Vehicle lock control method based on trajectory prediction and intention recognition and related device

Abstract

The application discloses a kind of based on trajectory prediction and intention recognition car lock control method and related equipment, it is related to car lock control field.The method includes: through target vehicle bluetooth module acquisition mobile terminal's bluetooth signal time series data, and receive inertial measurement data;Based on bluetooth data calculation dynamic characteristic value (slope and variance), based on inertial data generates first judgment result;Input finite state machine generates second judgment result (far away state, close state, in-car state);Compare results, conflict when effective output is selected according to arbitration logic, otherwise second result is used;Update user state, by far away state and close state trigger unlock instruction;When close state or in-car state is converted to far away state and continues to be stable and exceeds preset time threshold, trigger locking instruction.The application can accurately predict user motion trajectory, and distinguish the real leaving intention of user and the temporary, non-leaving nature activity in the periphery of vehicle, avoid false locking caused by intention misjudgment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of door lock access and control, and particularly relates to a vehicle lock control method based on trajectory prediction and intention recognition and related equipment. BACKGROUND

[0002] In recent years, with the popularity of Bluetooth communication technology in the automotive field, the Phone-as-a-Key (PaaK) system based on mobile terminals such as smart phones has been widely used. The system allows users to automatically unlock and lock the vehicle without a physical key when carrying an authorized mobile terminal, greatly improving the convenience of use.

[0003] At present, the mainstream PaaK system mainly adopts a short-range induction scheme based on Received Signal Strength Indication (RSSI). The typical working logic is: when the system detects that the RSSI value of the mobile terminal crosses a preset unlocking threshold, it is judged that the user is close and an unlocking operation is performed; when the RSSI value is lower than a preset locking threshold, it is judged that the user is far away and a locking (locking the vehicle) operation is performed.

[0004] However, this control logic which relies on threshold judgment of single-time signal strength has inherent technical defects. In actual use scenarios, the user's behavior is complex and continuous. For example, after getting off the vehicle, the user may not immediately walk away in a straight line, but may stay around the vehicle for a short time, such as checking the vehicle door, walking to the driver's cabin after taking or putting things from the trunk, or talking with people near the vehicle. In these scenarios, although the user's mobile terminal has temporarily left the signal area that triggers the locking, the user's true intention is not to lock the vehicle. At this time, the existing system will incorrectly judge the user's intention due to the instantaneous signal being lower than the threshold, and perform a locking operation that is not intended by the user. SUMMARY

[0005] In view of the above technical problems and defects, the purpose of the present application is to provide a vehicle lock control method based on trajectory prediction and intention recognition and related equipment, which can accurately distinguish the user's true intention to leave and the temporary and non-leaving nature of the activities around the vehicle, and avoid false locking caused by intention misjudgment.

[0006] To achieve the above object, in a first aspect, the application provides a vehicle lock control method based on trajectory prediction and intention recognition, comprising: continuously collecting, by a Bluetooth module of a target vehicle, Bluetooth signal characteristic parameters from a mobile terminal at a preset frequency to form Bluetooth signal time series data; receiving real-time collected inertial measurement data from the mobile terminal; calculating a dynamic characteristic value for representing a movement process of the mobile terminal relative to the target vehicle based on the Bluetooth signal time series data, the dynamic characteristic value including a slope and a variance of the Bluetooth signal time series; generating a first judgment result for representing an absolute motion state of the mobile terminal itself based on the inertial measurement data; inputting the dynamic characteristic value and the first judgment result into a preset finite state machine, and generating a second judgment result about a user state based on the dynamic characteristic value, the user state including a far-away state, a close-in state and an in-vehicle state; comparing the first judgment result and the second judgment result to obtain a comparison result; if the comparison result meets a preset conflict condition, selecting the first judgment result or the second judgment result as an effective output according to a preset arbitration logic; if the comparison result does not meet the preset conflict condition, adopting the second judgment result as the effective output; updating a current user state of the mobile terminal according to the effective output; when the updated user state of the finite state machine is a transition from the far-away state to the close-in state, triggering an unlocking instruction to make the target vehicle in an unlocked state; when the updated user state of the finite state machine is a transition from the close-in state or the in-vehicle state to the far-away state, and the far-away state lasts stably for more than a preset time threshold, triggering a locking instruction to make the target vehicle in a locked state.

[0007] The application upgrades the judgment and prediction of the user position and motion trajectory from the analysis of instantaneous "point" to the understanding of continuous "process" by collecting the time series data of the Bluetooth signal and calculating the slope and variance and other dynamic characteristics. More importantly, the application innovatively introduces the inertial measurement data of the mobile terminal as a second information source to independently judge the absolute motion state of the user. By comparing and intelligently arbitrating the judgment results from the two sources, the application can effectively identify and correct the misjudgment caused by signal fluctuation or the complex behavior of the user lingering around the vehicle. Finally, by adding a time threshold to the locking instruction, the application ensures that the locking is executed only after the intention of the user to leave continuously and stably is confirmed, thereby fundamentally solving the frequent misoperation problem of the prior art in complex scenarios, realizing the accurate prediction of the user motion trajectory, accurately distinguishing the real leaving intention of the user from the temporary and non-leaving activities around the vehicle, avoiding the mislocking caused by intention misjudgment, and significantly improving the intelligent level and reliability of the vehicle lock control system.

[0008] Optionally, in some embodiments, based on the inertial measurement data, a first judgment result is generated for characterizing an absolute motion state of the mobile terminal itself, including: extracting acceleration and angular velocity data from the inertial measurement data; based on the acceleration and angular velocity data, calculating a real-time motion trajectory of the mobile terminal in a three-dimensional space through a pose solving algorithm; and according to a projection component of the real-time motion trajectory on a horizontal plane, generating the first judgment result for indicating whether the mobile terminal is approaching the target vehicle, moving away from the target vehicle, or lingering around the target vehicle.

[0009] The technical solution of the above embodiment provides a specific implementation of converting original inertial measurement data into a high-level motion state description that can be directly used for intention judgment. Through pose solving and trajectory projection, the abstract acceleration and angular velocity data can be accurately analyzed into intuitive physical motion trajectories such as "approaching", "moving away", or "lingering". This makes the first judgment result no longer limited to simple "moving" or "stationary", but provides a more informative motion trend description that matches the judgment dimension of the Bluetooth signal, greatly enhancing the accuracy and effectiveness of subsequent comparison and arbitration.

