t BtPara represents the power supply status of the entire vehicle at time t. t RSSI represents the operating parameters of the vehicle's Bluetooth hardware at time t. t Delay represents the Bluetooth signal strength received by the key at time t. t KeyHw represents the signal transmission delay at time t. t Comm represents the working state of the key hardware at time t. t This represents the real-time communication parameters of the key at time t.

[0047] EMI t The meaning is the electromagnetic noise power value generated by the electrical and electronic control equipment inside the cabin, collected by the vehicle-mounted radio frequency detection module at time point t; EMI t Its function is to quantify the degree of interference from in-vehicle electrical equipment to Bluetooth communication, and to compensate for signal attenuation errors; EMI t The example value is 32dBμV / m (which means that there is strong electromagnetic interference in the vehicle, and the Bluetooth signal is prone to abnormal attenuation).

[0048] Block t This means quantifying the level of obstruction to the Bluetooth communication link by vehicle body metal and interior metal ornaments through Bluetooth signal reflection loss; Block t Its function is to distinguish between line-of-sight communication and non-line-of-sight communication caused by obstruction, and to avoid misjudgment of signal strength caused by obstruction; Block t Example values ​​are 2 (moderate obstruction, metal interior trim blocks the direct Bluetooth path) and 0 (no obstruction, signal transmission is lossless).

[0049] RFInt t The meaning is the amplitude of noise signals generated by Bluetooth and wireless devices on the same frequency band outside the vehicle; RFInt t Its function is to identify external wireless interference sources and correct signal jumps caused by external clutter; RFInt t The example value is 8dBm (there are multiple wireless devices on the same frequency in the vicinity, and the interference level is relatively high).

[0050] Door t The meaning is a unified indicator of the opening and closing of the four doors based on feedback from the vehicle's door lock sensors; Door t The purpose is to address the disruption of the metal shielding space in the cabin when the car door is opened, and to correct the Bluetooth propagation model based on this parameter; Door t Example values ​​are 1 (any door is open) and 0 (all doors are closed).

[0051] Eng t The meaning is the vehicle power operating level indicator output by the VCU, distinguishing between engine off and stationary, power on and standby, and driving operation; Eng t Its function is to distinguish the source of electromagnetic interference, which is continuously emitted by the motor and inverter after the vehicle starts. t Example values ​​are 0 (vehicle is off) and 2 (vehicle is running).

[0052] Vbat t The meaning is the real-time output voltage value of the vehicle's low-voltage battery; Vbat t The function of Vbat is to reduce the transmission power of the vehicle's Bluetooth when the battery is low; the voltage value is used for hardware power compensation. t Example values ​​are 12.6V (low voltage fully charged) and 11.2V (low voltage depleted, Bluetooth transmission power attenuated).

[0053] BtPara t This refers to the real-time data collected by the vehicle-mounted Bluetooth transceiver module, including transmit power, communication bit error rate, and occupied channel number; BtPara t Its function is to characterize whether the in-vehicle Bluetooth hardware itself is malfunctioning and to identify signal fluctuations caused by hardware failures; BtPara t Example values ​​are a bit error rate of 0.001 (hardware is working normally) and a bit error rate of 0.08 (hardware communication has packet loss anomalies).

[0054] RSSI t The meaning is the RSSI (Received Signal Strength Indication) value received by the vehicle's Bluetooth terminal from the digital key; RSSI t Its function is to determine the basic distance between the digital key and the vehicle, and it is a core basic indicator for location determination; RSSI t Example values ​​are -48dBm (high signal strength, key close to vehicle) and -82dBm (weak signal, key far from vehicle).

[0055] Delay t The meaning is the time difference between the vehicle's Bluetooth sending a detection command and the digital key returning a response; Delay t Its function is to distinguish between true long-distance signals and false weak signals caused by interference and obstruction, thus assisting in correcting distance judgments; Delay t Example values ​​are 14ms (communication link is unobstructed) and 68ms (link is obstructed or has strong interference).

[0056] KeyHw tThis refers to the built-in battery voltage and RF transmission level parameters transmitted back by the digital key; KeyHw t Its function is to identify the decrease in transmission power caused by low key battery, and to eliminate the judgment bias caused by the key hardware condition; KeyHw t Example values ​​are 3.1V battery voltage (key has sufficient power) and 2.2V battery voltage (key has low power consumption and attenuated transmission power).

[0057] Comm t This refers to the packet loss rate and channel switching frequency in the communication between the digital key and the vehicle's Bluetooth; Comm t Its function is to characterize the stability of the communication link; when the packet loss rate is too high, a single RSSIt has no reference value for judgment; Comm t Example values ​​are a packet loss rate of 0.002 (stable communication) and a packet loss rate of 0.15 (frequent link disconnection, low reliability of signal data).

[0058] S1022, Obtain the digital key's in-vehicle position probability discrimination model. Input the multimodal feature vector sequence into the digital key's in-vehicle position probability discrimination model to obtain the digital key's in-vehicle position probability.

[0059] In this embodiment, the digital key in-vehicle location probability discrimination model uses a one-dimensional convolutional neural network (1D-CNN) as its underlying architecture. This 1D-CNN consists of an input layer, a one-dimensional convolutional layer, a pooling layer, a memory network layer, a fully connected feature fusion layer, and a probability output layer connected in series. Model training uses training samples collected synchronously under multiple real-vehicle operating conditions and labeled with real location tags. The training sample collection scenarios cover multiple placement areas inside the vehicle, near-field and far-field locations outside the vehicle, and the critical boundary between inside and outside the vehicle. Simultaneously, various interference combinations are superimposed, including different in-vehicle electromagnetic interference intensities, various occlusion states, external wireless clutter, different vehicle start-stop conditions, different vehicle power supply levels, various operating states of the in-vehicle Bluetooth hardware, and different battery levels and communication conditions of the digital key. To address the problem of scarce extreme boundary samples, data augmentation is performed by superimposing compliant small-amplitude noise on continuous features, reorganizing various operating condition matching relationships, and fine-tuning the data time sequence window to supplement boundary samples and improve the model's generalization performance under complex interference conditions. The one-dimensional convolutional neural network (1D-CNN) is iteratively trained using the training samples processed above. Training stops when the loss function value tends to stabilize, thus completing the construction of the digital key car interior position probability discrimination model.