[0010] Optionally, in some embodiments, the arbitration logic includes: when a preset conflict condition is that the first judgment result indicates that the mobile terminal is in a lingering state and the second judgment result indicates that the user state has been converted to a moving away state, then the first judgment result is selected as the valid output; and when the preset conflict condition is that the first judgment result indicates that the mobile terminal is in a stable approaching or stable moving away state, and the second judgment result indicates that the user state is a stationary state outside the vehicle because the variance of the Bluetooth signal time sequence continues to be higher than the variance threshold, then the first judgment result is selected as the valid output.

[0011] The technical solution of the above embodiment provides clear and specific decision rules for the arbitration logic, significantly enhancing the decision robustness of the system in critical conflict scenarios. The technical solution presets two most typical conflict situations and provides the optimal solution based on common sense in the physical world. For example, when the Bluetooth signal misjudges that the user has left due to interference, but the inertial measurement data accurately identifies that the user is still lingering, the system can make the correct judgment according to the rules. This rule-based arbitration based on expert knowledge makes the system's decision no longer ambiguous, but stable and reliable, effectively avoiding misoperation in critical scenarios.

[0012] Optionally, in some embodiments, the arbitration logic is a dynamic weight arbitration logic, and the execution of the dynamic weight arbitration logic comprises: calculating a data credibility weight of the inertial measurement data and a data credibility weight of the Bluetooth signal time series data in real time, wherein the data credibility weight of the inertial measurement data is calculated based on the signal noise level and the zero bias stability, and the data credibility weight of the Bluetooth signal time series data is calculated based on the signal-to-noise ratio and the multipath effect interference degree; and weighting and fusing the first judgment result and the second judgment result with the corresponding data credibility weights respectively to generate a final user state judgment after fusion, and taking the final user state judgment as an effective output.

[0013] The technical solution of the above embodiment promotes the arbitration logic from a fixed rule to a dynamic adaptive level. By evaluating the signal quality of the two data sources in real time and calculating the credibility weight, the system can intelligently adjust the degree of trust in different sensors according to the current electromagnetic environment and user motion state. When the Bluetooth signal is stable, the system can focus more on the Bluetooth judgment; and when the signal interference is serious, the system can rely more on the inertial measurement. This dynamic weight fusion mechanism makes the arbitration logic more flexible and accurate, and can always maintain optimal judgment performance in various complex and variable environments.

[0014] Optionally, in some embodiments, the user state further comprises an out-of-vehicle stationary state, and the method further comprises: when the variance of the Bluetooth signal time series data is continuously higher than a preset variance threshold in the approaching state or the away state, judging that the current user state is the out-of-vehicle stationary state, and delaying the triggering of the car lock control instruction.

[0015] The technical solution of the above embodiment introduces a new "out-of-vehicle stationary state" to the system, which introduces a "buffer" mechanism for handling uncertainty. When the variance of the Bluetooth signal is too large, indicating that the signal quality is unreliable or the user behavior is unstable (such as wandering), the system will enter this specific state and pause decision-making. This design avoids making wrong judgments in ambiguous information, and significantly improves the stability and fault tolerance of the entire control system by actively waiting for clearer signals or user intentions, preventing unexpected car lock actions caused by temporary signal fluctuations.

[0016] Optionally, in some embodiments, the time threshold is a variable value calculated dynamically, and the step of dynamically calculating comprises: when the user state is converted from the approaching state or the in-vehicle state to the away state, extracting the slope stability and the slope absolute value of the Bluetooth signal time series data before the state conversion; quantitatively evaluating the decisiveness of the one-off departure behavior based on the slope stability and the slope absolute value; and setting the time threshold according to the decisiveness, wherein the time threshold and the decisiveness have an inverse relationship, so that a high decisiveness corresponds to a shorter time threshold, and a low decisiveness corresponds to a longer time threshold.

[0017] By introducing the dynamically calculated time threshold, the locking logic is more intelligent and user-friendly. The method can quantify the "resolute degree" of the user by analyzing the speed and stability when the user leaves, and adjust the waiting time before locking accordingly. For a decisive user who leaves, the system will quickly respond to execute the locking; while for a user who may only be temporarily away, the system will provide a longer buffer time. This adaptive delay strategy makes the system response accurately match the user's behavior intention, greatly optimizing the user experience.

[0018] Optionally, in some embodiments, after the user state is converted from the approaching state or the in-vehicle state to the away state, and before the locking instruction is triggered, the method further comprises: starting the countdown of the time threshold in the case that the locking control system enters a locking preparation state; during the duration of the locking preparation state, calculating a return intention index for quantifying the possibility of the user's return based on the latest collected signal time series data; judging whether the timing value of the timer reaches the time threshold; if the timing value reaches the time threshold and the return intention index does not exceed a preset cancellation threshold since entering the locking preparation state, triggering the locking instruction; if the return intention index exceeds the cancellation threshold at any time before the timing value reaches the time threshold, exiting the locking preparation state and canceling the current locking operation.

[0019] The technical solution of the above-mentioned embodiments adds a final and active "cancellation" defense line in the locking decision-making process. After entering the "preparation state" of the locking countdown, the system does not stop analyzing, but continues to calculate the "return intention index" of the user. This enables the system to "change its mind" at the last moment, even if the user has triggered the leaving logic, as long as there is a sign of return before locking, the system can suspend the locking operation in time. This design is the ultimate guarantee against false locking, maximizes the accuracy of the locking operation, and reflects a high level of intelligence and fault tolerance design.

[0020] In a second aspect, the embodiments of the present application provide an electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code comprises computer instructions, the one or more processors invoke the computer instructions to make the electronic device execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In a third aspect, the present application provides a computer-readable storage medium, which stores instructions, when the instructions run on an electronic device, make the electronic device execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0022] In a fourth aspect, the present application provides a computer program product comprising instructions which, when the computer program product is executed on an electronic device, cause the electronic device to carry out the method according to the first aspect, and any possible implementation of the first aspect.

[0023] It can be understood that the electronic device provided by the second aspect, the storage medium provided by the third aspect and the computer program product provided by the fourth aspect are all used to execute the method provided by the present application. Therefore, the beneficial effects achieved thereby can refer to the beneficial effects in the corresponding method, which will not be described here again. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is a flow chart of a vehicle lock control method based on trajectory prediction and intention recognition according to an embodiment of the present application;

[0025] Figure 2 is a technical path diagram of a vehicle lock control method based on trajectory prediction and intention recognition according to an embodiment of the present application;

[0026] Figure 3 is a hardware architecture schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] An embodiment of the present application provides a vehicle lock control method based on trajectory prediction and intention recognition, as shown in Figure 1 The method comprises the following steps:

[0028] In step 101, the Bluetooth module of the target vehicle is used to continuously collect Bluetooth signal characteristic parameters from a mobile terminal at a preset frequency to form a Bluetooth signal time sequence data, and to receive real-time collected inertial measurement data from the mobile terminal.

[0029] The mobile terminal includes but is not limited to a mobile phone, a smart watch (a bracelet), a smart key card and other portable user devices with Bluetooth communication and motion sensing functions.

[0030] Specifically, the Bluetooth module of the target vehicle will actively or passively establish a Bluetooth Low Energy connection with the user's mobile terminal. After the connection is established, the Bluetooth module of the target vehicle will continuously scan and receive the Bluetooth signals broadcasted from the mobile terminal at a fixed, pre-set sampling frequency, for example, 5 times per second (5 Hz). In each receiving process, the Bluetooth module of the target vehicle will extract key Bluetooth signal feature parameters, of which the most core parameter is the Received Signal Strength Indication (RSSI). In addition to RSSI, the Bluetooth signal feature parameters can also include other information such as Angle of Arrival (AoA) or Angle of Departure (AoD) for more accurate positioning.

[0031] The vehicle lock control system (hereinafter referred to as the system) receives the Bluetooth signal feature parameters collected by the Bluetooth module at consecutive time points, and then arranges these discrete measurement values in chronological order, thereby constructing a Bluetooth signal time series data. The Bluetooth signal time series data reflects the continuous change process of signal strength over a period of time.

[0032] At the same time, the inertial measurement unit (IMU) built-in the mobile terminal will collect its own motion data in real time, which is the inertial measurement data (IMU data). The inertial measurement data is usually composed of data from a three-axis accelerometer and a three-axis gyroscope, the former for measuring the linear acceleration of the mobile terminal, and the latter for measuring the angular velocity. The mobile terminal will periodically send these real-time collected inertial measurement data to the Bluetooth module of the target vehicle through the established Bluetooth communication link.

[0033] At this point, the system has obtained two independent data sources through the Bluetooth module: one is the Bluetooth signal time series data representing the relative position relationship, and the other is the inertial measurement data representing the absolute motion state of the mobile terminal. These two sets of data can be used to predict the user's motion trajectory.

[0034] Step 102, based on the Bluetooth signal time series data, calculate the dynamic characteristic value for representing the movement process of the mobile terminal relative to the target vehicle, the dynamic characteristic value includes the slope and variance of the Bluetooth signal time series; and based on the inertial measurement data, generate a first judgment result for representing the absolute motion state of the mobile terminal.

[0035] After obtaining the time series data of Bluetooth signal, the system does not directly use the instantaneous value, but carries out deep processing to extract dynamic characteristic values. Specifically, the system uses a sliding time window technology to analyze the Bluetooth signal time series data in the recent period (for example, the past 2 seconds). Within this window, the system calculates the slope of the Bluetooth signal time series by applying linear regression algorithm and the like. The slope, as a key dynamic characteristic value, directly quantifies the trend of signal strength: a significantly positive slope indicates that the signal strength is continuously enhanced, inferring that the mobile terminal is approaching the target vehicle; a significantly negative slope indicates that the signal strength is continuously weakened, inferring that the mobile terminal is moving away from the target vehicle; and a slope close to zero indicates that the relative distance is basically stable.

[0036] At the same time, the system also calculates the variance of the Bluetooth signal time series within the time window. The variance is another important dynamic characteristic value, which is used to measure the fluctuation degree or stability of the signal. A smaller variance indicates that the signal changes smoothly, which usually corresponds to the user's continuous movement with clear intention (such as straight walking); while a larger variance may indicate that the signal is affected by multipath effect, human body shielding, or the user is performing irregular and non-directional activities around the vehicle (such as wandering, turning around).

[0037] In addition, the system processes the received inertial measurement data to determine the absolute motion state of the mobile terminal by analyzing the patterns of acceleration and angular velocity data, such as gait detection algorithm. The absolute motion state refers to the physical movement of the mobile terminal itself, which is independent of the target vehicle.

[0038] After algorithm analysis, the system generates a first judgment result, which classifies the absolute motion state into several preset states, such as "walking", "stationary" or "in the vehicle".

[0039] Step 103, input the dynamic characteristic values and the first judgment result into the preset finite state machine, and generate a second judgment result about the user state based on the dynamic characteristic values, the user state including the moving away state, the approaching state and the in-vehicle state.

[0040] In the system, a finite state machine (FSM) is constructed in advance, which is a core logic unit for modeling and decision-making of the relative relationship between the user and the vehicle. The finite state machine defines several mutually exclusive user states, mainly including a far-away state, an approaching state and an in-vehicle state. The far-away state represents that the relative distance between the mobile terminal and the target vehicle continues to increase, and the system determines that the user has a clear intention to leave. The approaching state represents that the relative distance between the mobile terminal and the target vehicle continues to decrease, and the system determines that the user has an intention to approach and use the vehicle. The in-vehicle state represents that the mobile terminal is located inside the target vehicle, the Bluetooth signal strength reaches the highest and is stable, and the system determines that the user has entered the vehicle and has the permission to start the vehicle.

[0041] In this step, the system provides the obtained dynamic characteristic values (i.e., the slope and variance of the Bluetooth signal time sequence) and the first judgment result about the absolute motion state as inputs to the finite state machine.