[0060] In this embodiment of the application, a gradient boosting decision tree model can be used, combined with the aforementioned training samples, to complete the model training and construction, thereby constructing a digital key in-vehicle location probability discrimination model suitable for in-vehicle scenarios.

[0061] In this embodiment, a probability discrimination model for the in-vehicle position of the digital key is first obtained. Then, a multimodal feature vector sequence is input into the model to obtain the in-vehicle position probability of the digital key. The in-vehicle position probability of the digital key is a scalar value between 0 and 1 output by the multimodal feature vector sequence obtained through temporal alignment and concatenation, after inference calculation by the model. This scalar value represents the credibility of the digital key being inside the vehicle cabin; the closer the value is to 1, the higher the credibility of the digital key being inside the vehicle; the closer the value is to 0, the higher the credibility of the digital key being outside the vehicle.

[0062] To perform deep reasoning on multimodal feature vector sequences using a digital key in-vehicle location probability discrimination model, the probability of the digital key's in-vehicle location is quantified, such as... Figure 3 As shown in Figure S1022, the specific steps for "inputting the multimodal feature vector sequence into the digital key vehicle interior position probability discrimination model to obtain the digital key vehicle interior position probability" are as follows: S10221, Extract local features from the multimodal feature vector sequence to obtain the feature map sequence.

[0063] In this embodiment, a one-dimensional convolutional layer is used to extract local features from the multimodal feature vector sequence to obtain a feature map sequence.

[0064] The process of extracting local features from a multimodal feature vector sequence using a one-dimensional convolutional layer to obtain a feature map sequence is as follows: The one-dimensional convolutional layer includes convolutional kernels of different sizes; the multimodal feature vector sequence is input into the one-dimensional convolutional layer, and each convolutional kernel performs a sliding dot product operation on the multimodal feature vector sequence along the time axis, outputting the corresponding sequence; all sequences are concatenated along the channel dimension to obtain the feature map sequence.

[0065] For example, a one-dimensional convolutional layer contains small convolutional kernels, medium convolutional kernels, and large convolutional kernels; for example, the size of the small convolutional kernel is 3, the size of the medium convolutional kernel is 5, and the size of the large convolutional kernel is 7. The small convolutional kernel, medium convolutional kernel, and large convolutional kernel perform a sliding dot product operation on the input multimodal feature vector sequence along the time axis to obtain the small convolutional kernel sequence, medium convolutional kernel sequence, and large convolutional kernel sequence. The small convolutional kernel sequence, medium convolutional kernel sequence, and large convolutional kernel sequence are concatenated along the channel dimension to generate a feature map sequence containing multi-scale temporal feature information.

[0066] Among them, small convolutional kernels, due to their extremely short receptive field, focus on capturing instantaneous fluctuations in Bluetooth signal strength and brief spikes caused by sudden environmental changes. Medium convolutional kernels respond to the gradual changes in the electromagnetic environment and the continuous rise or fall of the signal within a medium time window. Large convolutional kernels cover a longer time range and are used to extract the overall signal drift and long-term contextual dependencies caused by slow changes in the vehicle's operating conditions.

[0067] S10222, extract the effective features from the feature map sequence to obtain the initial compressed feature sequence.

[0068] In this embodiment, effective features are extracted from the feature map sequence using a pooling layer to obtain an initial compressed feature sequence. Effective features specifically refer to feature components in the feature map sequence that possess discriminative power and characterize the core changes in the business within the time-series dimension. These components are obtained by sampling the maximum value of each channel through a pooling window of limited size and step size sliding along the time axis. This approach not only fully preserves key temporal evolution information but also reduces the length of the feature sequence.

[0069] The process of extracting effective features from the feature map sequence using pooling layers to obtain the initial compressed feature sequence is as follows: the pooling window size and sliding step size in the pooling layer are the first size and the first step length; the feature map sequence is input into the pooling layer, the pooling window slides along the time dimension of the feature map sequence, and the maximum value of the features of each channel within the window coverage area is taken at each sliding position to obtain the initial compressed feature sequence, and the initial compressed feature sequence is output to the memory network layer.

[0070] For example, the pooling layer sets the pooling window size to 2 (the pooling window covers two consecutive time steps along the time axis) and the sliding stride to 2. The pooling window slides along the time dimension of the feature map sequence for calculation, while the channel dimension remains unchanged during the operation, fully preserving all channels output by the one-dimensional convolutional layer. For each sliding position of the pooling window, the maximum value selection operation is performed on each channel feature within the window coverage area, outputting the initial compressed feature sequence. After the above pooling processing, the time dimension length of the resulting initial compressed feature sequence is compressed to 50% of the feature map sequence, and the overall parameter count is reduced by 50% simultaneously. This effectively preserves the peak response features of the Bluetooth signal within a local time window while suppressing insignificant noise interference.

[0071] S10223, based on the temporal sequence of the initial compressed feature sequence and the corresponding sequence changes, the global temporal feature sequence is obtained.

[0072] In this embodiment, the memory network layer obtains a global temporal feature sequence based on the temporal sequence of the initial compressed feature sequence and the corresponding sequence changes.

[0073] The process by which the memory network layer obtains the global temporal feature sequence based on the temporal sequence of the initial compressed feature sequence and the corresponding sequence changes is as follows: The memory network layer includes N bidirectional long short-term memory networks. The initial compressed feature sequence is input into the memory network layer to obtain the global temporal feature sequence. The specific steps are as follows: The following steps are executed iteratively until the Nth first intermediate time-series feature sequence is obtained; The i-th first compressed feature sequence to be processed is input into the i-th bidirectional long short-term memory network to obtain the i-th first intermediate time-series feature sequence. The i-th first intermediate time-series feature sequence is the (i+1)-th first compressed feature sequence to be processed, the 1st first compressed feature sequence to be processed is the initial compressed feature sequence, and the Nth first intermediate time-series feature sequence is the global time-series feature sequence. i is 1, 2, 3, ..., N in sequence.