[0042] The finite state machine is internally provided with a set of detailed state transition rules, which mainly drive the switching of the state according to the input dynamic characteristic values. For example, when the current user state of the finite state machine is the approaching state, if the received dynamic characteristic values show that the slope is a continuous negative value and the variance is small, it strongly indicates that the user is stably leaving, and the state machine will trigger the state transition and update the current user state to the far-away state. Conversely, if the current user state is the far-away state and the input slope turns to a positive value, the state machine will switch to the approaching state.

[0043] It is worth noting that in this step, the finite state machine mainly generates the second judgment result about the user state based on the dynamic characteristic values representing the relative position change. For example, whether the user state is the far-away state or the approaching state is directly determined by the dynamic characteristic values such as the slope of the Bluetooth signal time sequence. This generated judgment is the second judgment result. The first judgment result (such as "walking" or "still") is input into the state machine as auxiliary information in this stage, which can be used to adjust the confidence of state transition or as a prerequisite condition for certain specific transitions, but the core basis for generating the second judgment result is the dynamic characteristic value.

[0044] In some embodiments, the construction process of the finite state machine includes:

[0045] Firstly, the basic elements of the state machine are defined: states, events and transitions. The states are the above-mentioned far state, close state and in-car state, in addition to which an initial "unknown state" or "connection establishment state" can also be included. The events are the trigger conditions that drive the switching between states, which in this embodiment are mainly defined by the specific values or combinations of the input dynamic characteristic values (slope and variance of the Bluetooth signal time series) and the first judgment result (absolute motion state of the mobile terminal). The transitions are the paths connecting different states, and each transition path is associated with one or a group of specific event rules.

[0046] In the specific construction, the developer needs to design and calibrate the detailed logic rules of state transition according to a large amount of actual scene data and user behavior patterns. For example, the rule for transitioning from the far state to the close state can be defined as: when the slope of the Bluetooth signal time series changes from negative to positive continuously for multiple sampling periods, and the value exceeds a preset slope threshold, it is triggered. The rule for transitioning from the close state to the in-car state can be defined as: when the mean value of the RSSI of the Bluetooth signal exceeds an extremely high in-car threshold (for example, -40 dBm), and the variance is lower than an extremely small stability threshold, it is triggered.

[0047] By defining the states, events and transition rules in such a rigorous manner, a finite state machine that can accurately reflect the user-vehicle interaction process is constructed.

[0048] Step 104, comparing the first judgment result with the second judgment result to obtain a comparison result.

[0049] This step is the key link of information fusion and cross-validation. The system obtains two independent judgment results in the same analysis period in parallel: the first judgment result, which is derived from the analysis of the mobile terminal inertial measurement data, describes the absolute motion state of the mobile terminal itself, such as "walking" or "still"; and the second judgment result, which is derived from the analysis of the dynamic characteristic values of the Bluetooth signal time series based on the finite state machine, describes the user state of the user relative to the target vehicle, such as "far state" or "close state". The comparison process is to logically match and verify these two judgment results with different sources but synchronized in time. The comparison result is the conclusion of whether there is a conflict after logically consistent comparison of the first judgment result and the second judgment result.

[0050] Specifically, the system checks whether the first result and the second result are logically self-consistent according to a preset logical correlation matrix or rule set. For example, if the first result is "walking" and the second result is "away state", the two results are highly consistent in the physical world, because the user leaving by walking will cause the Bluetooth signal strength to continuously decrease. In this case, the comparison result can be "consistent" or "no conflict".

[0051] On the contrary, if the first result is "static" and the second result is "away state", this constitutes a contradiction in physical logic, because a static device should not produce a continuously away signal feature. In this case, the comparison result can be "conflict".

[0052] The essence of the comparison process is to check the validity of the relative position judgment based on wireless signals by introducing independent physical motion sensor data, thereby providing more reliable basis for subsequent decision-making.

[0053] Step 105, if the comparison result meets the preset conflict condition, the first result or the second result is selected as the effective output according to the preset arbitration logic.

[0054] When the comparison result of the previous step is determined to be "conflict", it indicates that the two independent data sources give mutually contradictory conclusions, and the system activates the preset conflict condition and starts the arbitration logic to make a decision. The preset conflict condition is a specific definition of the scene combination that needs to be arbitrated, for example, (first result = "static", second result = "away state"). The arbitration logic is a set of decision rules based on prior knowledge and sensor physical characteristics, used to distinguish which result is closer to the real physical world when a conflict occurs, and select it as the final effective output.

[0055] In specific implementation, the arbitration logic analyzes the specific type of conflict. For example, in the scenario where the user just gets off the vehicle and stands still near the vehicle, the Bluetooth signal strength may instantaneously decrease due to signal multipath effect, causing the second result to be misjudged as "away state", while the inertial measurement data of the mobile terminal accurately outputs the first result as "static". Under this preset conflict condition, the arbitration logic gives higher weight to the inertial measurement data, because it directly measures physical motion and is less susceptible to electromagnetic environment interference. Therefore, the arbitration logic selects the "static" state indicated by the first result to correct or reject the second result, and the final effective output determines that the user has not left, thereby avoiding false locking operation. On the contrary, in some specific scenarios, the arbitration logic may prefer to adopt the second result, achieving dynamic evaluation and optimal decision of sensor reliability in different scenarios.

[0056] Step 106, if the comparison result does not meet the preset conflict condition, the second judgment result is used as the effective output.

[0057] When the system performs comparison and finds that the first judgment result and the second judgment result do not meet any preset conflict condition, it means that the information of the two data sources is mutually confirmed and logically consistent. For example, the first judgment result shows that the mobile terminal is in "walking", and the second judgment result shows that the user state is "approaching". Such consistency greatly enhances the confidence of system decision, indicating that the user is indeed walking towards the target vehicle. In this case, the system will directly adopt the second judgment result as the current effective output.

[0058] The reason for choosing the second judgment result (i.e. "far away", "approaching" or "in the car") as the effective output is that these user states are specifically designed for car lock control logic and are directly related to the user's core intention (leaving, approaching or entering the vehicle), which is the most direct basis for triggering the unlock or lock instruction.

[0059] The first judgment result (such as "walking" or "still") provides important context verification information, but it does not directly map to the car lock control instruction. Therefore, in the case where there is no conflict between the two judgment results and they mutually confirm each other, the system adopts the second judgment result which directly describes the relative relationship between the user and the vehicle for subsequent state update and instruction triggering, which is the most efficient and logically clearest processing method.

[0060] This step ensures that in the high-confidence scenario of consistent information, the system can quickly and accurately respond, maintaining the smoothness and stability of the control flow.

[0061] Step 107, according to the effective output, updating the current user state of the mobile terminal.

[0062] This step is the execution link of the system decision-making closed loop, responsible for formalizing the final judgment result determined after comparison and arbitration as the internal state of the finite state machine. Among them, the effective output, i.e. the first judgment result or the second judgment result selected by the arbitration logic in the previous step, or the second judgment result directly adopted in the case of no conflict, represents the best estimate of the system for the user's real state in this sampling period. When the finite state machine receives this effective output, it will perform state update operation.

[0063] Specifically, the finite state machine compares the incoming valid output (e.g. "far state") with the current user state (e.g. "close state") currently stored inside the finite state machine. If the two are inconsistent, the finite state machine will update the "current user state" from "close state" to "far state" strictly according to the preset transition rules. If the two are consistent, the "current user state" remains unchanged, which can be regarded as a confirmation of the current state, enhancing the stability of the state. This updating process is crucial because the "current user state" is the "memory" of the entire system for the user's position and intention, and is the only basis for all subsequent control instructions (such as unlocking or locking).

[0064] Through this step, the system converts discrete, possibly noisy instantaneous judgments into a continuous, stable and error-corrected internal logic state, laying a solid foundation for subsequent accurate instruction triggering.

[0065] Step 108, when the updated user state of the finite state machine is from the far state to the close state, triggering the unlocking instruction to make the target vehicle in the unlocked state.

[0066] This step specifies the precise conditions for triggering the vehicle unlocking operation. The system does not simply execute the unlocking when the user is in the "close state", but strictly monitors the dynamic changes of the "current user state" inside the finite state machine.

[0067] Specifically, the system continuously monitors whether the current user state has undergone a specific state transition, i.e. from "far state" to "close state". This state transition-based triggering logic can accurately capture the complete behavior intention of the user approaching the vehicle from a distance, effectively avoiding unnecessary or repeated unlocking operations due to the user's wandering around the vehicle.

[0068] Once the system detects that the current user state has exactly completed the update from "far state" to "close state", it determines that the user's approach behavior is clear and effective. At this time, the system will immediately generate and send a digital unlocking instruction. The unlocking instruction is transmitted to the body control module through the vehicle's internal bus network (e.g. CAN bus).

[0069] After receiving and verifying the legality of the unlocking instruction, the body control module will drive the door lock actuator of the target vehicle to act, so that all doors of the target vehicle are switched from the locked state to the unlocked state, completing the entire automatic unlocking process.

[0070] Step 109, when the updated user state of the finite state machine is from the approaching state or the in-vehicle state to the away state, and the away state lasts stably for more than a preset time threshold, triggering the locking instruction to make the target vehicle in the locked state.

[0071] This embodiment defines a more intelligent and humanized locking logic containing double conditions. First, the system monitors whether the state transition of the finite state machine from the approaching state or the in-vehicle state to the away state occurs. This transition event marks that the user may have started to leave the target vehicle, which is the starting point of triggering the locking judgment.

[0072] However, the system does not immediately execute the locking, but starts an internal timer and enters an observation period. During this observation period, the system continuously verifies whether the current user state stably remains in the away state. The so-called continuous stability means that the current user state does not jump to the approaching state or other states again during the timing period. Only when the accumulated time of the timer exceeds a preset time threshold and the user state is always in the away state during this period, the system finally confirms that the user's leaving intention is continuous and clear.

[0073] Among them, the time threshold is a configurable parameter, for example, 3 seconds, which is long enough to accommodate the user's short stay at the vehicle side, opening and closing the trunk, and other complex behaviors. Once the two conditions (state transition and time delay) are met at the same time, the system generates and sends a locking instruction to the body control module to make the target vehicle enter a safe locked state. This design greatly improves the fault tolerance and user experience of the system.

[0074] Through the above series of method steps, this embodiment can overcome the control logic defects caused by relying on a single moment RSSI threshold in the related art, and through the introduction of multi-dimensional information fusion and dynamic process analysis, it realizes accurate prediction of user motion trajectory and accurate identification of real unlocking and locking intention, thereby significantly improving the accuracy and intelligent level of the keyless entry and start system.

[0075] Firstly, unlike the related art which only relies on instantaneous signal strength, this embodiment collects time series data of Bluetooth signals and calculates dynamic characteristic values such as slope and variance, upgrading from "point" judgment to "process" analysis. The slope can reveal the trend of the relative distance between the user and the vehicle (whether it is continuously approaching or continuously moving away), and the variance can reflect the stability of the signal, effectively distinguishing whether the user is stably moving or lingering around the vehicle. This dynamic process-based analysis accurately predicts the user's motion trajectory and effectively solves the false locking problem caused by the instantaneous signal fluctuation due to the user's complex behaviors such as short stay around the vehicle and detour.

[0076] Secondly, the embodiment creatively introduces the inertial measurement (IMU) data of the mobile terminal as a second information source. The IMU data reflects the absolute motion state of the mobile terminal itself (such as stationary, walking), which forms an effective cross-verification with the relative position relationship represented by the Bluetooth signal. By comparing the Bluetooth dynamic feature judgment result with the IMU motion state judgment result, and setting up conflict arbitration logic, the application constructs a more robust decision system. For example, when the Bluetooth signal shows that the user is "far away" but the IMU data shows that the user is "stationary", the system can intelligently judge that this may be a signal interference or a scenario where the user is standing by the car, rather than truly leaving, thereby avoiding false locking instructions.

[0077] Finally, the embodiment increases the time threshold constraint in the logic of triggering the locking instruction, that is, it requires the "far away state" to be stable for more than a preset time. This design provides sufficient buffer for the non-immediate leaving behavior of the user (such as walking to the cab after taking something from the trunk), constitutes the last line of defense against misoperation, and ensures that the locking operation fully meets the user's final intention.