[0074] The bidirectional long short-term memory network includes a forward long short-term memory network layer and a reverse long short-term memory network layer. The i-th first compressed feature sequence to be processed is input into the i-th bidirectional long short-term memory network to obtain the i-th first intermediate temporal feature sequence. The specific steps are as follows: A forward long short-term memory network is used to perform a forward traversal of the i-th first compressed feature sequence to be processed in temporal order; and a reverse long short-term memory network is used to perform a reverse traversal of the i-th first compressed feature sequence to be processed in temporal order. During the forward traversal process, the forward long short-term memory network obtains the forward forget gate, forward input gate, forward candidate cell state, and forward output gate for each time step based on the compressed features to be processed at that time step. Based on the forward forget gate, forward input gate, and forward candidate cell state, the forward target cell state at that time step is obtained. Based on the forward target cell state and the forward output gate, the forward hidden state at that time step is obtained.

[0075] Specifically, Formula 3 is used to calculate the forward forget gate, Formula 4 is used to calculate the forward input gate, Formula 5 is used to calculate the forward candidate cell state, Formula 6 is used to calculate the forward output gate, Formula 7 is used to calculate the forward target cell state, and Formula 8 is used to calculate the forward hidden state.

[0076] (Formula 3), where, Represents the forward forgetting gate at time t. Represents the activation function. Represents the forward forget gate weight matrix. This represents the positive hidden state at time t-1. The compressed feature sequence to be processed at time t; Represents the positive forgetting gate bias term; (Formula 4), where, This represents the positive input gate at time t. Represents the forward input gate weight matrix. Represents the positive input gate bias term; (Formula 5), ​​where This represents the positive candidate cell state at time t. This represents the weight matrix of positive candidate cell states. Represents the bias term for the positive candidate cell state; (Formula 6); where, This represents the positive output gate at time t. Represents the forward output gate weight matrix. Represents the positive output gate bias term; (Formula 7); where, Represents the Hadamard product operator. This represents the positive target cell state at time t-1; This represents the positive target cell state at time t; (Formula 8); where, This represents the positive hidden state at time t.

[0077] In the reverse long short-term memory network, during the reverse traversal process, for each time step, based on the compressed features to be processed at that time step, the reverse forget gate, reverse input gate, reverse candidate cell state, and reverse output gate are obtained respectively. Based on the reverse forget gate, reverse input gate, and reverse candidate cell state, the reverse target cell state at that time step is obtained. Based on the reverse target cell state and the reverse output gate, the reverse hidden state at that time step is obtained. Specifically, Formula 9 is used to calculate the reverse forget gate, Formula 10 is used to calculate the reverse input gate, Formula 11 is used to calculate the reverse candidate cell state, Formula 12 is used to calculate the reverse output gate, Formula 13 is used to calculate the reverse target cell state, and Formula 14 is used to calculate the reverse hidden state.

[0078] (Formula 9), where, Represents the reverse forgetting gate at time t. Represents the activation function. Represents the weight matrix of the reverse forget gate. This represents the reverse hidden state at time t+1. The compressed feature sequence to be processed at time t; Represents the reverse forgetting gate bias term; (Formula 10), where, This represents the inverted input gate at time t. Represents the inverse input gate weight matrix. Represents the inverse input gate bias term; (Formula 11), where This represents the reverse candidate cell state at time t. Represents the reverse candidate cell state weight matrix. This represents the reverse candidate cell state bias term; (Formula 12); where, This represents the inverse output gate at time t. Represents the reverse output gate weight matrix; Represents the inverted output gate bias term; (Formula 13); where, Represents the Hadamard product operator. This represents the reversed target cell state at time t+1; This represents the reversed target cell state at time t; (Formula 14); where, This represents the reverse hidden state at time t.

[0079] The forward and reverse hidden states corresponding to the same time point are superimposed to obtain the hidden state at that time point; the hidden states corresponding to all times are arranged in their original time sequence order to obtain the i-th first intermediate time sequence feature sequence.

[0080] This global temporal feature sequence integrates forward and backward long-distance dependencies within the complete time window of the compressed feature sequence, enabling the model to fully capture all temporal context information.

[0081] S10224 fuses the feature information of the global time series feature sequence to obtain the deep feature vector.

[0082] In this embodiment, the fully connected feature fusion layer fuses feature information from the global temporal feature sequence to obtain a deep feature vector.

[0083] The fully connected feature fusion layer fuses feature information from the global temporal feature sequence to obtain a deep feature vector. First, the global temporal feature sequence is flattened into a one-dimensional feature vector. This one-dimensional feature vector is then fed into the first fully connected layer, which performs a linear transformation and adds a bias, and outputs intermediate features through activation function operations. The resulting intermediate features are then input into subsequent fully connected layers, which sequentially repeat the linear transformation, bias addition, and activation function operations to achieve feature propagation layer by layer. Finally, after the last fully connected layer completes the computation, the deep feature vector is output.

[0084] For example, the fully connected feature fusion layer contains two fully connected layers: the first fully connected layer has 128 neurons, and the second fully connected layer has 64 neurons. The global temporal feature sequence output from the memory network layer is input into the fully connected feature fusion layer. The global temporal feature sequence is first flattened into a one-dimensional feature vector. The flattened one-dimensional feature vector is then input into the first fully connected layer. The first fully connected layer performs a linear transformation on the input features and adds a bias, then outputs a 128-dimensional intermediate feature after passing through an activation function. The 128-dimensional intermediate feature is then input into the second fully connected layer. The second fully connected layer performs a linear transformation on the input features and adds a bias, then outputs a 64-dimensional deep feature vector after passing through an activation function.

[0085] A Dropout layer is configured between the two fully connected layers, with a dropout rate set to 0.2. The fully connected feature fusion layer, through the gradual abstraction and compression of multiple fully connected layers, performs deep fusion mapping of local detail information and global context information in the global temporal feature sequence, generating a fixed-dimensional deep feature vector, providing a fully abstract high-level semantic feature representation for the probability output layer.

[0086] S10225, calculates the probability of the in-vehicle location corresponding to the digital key based on deep feature vectors.

[0087] In this embodiment, the probability output layer calculates the probability of the in-vehicle location corresponding to the digital key based on deep feature vectors.