[0078] The embodiment constructs a two-dimensional perception model that combines the dynamic characteristics of the Bluetooth signal and the inertial state of the terminal, and combines the multi-state machine and the delay confirmation mechanism, upgrades the simple signal threshold judgment to a deep understanding and prediction of the user's continuous behavior trajectory and real intention, thereby effectively solving the technical pain points of frequent misjudgment and misoperation of related technologies in complex scenarios, greatly enhancing the reliability, safety and user experience of the system.

[0079] In some embodiments, in step 101, based on the inertial measurement data, a first judgment result is generated for representing the absolute motion state of the mobile terminal itself, which can specifically include:

[0080] S1011, extracting acceleration and angular velocity data from the inertial measurement data.

[0081] The system first receives the inertial measurement data from the mobile terminal through the Bluetooth module of the target vehicle, which is encapsulated in a data packet. The inertial measurement data is the raw or preprocessed data stream generated by the inertial measurement unit built in the mobile terminal. The inertial measurement unit usually integrates a three-axis accelerometer and a three-axis gyroscope.

[0082] In this step, the processing module of the system parses the received data packet, thereby separating the composite inertial measurement data into two independent data sets. The first data set is acceleration data, measured by the three-axis accelerometer, which contains the specific force information of the mobile terminal on the three orthogonal axes (x, y, z) in its own carrier coordinate system. The specific force is the vector sum of the true linear acceleration of the mobile terminal and the gravity acceleration component. The second data set is angular velocity data, measured by the three-axis gyroscope, which represents the speed of rotation of the mobile terminal around the three orthogonal axes (x, y, z) in its own carrier coordinate system.

[0083] After the extraction is completed, the system obtains two basic input quantities for subsequent kinematic calculations: a set of time-sequenced three-dimensional acceleration data containing gravity components, and a set of synchronized three-dimensional angular velocity data representing the rate of attitude change.

[0084] S1012, based on the acceleration and angular velocity data, the real-time motion trajectory of the mobile terminal in three-dimensional space is calculated by an attitude solving algorithm.

[0085] The core of this step is to convert the physical quantities measured from the sensor coordinate system to the position change in the global navigation coordinate system. First, the system uses an attitude solving algorithm, such as complementary filtering, Kalman filtering, or Mahony / Madgwick gradient descent algorithm, to fuse the extracted acceleration data and angular velocity data. The attitude solving algorithm can use angular velocity data to estimate the angle change by time integration, and use acceleration data to perceive the gravity direction under static or quasi-static conditions to correct the integral drift of the gyroscope, thereby accurately calculating the real-time attitude of the mobile terminal relative to the earth-fixed coordinate system (navigation coordinate system), usually in the form of a quaternion or a rotation matrix.

[0086] After obtaining the real-time attitude, the system converts the original acceleration data from the carrier coordinate system of the mobile terminal to the navigation coordinate system using the attitude information. In this process, the system can subtract the gravity acceleration vector (a constant quantity in the navigation coordinate system, about 9.8 m / s²) from the converted acceleration data, thereby obtaining the pure linear acceleration of the mobile terminal.

[0087] Finally, the system performs two consecutive time integration operations on the linear acceleration: the first integration obtains the velocity of the mobile terminal in the navigation coordinate system, and the second integration obtains the displacement relative to the initial position. This series of consecutive displacement points constitutes the real-time motion trajectory of the mobile terminal in three-dimensional space.

[0088] S1013, generating a first judgment result indicating that the mobile terminal is approaching the target vehicle, is moving away from the target vehicle, or is lingering around the target vehicle according to the projection component of the real-time motion trajectory on the horizontal plane.

[0089] After obtaining the real-time motion trajectory in three-dimensional space, the system first performs dimension reduction processing on the real-time motion trajectory in order to simplify analysis and focus on the core behavior related to vehicle access. Specifically, the system projects the three-dimensional trajectory point sequence onto a fixed horizontal plane, ignoring the height change in the vertical direction, thereby obtaining a motion trajectory on a two-dimensional plane.

[0090] Subsequently, the system needs to analyze the relative relationship between the two-dimensional trajectory and the target vehicle. The system sets the position of the target vehicle as the origin or a fixed reference point of the horizontal plane coordinate system. Then, the system analyzes the projection position sequence of the mobile terminal in the recent period of time in a sliding time window manner. By calculating the distance change trend between the position point of the mobile terminal and the reference point of the target vehicle in the time window, the system can make a judgment: if the distance continuously and significantly decreases, the first judgment result of "approaching the target vehicle" is generated; if the distance continuously and significantly increases, the first judgment result of "moving away from the target vehicle" is generated.

[0091] For the judgment of "lingering around the target vehicle", the system analyzes more complex trajectory features. For example, when it is found that the motion speed of the mobile terminal is not zero, but the distance between the mobile terminal and the target vehicle fluctuates within a small range, and the curvature of the trajectory is large or the direction changes frequently, the system generates the first judgment result of "lingering around the target vehicle".

[0092] In some embodiments, the arbitration logic described above can include:

[0093] (1) When the preset conflict condition is that the first judgment result indicates that the mobile terminal is in a lingering state and the second judgment result indicates that the user state has been converted to a moving away state, the first judgment result is selected as the valid output.

[0094] This arbitration logic is specifically used to handle the complex scenario of non-directional movement of the user around the vehicle. In this scenario, the two judgment results generated by the system in parallel are contradictory: first, by analyzing the real-time motion trajectory calculated from the inertial measurement data, the system finds that the projection component of the mobile terminal on the horizontal plane continuously moves, but the distance between the projection component and the reference point of the target vehicle always fluctuates within a small range, and the motion direction changes frequently, without a clear trend, so the first judgment result generated is "lingering state".

[0095] However, during the same period, the user's body might block the Bluetooth signal due to walking around the vehicle or turning around, or the mobile terminal might enter an area with weak signal, causing a short and sharp drop in the strength value (RSSI) of the Bluetooth signal time series. This downward trend might cause the slope of the Bluetooth signal time series to be calculated as a negative value, leading the finite state machine to misjudge that the user is leaving, and thus generate a second judgment result that the user state has been converted to the "far away state".