[0088] The process of calculating the probability of the in-vehicle position corresponding to the digital key based on the deep feature vector is as follows: the probability output layer performs nonlinear mapping calculation on the deep feature vector and outputs the probability of the in-vehicle position corresponding to the digital key.

[0089] For example, the probability output layer consists of a single-neuron fully connected layer and an activation function. The deep feature vector output from the fully connected feature fusion layer is input into the probability output layer. The deep feature vector undergoes a linear transformation through the single-neuron fully connected layer to obtain a logical value. The logical value is then non-linearly mapped by the activation function, outputting a continuous value between 0 and 1 as the probability of the digital key's location inside the vehicle. The closer the probability value is to 1, the higher the confidence level of the model in determining that the digital key is inside the vehicle; the closer it is to 0, the higher the confidence level of the model in determining that the digital key is outside the vehicle.

[0090] To avoid the judgment bias caused by using a single signal threshold to determine the position of the digital key and to improve the reliability of digital key position determination, such as Figure 4 As shown, the specific steps in S103 for "determining the actual location of the digital key based on the probability of its location inside the vehicle" are as follows: S1031, Obtain the set of location probability thresholds.

[0091] In this embodiment, a set of location probability thresholds is obtained. This provides a standardized and quantifiable basis for comparing and determining the location probability inside the digital key vehicle.

[0092] In the embodiments of this application, such as Figure 5 As shown, the specific steps for "obtaining the location probability threshold set" in S1031 are as follows: S10311, Obtain historical sample data, which includes location ground truth labels.

[0093] In this embodiment, historical sample data is acquired. This historical sample data includes multiple sets of samples. Each set of historical sample data includes a historical environmental data sequence, a historical vehicle state data sequence, a historical digital key data sequence, and a location truth label corresponding to that set of data. The historical environmental data sequence includes Bluetooth signal interference data, ambient temperature and humidity data, and scene type data. The historical vehicle state data sequence includes vehicle speed, door opening / closing status, vehicle gear position, and engine start / stop status. The historical digital key data sequence includes Bluetooth RSSI strength, RSSI rate of change, RSSI temporal variance, and mobile phone battery level data at different times. The location truth label is used to mark the actual location of the digital key corresponding to that set of samples. The truth labels are divided into two categories: in-vehicle truth labels and out-of-vehicle truth labels. Specifically, the location truth label is used to mark the actual location (inside or outside the vehicle) of the digital key in the historical sample data, representing the true category.

[0094] S10312, input historical sample data into the digital key vehicle interior position probability discrimination model to obtain the vehicle interior position prediction probability corresponding to the historical sample data.

[0095] In this embodiment, the historical environment data sequence, historical vehicle state data sequence, and historical digital key data sequence in each group of historical sample data are time-aligned and concatenated to generate a multimodal feature vector sequence. All multimodal feature vector sequences are sequentially input into the digital key in-vehicle position probability discrimination model. Feature operations are performed layer by layer through the model's input layer, one-dimensional convolutional layer, pooling layer, memory network layer, and fully connected feature fusion layer. Finally, a nonlinear mapping is performed through the probability output layer to output the in-vehicle position prediction probability corresponding to each group of historical sample data. The in-vehicle position prediction probability refers to a continuous value between 0 and 1 output by the digital key in-vehicle position probability discrimination model based on the multimodal feature vector sequence. This value represents the confidence level that the digital key is currently inside the vehicle; the closer the value is to 1, the higher the confidence level that the model determines the digital key is inside the vehicle; the closer the value is to 0, the higher the confidence level that the model determines the digital key is outside the vehicle.

[0096] S10313, generate a performance evaluation curve based on the predicted in-vehicle position probability and the true position label corresponding to all historical sample data.

[0097] In this embodiment, a performance evaluation curve is generated based on the predicted in-vehicle position probability and the true position label corresponding to all historical sample data. This performance evaluation curve allows for a clear quantification of the likelihood of false or false key position determinations at each probability threshold, facilitating a comprehensive balance between safety performance and digital key position detection sensitivity in vehicle control scenarios.

[0098] In the embodiments of this application, such as Figure 6 As shown, the specific steps in S10313, "generating a performance evaluation curve based on the predicted in-vehicle position probability and the true position label corresponding to all historical sample data," are as follows: S103131, set multiple candidate probability thresholds, and for any candidate probability threshold, compare the predicted probability of the in-vehicle position corresponding to each historical sample data with the candidate probability threshold to obtain the predicted position determination result corresponding to each historical sample data.

[0099] In this embodiment, specifically, within the value range of 0 to 1 for the predicted probability of the vehicle's interior location, multiple candidate probability thresholds with different values ​​are set, uniformly covering the complete numerical range of the predicted probability of the vehicle's interior location. For any candidate probability threshold, all historical sample data are iterated one by one, and the predicted probability of the vehicle's interior location corresponding to each set of historical sample data is compared with the current candidate probability threshold. If the predicted probability of the vehicle's interior location corresponding to the historical sample data is greater than or equal to the current candidate probability threshold, the predicted digital key location corresponding to that set of historical sample data is determined to be inside the vehicle; if the predicted probability of the vehicle's interior location corresponding to the historical sample data is less than the current candidate probability threshold, the predicted digital key location corresponding to that set of historical sample data is determined to be outside the vehicle. After completing the comparison and determination process for all historical sample data, the determination results of all digital key predicted locations corresponding to the current candidate probability threshold are obtained.

[0100] S103132, compare the predicted location determination results and location true value labels corresponding to each historical sample data, statistically obtain the number of true positives and false positives in the historical sample data, and calculate the true positive rate and false positive rate under the candidate probability threshold based on the number of true positives and false positives.

[0101] In this embodiment of the application, specifically, for the digital key prediction location determination results of all historical sample data corresponding to a single candidate probability threshold, the digital key prediction location determination results of each group of historical sample data are compared with the location truth labels corresponding to the historical sample data one by one, and the number of true positives and false positives corresponding to the current candidate probability threshold are statistically obtained. Based on the number of true positives and false positives, the true positive rate and false positive rate corresponding to the current candidate probability threshold are further calculated.

[0102] Among them, the true positive number refers to the number of samples in all historical sample data where the true location label is "inside the vehicle" and the digital key predicts the location as "inside the vehicle"; the false positive number refers to the number of samples in all historical sample data where the true location label is "outside the vehicle" and the digital key predicts the location as "inside the vehicle".

[0103] The true positive rate is used to characterize the accuracy of the digital key in-vehicle location probability discrimination model in identifying real in-vehicle samples. The true positive rate is calculated as the ratio of the number of true positives to the total number of historical sample data with true location labels of in-vehicle. The false positive rate is used to characterize the degree of misidentification of real out-of-vehicle samples by the digital key in-vehicle location probability discrimination model. The false positive rate is calculated as the ratio of the number of false positives to the total number of historical sample data with true location labels of out-of-vehicle.

[0104] S103133, with the false positive rate on the x-axis and the true positive rate on the y-axis, connect the coordinate points corresponding to the true positive rate and false positive rate obtained under all candidate probability thresholds in sequence to generate a performance evaluation curve.

[0105] In this embodiment, after obtaining the false positive rate and true positive rate values ​​corresponding to all candidate probability thresholds, the performance evaluation curve is generated. First, a two-dimensional coordinate system is established, with the horizontal axis defined as the false positive rate and the vertical axis as the true positive rate. Second, for each candidate probability threshold, the corresponding false positive rate and true positive rate values ​​are used as a set of coordinate parameters, and a unique coordinate point is marked in the two-dimensional coordinate system, completing the marking of coordinate points corresponding to all candidate probability thresholds. Finally, according to the ascending order of candidate probability thresholds, all coordinate points in the two-dimensional coordinate system are smoothly connected sequentially to generate a complete performance evaluation curve.

[0106] The performance evaluation curve is a two-dimensional performance curve constructed with the false positive rate on the horizontal axis and the true positive rate on the vertical axis. The performance evaluation curve is composed of the coordinate points corresponding to all candidate probability thresholds connected in sequence. The performance evaluation curve is used to comprehensively and visually characterize the overall classification and recognition performance of the digital key vehicle in-vehicle location probability discrimination model under different probability judgment criteria. It can intuitively reflect the accuracy and degree of misjudgment of the digital key vehicle in-vehicle location judgment.

[0107] For example, historical sample data is shown in Table 1:

[0108] Table 1 Statistical analysis was performed on the historical sample data in Table 1. The total number of samples was 10, with 5 true positives and 5 true negatives. Candidate probability thresholds were set from low to high as follows: 0.1, 0.3, 0.5, 0.6, 0.8, and 0.9. Coordinate points were calculated for each candidate threshold, as shown in Table 2.

[0109]

[0110] Table 2 In Table 2, TP represents the number of true positives: true value = number of samples inside the vehicle and predicted = number of samples inside the vehicle; FP represents the number of false positives: true value = number of samples outside the vehicle and predicted = number of samples inside the vehicle. TPR represents the true positive rate, and FPR represents the false positive rate. Performance evaluation curves are generated based on the coordinate points in Table 2; the generated performance evaluation curves are shown below. Figure 7 As shown, the horizontal axis represents the false positive rate, and the vertical axis represents the true positive rate.

[0111] S10314, Based on the performance evaluation curve, determine the first position probability threshold and the second position probability threshold; combine the first position probability threshold and the second position probability threshold to obtain the position probability threshold set.

[0112] In this embodiment, a first position probability threshold and a second position probability threshold are determined based on a performance evaluation curve. The first and second position probability thresholds are then combined to obtain a position probability threshold set. Using the performance evaluation curve for dual-threshold screening allows for the use of quantitative data as the basis for parameter selection, avoiding biases caused by manually setting thresholds based on experience, and simultaneously reducing the probability of misjudgments and missed judgments in digital key position determination. The position probability threshold set, composed of the first and second position probability thresholds, can divide the determination into in-vehicle, out-of-vehicle, and hesitant determination intervals, mitigating the problem of frequent switching of position determination results caused by environmental interference and continuously improving the stability of digital key position recognition results.

[0113] In the embodiments of this application, such as Figure 8 As shown, the specific steps of S10314 "Determine the first position probability threshold and the second position probability threshold based on the performance evaluation curve" are as follows: S103141, Obtain the business requirement type, and determine the upper limit of the false positive rate and the lower limit of the true positive rate based on the business requirement type.

[0114] In this embodiment, the current vehicle application scenario is identified based on vehicle environmental data sequences, vehicle status data sequences, and digital key data sequences. For example, when the environmental data sequence indicates that the electromagnetic interference intensity inside the vehicle is 10~18dBμV / m, the vehicle body occlusion level is 0, and the external wireless noise interference is 2~5dBm, and the vehicle status data sequence indicates that all four doors are fully closed, the vehicle is turned off and stationary, the battery has sufficient power, and the vehicle Bluetooth hardware communication status is stable, while the key data sequence indicates that the Bluetooth signal strength RSSI is -78~-90dBm, the communication latency is 55~80ms, the key hardware has sufficient power, and the long-distance communication packet loss rate is relatively high, the current vehicle application scenario is determined to be a parking and locking security priority scenario. When the environmental data sequence indicates that the electromagnetic interference intensity is 25~38dBμV / m, the interior is slightly occluded, and multiple wireless devices around the garage generate 6~10dBm of external radio frequency interference, and the vehicle status... When the data sequence indicates that the driver's door is opened and then closed, the vehicle is off and in standby mode, the battery voltage is normal, the vehicle Bluetooth only slightly switches communication channels, and the Bluetooth signal strength RSSI of the key data sequence is between -42 and -60 dBm, the communication latency is between 12 and 22 ms, and the packet loss rate is extremely low, the current vehicle application scenario is determined to be a scenario prioritizing in-vehicle start sensitivity. When the environmental data sequence is superimposed with medium to high intensity electromagnetic interference and radio frequency interference formed by shopping mall equipment and surrounding vehicles, and the metal frame of the car window causes moderate signal obstruction, and the vehicle status data sequence indicates that the four doors are closed, the vehicle is off, the battery voltage fluctuates slightly, the bit error rate of the vehicle Bluetooth hardware is slightly increased, and the Bluetooth signal strength RSSI of the key data sequence is between -62 and -70 dBm, the communication latency is between 30 and 45 ms, and the packet loss rate is in the critical range of 0.05 to 0.1, the current vehicle application scenario is determined to be a critical transition scenario that balances recognition sensitivity and anti-false judgment capability.