[0096] After the system identifies the conflict between the "wandering state" (from the first judgment result) and the "far away state" (from the second judgment result), arbitration is started. The arbitration logic determines, based on prior knowledge, that the physical trajectory reflected by the inertial measurement data is more indicative of the user's true intention in this situation than the Bluetooth signal, which is susceptible to environmental interference.

[0097] Therefore, the arbitration logic decides to select the first judgment result as the valid output, and identifies the user's true state as "wandering", thereby suppressing the false locking process that might be triggered by the second judgment result.

[0098] (2) When the preset conflict condition indicates that the mobile terminal is in a stable approaching or stable leaving state according to the first judgment result, and the second judgment result indicates that the user state is the out-of-vehicle stationary state because the variance of the Bluetooth signal time series is continuously higher than the variance threshold, the first judgment result is selected as the valid output.

[0099] This arbitration logic mainly solves the problem of misjudgment caused by a serious decline in Bluetooth signal quality in a complex electromagnetic environment. In this scenario, the user is actually walking straight to or away from the target vehicle at a relatively stable speed. Through gait detection and trajectory integration, the first judgment result generated based on the inertial measurement data can accurately determine that the mobile terminal is in a "stable approaching" or "stable leaving" state.

[0100] However, if the vehicle is parked in an underground parking lot or near a large metal structure, the Bluetooth signal will suffer from severe multipath effects, that is, the signal reaches the receiving end after being reflected through multiple paths, causing interference and resulting in sharp and irregular rapid fluctuations in the received signal strength indication (RSSI). Such fluctuations will cause the variance value of the Bluetooth signal time series calculated by the system to be continuously higher than the preset variance threshold. In the rules of the finite state machine, a continuous high variance is usually interpreted as an unreliable signal or irregular micro-movement by the user, and thus might be classified as a special "out-of-vehicle stationary state" or "state unknown", thus generating a second judgment result that does not match the actual situation.

[0101] When the system detects the preset conflict condition of "steady approaching / away" (from the first judgment result) and "out-of-vehicle stationary state" (from the second judgment result, and the trigger cause is variance overrun), the arbitration logic is started. Since the inertial measurement data is basically not affected by the external electromagnetic environment, its judgment of stable linear motion has very high reliability. Therefore, the arbitration logic will prefer to accept the first judgment result, and select it as the effective output, so as to correct the misjudgment caused by the poor signal environment, and ensure that the system can continue to correctly track the real motion trajectory of the user.

[0102] In some embodiments, the arbitration logic described above can also be a dynamic weight arbitration logic, and the execution of the dynamic weight arbitration logic includes the following steps:

[0103] S1051, real-time calculation of data reliability weight of inertial measurement data and data reliability weight of Bluetooth signal time series data; wherein the data reliability weight of the inertial measurement data is calculated based on the signal noise level and the zero drift stability, and the data reliability weight of the Bluetooth signal time series data is calculated based on the signal-to-noise ratio and the multipath effect interference degree.

[0104] This step aims to quantitatively evaluate the instantaneous quality of the two independent data sources, and provides the basis for subsequent weighted fusion.

[0105] Firstly, for the data reliability weight of the inertial measurement data, the system will perform online sensor performance analysis. The system calculates the variance of the accelerometer and gyroscope output data in the short window when the mobile terminal is judged to be stationary, so as to quantify the current signal noise level; a lower variance corresponds to a lower noise, thereby positively contributing to the data reliability weight.

[0106] At the same time, the system will refer to a long-term maintained zero drift stability model, which records the drift characteristics of the sensor zero point over time and temperature; a smaller current drift and higher stability will also increase the data reliability weight. The system inputs the quantitative indicators of signal noise level and zero drift stability into a preset function or lookup table, and calculates the data reliability weight of the inertial measurement data between 0 and 1.

[0107] Secondly, for the data confidence weight of the Bluetooth signal time series data, the system mainly evaluates the quality of the wireless channel. The system analyzes the RSSI value of the received Bluetooth signal time series data, and indirectly estimates the signal-to-noise ratio and the degree of multipath effect interference by calculating the mean and variance thereof in a sliding time window. A higher RSSI mean and a lower variance generally mean a stronger direct signal and a weaker multipath interference, that is, a good channel quality, and thus a higher data confidence weight of the Bluetooth signal time series data is mapped. Conversely, a sharply fluctuating RSSI sequence (high variance) indicates that there is a serious multipath effect, and the data confidence weight thereof will be significantly lowered.

[0108] S1052, the first judgment result and the second judgment result are respectively weighted and fused with the corresponding data confidence weights to generate a fused final user state judgment.

[0109] This step is the core of the dynamic weight arbitration logic, which integrates two judgment results with different confidence levels into a unified conclusion through mathematical methods. In this process, the system no longer makes a two-choice decision like hard rule arbitration, but performs a more refined fusion calculation.

[0110] First, the system converts the discrete first judgment result (for example, "stable approach") and the second judgment result (for example, "out-of-car stationary state") into a quantifiable representation, such as assigning a probability or confidence score to each possible user state (far-away state, approach state, wandering state, etc.).

[0111] For the first judgment result, the state indicated thereby will obtain an initial confidence, which is then modulated by the data confidence weight of the inertial measurement data. Similarly, the initial confidence of the state indicated by the second judgment result is modulated by the data confidence weight of the Bluetooth signal time series data.

[0112] Next, the system uses an information fusion algorithm, such as Bayesian inference, Dempster-Shafer theory, or a conventional weighted average voting mechanism, to combine the two weighted confidence distributions. For example, for the "approach state" as the final state, the final score thereof can be obtained by multiplying the support of the "approach state" by the first judgment result by the data confidence weight of the inertial measurement data, adding the support of the "approach state" by the second judgment result by the data confidence weight of the Bluetooth signal time series data, and then performing normalization.

[0113] The fusion algorithm considers all possible states of the weighted evidence, and finally outputs a state with the highest probability or the highest comprehensive score, which is the fused final user state judgment.

[0114] S1053, the final user state judgment is designated as the valid output.