[0115] In this embodiment of the application, a vehicle application scenario is obtained; based on a preset mapping relationship between vehicle application scenarios and business requirement types, the business requirement type corresponding to the vehicle application scenario is determined; specifically, the vehicle application scenario is matched in the preset mapping relationship between vehicle application scenarios and business requirement types, and the successfully matched business requirement type is taken as the business requirement type corresponding to the vehicle application scenario.

[0116] For example, based on the preset mapping relationship between vehicle application scenarios and business requirement types, the following can be matched: parking lock safety priority scenario corresponds to the external lock anti-false judgment requirement; in-vehicle start sensitivity priority scenario corresponds to the power-on start high sensitivity requirement; and critical transition scenario that balances recognition sensitivity and anti-false judgment capability corresponds to the critical interval balance compatibility requirement.

[0117] In this embodiment, the business requirement type is obtained, and based on the business requirement type, an upper limit for the false positive rate and a lower limit for the true positive rate are determined. Specifically, the business requirement type is matched against a preset mapping relationship between the upper limit for the false positive rate, the lower limit for the true positive rate, and the business requirement type. The successfully matched upper limit for the false positive rate and the lower limit for the true positive rate are used as the upper limit for the false positive rate and the lower limit for the true positive rate corresponding to the business requirement type. The business requirement type is a scenario classification identifier obtained based on the vehicle application scenario and used to determine the upper limit for the false positive rate and the lower limit for the true positive rate.

[0118] For example, the preset upper limit of false positive rate and lower limit of true positive rate are mapped to the business requirement type. The record shows that "external locking anti-misjudgment requirement" corresponds to "false positive rate upper limit 0.05, true positive rate lower limit 0.85", so as to strictly control the risk of external locking being mistaken for internal locking and ensure parking locking safety; "high sensitivity requirement for power-on start" corresponds to "false positive rate upper limit 0.25, true positive rate lower limit 0.98", which improves the convenience of starting the vehicle; "critical range balanced compatibility requirement" corresponds to "false positive rate upper limit 0.15, true positive rate lower limit 0.92", which takes into account the recognition sensitivity and anti-misjudgment capability and adapts to transitional scenarios with critical signals and complex operating conditions. Based on the preset mapping relationship between the upper limit of the false positive rate, the lower limit of the true positive rate, and the business requirement type, the following results were obtained: the upper limit of the false positive rate for the requirement of preventing false judgment of vehicle external locking is 0.05, and the lower limit of the true positive rate is 0.85; the upper limit of the false positive rate for the requirement of high sensitivity upon power-on start is 0.25, and the lower limit of the true positive rate is 0.98; the upper limit of the false positive rate for the requirement of balanced compatibility in the critical interval is 0.15, and the lower limit of the true positive rate is 0.92.

[0119] S103142, select the node with the highest false positive rate and the highest true positive rate on the performance evaluation curve, and determine the candidate probability threshold corresponding to the node as the first position probability threshold.

[0120] In this embodiment of the application, specifically, the performance evaluation curve is retrieved, all nodes contained in the performance evaluation curve are traversed, and all nodes with a false positive rate not greater than the upper limit of the false positive rate are selected. Then, the node with the largest true positive rate value is selected from the selected nodes, the candidate probability threshold corresponding to the node is extracted, and the candidate probability threshold is set as the first position probability threshold. The first position probability threshold refers to the baseline probability value used to determine that the digital key is in the vehicle.

[0121] S103143, select the node with the lowest false positive rate and the true positive rate on the performance evaluation curve, and determine the candidate probability threshold corresponding to the node as the second position probability threshold.

[0122] In this embodiment of the application, specifically, the performance evaluation curve is retrieved, all nodes contained in the performance evaluation curve are traversed, and all nodes with a true positive rate not less than the lower limit of the true positive rate are selected. Then, the node with the smallest false positive rate value is selected from the selected nodes, the candidate probability threshold corresponding to the node is extracted, and the candidate probability threshold is set as the second position probability threshold. The second position probability threshold refers to the baseline probability value used to determine whether the digital key is outside the vehicle.

[0123] S1032, compare the probability of the position inside the vehicle with the set of position probability thresholds to obtain the probability comparison result.

[0124] In this embodiment of the application, specifically, the probability of the digital key's position inside the vehicle is compared with the first position probability threshold and the second position probability threshold one by one, and the numerical relationship between the digital key's position inside the vehicle and the first position probability threshold, and the numerical relationship between the digital key's position inside the vehicle and the second position probability threshold is fully recorded. Based on all the numerical comparison information, the corresponding probability comparison result is generated.

[0125] S1033, based on the probability comparison results, obtain the actual location of the digital key.

[0126] In this embodiment of the application, specifically, when the probability of the position inside the vehicle is greater than or equal to the first position probability threshold, the actual position of the digital key is determined to be inside the vehicle; when the probability of the position inside the vehicle is less than or equal to the second position probability threshold, the actual position of the digital key is determined to be outside the vehicle; when the probability of the position inside the vehicle is less than the first position probability threshold but greater than the second position probability threshold, the actual position of the digital key is determined to be a hesitant position.

[0127] S1034 enables vehicle control based on actual location.

[0128] In this embodiment of the application, specifically, when it is determined that the actual location of the digital key is inside the vehicle, the user permissions are set to a first permission set, and the vehicle is controlled according to the first permission set; wherein the first permission set is a set of vehicle control permissions corresponding to the scenario where the digital key is inside the vehicle, and the first permission set specifically includes vehicle start control permissions, in-vehicle device adjustment permissions, and vehicle driving parameter setting permissions.

[0129] When it is determined that the actual location of the digital key is outside the vehicle, the user permissions are set to the second permission set, and the vehicle is controlled according to the second permission set. The second permission set is the set of vehicle control permissions corresponding to the scenario where the digital key is outside the vehicle. The second permission set specifically includes vehicle door unlocking permission, vehicle door locking permission, and vehicle start control permission and vehicle driving control permission are disabled.

[0130] When the actual location of the digital key is determined to be a hesitant location, a digital key location confirmation prompt instruction is sent to the user to request the user to confirm whether the digital key is currently inside or outside the vehicle; in response to the user confirming that the digital key is inside the vehicle, the user's permissions are set to a first permission set, and the vehicle is controlled according to the first permission set; or, in response to the user confirming that the digital key is outside the vehicle, the user's permissions are set to a second permission set, and the vehicle is controlled according to the second permission set.

[0131] In this embodiment, a hesitant position is set to define scenarios where the probability of the digital key's position inside the vehicle falls within the critical range between a first position probability threshold and a second position probability threshold, making it impossible to accurately determine whether the digital key is inside or outside the vehicle. This avoids the forced misjudgment problem caused by complex in-vehicle environment disturbances in a single threshold judgment method and compensates for the scenario adaptation defects of the fixed threshold discrimination mechanism. The core technical advantage of setting a hesitant position is that it eliminates the forced classification judgment logic of the critical probability data and accurately calibrates the actual position of the digital key based on the user's active location confirmation result. This avoids the problem of vehicle controllable functions being disabled and poor user experience caused by the digital key being misjudged as being outside the vehicle when it is inside the vehicle, and also prevents the problem of abnormally open vehicle control permissions and vehicle use security risks caused by the digital key being misjudged as being inside the vehicle when it is outside the vehicle. It simultaneously ensures the discrimination accuracy of the digital key position discrimination model, vehicle use security, and user experience.

[0132] In this embodiment, in response to the user confirming the location of the digital key, the environmental data sequence, state data sequence, and key data sequence are associated with the user-confirmed location to generate incremental samples. Based on the incremental samples, the digital key in-vehicle location probability discrimination model is updated using an incremental learning method. This continuously accumulates effective labeled sample data in real-world in-vehicle scenarios, compensating for the shortcomings of traditional fixed-parameter models in adapting to dynamic and complex in-vehicle scenarios. It continuously optimizes the internal discrimination parameters and model fitting performance of the digital key in-vehicle location probability discrimination model, effectively reducing the probability of misjudging the digital key location under complex in-vehicle conditions, and steadily improving the scenario adaptability and location discrimination accuracy of the digital key in-vehicle location probability discrimination model.

[0133] In this embodiment, under standard vehicle usage conditions where the special control mode is not enabled, the first permission set corresponding to the digital key being inside the vehicle has a higher overall permission level than the second permission set corresponding to the digital key being outside the vehicle.

[0134] The standard permission logic is adapted to scenarios such as single-person commuting, adults riding with the digital key, and temporary rest stops while the vehicle is parked. When the owner enters the cabin with the digital key, the vehicle system determines that the digital key is inside the vehicle and loads the first permission set, fully granting all vehicle control permissions, including power-on start, vehicle driving, in-depth adjustment of in-vehicle devices, and customization of driving parameters. When the owner leaves the vehicle with the digital key, the vehicle system determines that the digital key is outside the vehicle and automatically switches to the second permission set, granting only basic access permissions such as door unlocking, door locking, and trunk opening, while blocking all high-risk operation permissions related to power-on and vehicle driving. This differentiated permission layering balances the convenience of daily use with vehicle anti-theft security.

[0135] This application configures multiple special control scenarios. Within these scenarios, the hierarchical relationship between the two sets of permissions is reversed. The first permission set corresponding to the digital key determining the person is inside the vehicle has a lower permission level than the second permission set corresponding to the person determining the person is outside the vehicle. Typical scenarios include child care mode and driving school instructor mode. Before operating a special control scenario, the vehicle monitoring and restriction logic must be activated in advance. The standard operating condition for a special control scenario is that the guardian holds the digital key and is outside the vehicle, while a minor or student remains alone inside. The first permission set matched to the person inside the vehicle only enables basic static comfort functions such as air conditioning ventilation and audio-visual playback, completely disabling driving permissions that could pose safety risks, such as power-on, starting the engine, and shifting gears. The second permission set activated when the guardian holds the digital key and is outside the vehicle incorporates remote vehicle control capabilities, supporting advanced control functions such as remote power cut-off, emergency locking, remote window control, and real-time monitoring of the vehicle's interior status. The control priority of the second permission set corresponding to the person determining the person is outside the vehicle is significantly higher than the first permission set matched to the person without a digital key inside the vehicle.

[0136] Two sets of permission hierarchy rules can be dynamically switched based on the vehicle's operating mode: When the vehicle is powered on and the monitoring restriction mode or instructor restriction mode is not activated, the in-vehicle system defaults to the conventional logic that the user's first permission set (inside the vehicle) is higher than the second permission set (outside the vehicle) determined by the digital key. When the user activates the monitoring restriction mode or instructor restriction mode via the in-vehicle system, the digital key's accompanying app, or in-vehicle voice commands, the in-vehicle system automatically switches to the permission inversion control logic. Each time the in-vehicle system executes special permission determination logic, it simultaneously collects real-time environmental data sequences, vehicle status data sequences, and digital key communication data. Combined with the current permission determination result, it generates incremental labeled samples. These incremental labeled samples are input into the digital key's in-vehicle position probability discrimination model to complete incremental iterative training, continuously optimizing the probability determination thresholds for different business scenarios and reducing vehicle permission control anomalies caused by misjudgments of positions within critical probability intervals.

[0137] Corresponding to the above method embodiments, this application also provides a vehicle control device, such as... Figure 9As shown, the device may include: a data acquisition module 901, a probability output module 902, and a vehicle control module 903.

[0138] The data acquisition module 901 is used to acquire the vehicle's environmental data sequence, status data sequence, and digital key key data sequence.

[0139] The probability output module 902 is used to determine the probability of the digital key's position inside the vehicle based on the environmental data sequence, the state data sequence, and the key data sequence.

[0140] The vehicle control module 903 is used to determine the actual location of the digital key based on the in-vehicle location probability; and to implement vehicle control based on the actual location.

[0141] This application also provides a vehicle, such as... Figure 10 As shown, it includes a processor 1001, a communication interface 1002, a memory 1003, and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other through the communication bus 1004. Memory 1003 is used to store computer programs; When processor 1001 executes a program stored in memory 1003, it performs the following steps: The system acquires a sequence of environmental data, a sequence of state data, and a sequence of key data for the digital key; determines the probability of the digital key's position inside the vehicle based on the environmental data sequence, the state data sequence, and the key data sequence; determines the actual position of the digital key based on the probability of its position inside the vehicle; and controls the vehicle based on the actual position.

[0142] The communication bus mentioned in the above vehicles can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0143] The communication interface is used for communication between the aforementioned vehicle and other devices.

[0144] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0145] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0146] In another embodiment provided in this application, a storage medium is also provided, which stores instructions that, when run on a computer, cause the computer to execute any of the vehicle control methods described in the above embodiments.

[0147] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform any of the vehicle control methods described in the above embodiments.

[0148] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted from one storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0149] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0150] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0151] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A vehicle control method, characterized in that, The method includes: Acquire the vehicle's environmental data sequence, status data sequence, and digital key key data sequence; Based on the environmental data sequence, the state data sequence, and the key data sequence, determine the probability of the digital key's position inside the vehicle; Based on the probability of the location inside the vehicle, the actual location of the digital key is determined; and based on the actual location, the vehicle is controlled.

2. The method according to claim 1, characterized in that, Determining the probability of the digital key's in-vehicle location based on the environmental data sequence, the state data sequence, and the key data sequence includes: The environmental data sequence, the state data sequence, and the key data sequence are aligned and concatenated according to time sequence to obtain a multimodal feature vector sequence; Local features are extracted from the multimodal feature vector sequence to obtain a feature map sequence; Extract the effective features from the feature map sequence to obtain the initial compressed feature sequence; Based on the temporal sequence of the initial compressed feature sequence and the sequence changes corresponding to the temporal sequence, a global temporal feature sequence is obtained; The global temporal feature sequence is fused with feature information to obtain a deep feature vector; The probability of the vehicle's interior location corresponding to the digital key is calculated based on the deep feature vector.

3. The method according to claim 1, characterized in that, Determining the actual location of the digital key based on the in-vehicle location probability includes: Obtain the set of location probability thresholds; The probability of the in-vehicle location is compared with the set of location probability thresholds to obtain the probability comparison result; Based on the probability comparison results, the actual location of the digital key is obtained.

4. The method according to claim 3, characterized in that, The set of location probability thresholds includes: Acquire historical sample data, which includes location ground truth labels; Determine the predicted probability of the in-vehicle position corresponding to the historical sample data; Based on the predicted in-vehicle location probabilities corresponding to all the historical sample data and the true location labels, a performance evaluation curve is generated. Based on the performance evaluation curve, a first position probability threshold and a second position probability threshold are determined; the first position probability threshold and the second position probability threshold are combined to obtain the position probability threshold set.

5. The method according to claim 4, characterized in that, The step of generating a performance evaluation curve based on the predicted in-vehicle position probabilities corresponding to all the historical sample data and the ground truth labels of the positions includes: Multiple candidate probability thresholds are set. For any one of the candidate probability thresholds, the predicted probability of the in-vehicle position corresponding to each of the historical sample data is compared with the candidate probability threshold to obtain the predicted position determination result corresponding to each of the historical sample data. The predicted location determination results and location truth labels corresponding to each of the historical sample data are compared, and the number of true positives and false positives in the historical sample data are statistically obtained. The true positive rate and false positive rate under the candidate probability threshold are calculated based on the number of true positives and the number of false positives. Using the false positive rate as the horizontal axis and the true positive rate as the vertical axis, the coordinate points corresponding to the true positive rate and false positive rate obtained under all candidate probability thresholds are connected sequentially to generate the performance evaluation curve.

6. The method according to claim 5, characterized in that, The step of determining the first position probability threshold and the second position probability threshold based on the performance evaluation curve includes: Obtain the business requirement type, and determine the upper limit of the false positive rate and the lower limit of the true positive rate based on the business requirement type; On the performance evaluation curve, select the node with the highest true positive rate and a false positive rate that does not exceed the upper limit of the false positive rate, and determine the candidate probability threshold corresponding to the node as the first position probability threshold. On the performance evaluation curve, select the node with the lowest false positive rate and the true positive rate that is not lower than the lower limit of the true positive rate. Then, determine the candidate probability threshold corresponding to the node as the second position probability threshold.

7. The method according to claim 6, characterized in that, The types of business requirements obtained include: Obtain vehicle application scenarios; Based on the preset mapping relationship between vehicle application scenarios and business requirement types, the business requirement type corresponding to the vehicle application scenario is determined.

8. The method according to claim 3, characterized in that, The set of location probability thresholds includes a first location probability threshold and a second location probability threshold; The step of obtaining the actual location of the digital key based on the probability comparison result includes: When the probability of the location inside the vehicle is greater than or equal to the first location probability threshold, the actual location of the digital key is determined to be inside the vehicle. When the probability of the position inside the vehicle is less than or equal to the second position probability threshold, the actual position of the digital key is determined to be outside the vehicle. When the probability of the position inside the vehicle is less than the first position probability threshold and greater than the second position probability threshold, the actual position of the digital key is determined to be a hesitant position.

9. The method according to claim 8, characterized in that, The method of controlling the vehicle based on the actual location includes: When it is determined that the actual location of the digital key is inside the vehicle, the user permissions are set to the first permission set, and the vehicle is controlled according to the first permission set; When it is determined that the actual location of the digital key is outside the vehicle, the user permissions are set to the second permission set, and the vehicle is controlled according to the second permission set; When the actual location of the digital key is determined to be a hesitant location, a digital key location confirmation prompt instruction is sent to the user to request the user to confirm whether the digital key is currently inside or outside the vehicle; in response to the user confirming that the digital key is inside the vehicle, the user's permissions are set to a first permission set, and the vehicle is controlled according to the first permission set; or, in response to the user confirming that the digital key is outside the vehicle, the user's permissions are set to a second permission set, and the vehicle is controlled according to the second permission set.

10. A vehicle, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the vehicle control method according to any one of claims 1-9.

Images