[0115] This step is the end point of the dynamic weight arbitration process, responsible for formalizing the conclusion derived from the complex fusion calculation and passing it to the next processing link of the system. In the previous step, the system has generated a single final user state judgment through the weighted fusion algorithm, which represents the most reliable evaluation of the user's intention at the current moment, such as the "approaching state".

[0116] In this step S1053, the system directly designates this final user state judgment as the valid output for updating the finite state machine. This means that, regardless of whether there is a conflict between the original first judgment result and the second judgment result, and regardless of the degree of conflict, the dynamic weight arbitration logic can give a harmonized, quantitative optimal solution. This valid output is then input into the finite state machine to update the current user state of the mobile terminal.

[0117] Compared with the arbitration logic based on fixed rules, the valid output generated by this dynamic weight method has higher adaptability and robustness. For example, in a scenario where the Bluetooth signal quality is extremely poor but the inertial measurement data is very stable, the fusion result will greatly favor the judgment of the inertial measurement data; conversely, the opposite is also true.

[0118] In this way, the system realizes real-time self-adaptation of sensor data quality, ensuring that the output decision is based on the comprehensive judgment of the current most reliable information source in any environment, thereby significantly improving the accuracy and reliability of the entire car lock control method.

[0119] In some embodiments, the user state also includes an outside stationary state. The outside stationary state is a logical state determined based on the variance of the Bluetooth signal time series being continuously higher than a preset threshold, used to represent the scenario where the user's approaching or away intention is unclear and temporarily stays outside the vehicle due to unstable signal or complex user behavior.

[0120] Therefore, in step 103, the method of the present embodiment can further include: when in the approaching state or the away state, detecting that the variance of the Bluetooth signal time series data is continuously higher than a preset variance threshold, judging the current user state as the outside stationary state, and delaying the triggering of the car lock control instruction.

[0121] This step aims to build a buffer and fault-tolerant mechanism for signal uncertainty. First, the system continuously monitors the current user state inside the finite state machine. When the current user state is in the "approaching state" or "moving away state", it indicates that the system initially judges that the user is performing an action with a clear direction. During this period, the system processing module uses a sliding time window to perform real-time statistical analysis on the latest collected Bluetooth signal time series data, and the core is to calculate the variance of the Bluetooth signal time series. The variance of the Bluetooth signal time series is a key dynamic characteristic value for quantifying the fluctuation degree of the signal strength; a smooth movement corresponds to a smaller variance, while signal interference or user wandering and other irregular activities will lead to a larger variance.

[0122] The system continuously compares the real-time calculated variance value with a pre-calibrated preset variance threshold. Once the system detects that the variance of the Bluetooth signal time series is continuously higher than the preset variance threshold in multiple calculation periods, the system triggers state transition and forcibly updates the current user state to the "out-of-vehicle stationary state".

[0123] After entering the "out-of-vehicle stationary state", the core control logic is to delay the triggering of the car lock control instruction, which means that the system's active unlocking or locking decision-making process will be temporarily suspended. Even if the average strength of the Bluetooth signal meets the threshold condition for unlocking or locking at this time, the system will not issue the corresponding car lock control instruction. The system will remain in the "out-of-vehicle stationary state" until the variance of the Bluetooth signal time series falls below the preset variance threshold, at which time the system will exit the "out-of-vehicle stationary state" and return to the normal judgment process.

[0124] This mechanism effectively avoids unexpected car lock operations caused by transient signal fluctuations or user activities with no clear intention near the car, significantly enhancing the intelligence and robustness of the system.

[0125] In some embodiments, the preset time threshold in step 109 is a variable value calculated dynamically, wherein the step of dynamic calculation includes:

[0126] S1091, when the user state is detected to be converted from the approaching state or the in-vehicle state to the moving away state, the slope stability and absolute value of the Bluetooth signal time series data before the state conversion are extracted.

[0127] This step aims to perform in-depth feature mining on the behavior process that triggers the determination of "away state". When the current user state of the finite state machine undergoes a specific transition from "approaching state" or "in-vehicle state" to "away state", the system does not immediately start timing, but first traces back and analyzes the Bluetooth signal time series data immediately before the transition point. The system defines a fixed length of the review window, for example, 2 seconds before the state transition. For the Bluetooth signal time series data in the time window, the system first calculates an overall slope, and the absolute value of the slope is the slope absolute value. The slope absolute value directly quantifies the average rate of signal strength decline in this period, and a larger slope absolute value usually means that the user is quickly leaving the target vehicle.

[0128] At the same time, in order to evaluate the stability of the leaving process, the system further divides the review time window into multiple smaller, possibly overlapping sub-windows. The system calculates a local slope value for each sub-window, thereby obtaining a slope value sequence. Then, the system calculates the variance or standard deviation of the slope value sequence, which is used to represent the slope stability. A lower variance indicates that the speed and direction are very consistent throughout the leaving process, i.e. high slope stability; on the contrary, a higher variance means that the user's leaving process is accompanied by speed changes or direction hesitation, i.e. low slope stability.

[0129] S1092, based on the slope stability and slope absolute value described above, quantitatively evaluate the decisiveness of a leaving behavior.

[0130] This step aims to fuse the two independent physical features extracted in the previous stage into a single indicator that can intuitively reflect the strength of the user's leaving intention, i.e. decisiveness. The decisiveness is a comprehensive quantitative score that describes the certainty and irreversibility of the user's leaving behavior.

[0131] The system will use a pre-set multi-dimensional evaluation model, such as a weighted sum function or a fuzzy logic reasoning system, to calculate the decisiveness. In this model, the slope absolute value and the slope stability are two main input variables. A specific calculation logic can be:

[0132] A high absolute value of the slope and a high level of slope stability (i.e. low variance) both contribute positively and significantly to the final score of decisiveness. This means that when the system observes that the user not only leaves quickly (high absolute value of the slope), but also leaves in a very smooth and unhesitant trajectory (high slope stability), the evaluation model outputs a very high decisiveness score. Conversely, if the user leaves very slowly (low absolute value of the slope) and the signal fluctuates dramatically during the leaving process, showing signs of hesitation or turning back (low slope stability), the model outputs a very low decisiveness score.

[0133] Through this quantitative evaluation, the system successfully maps the complex signal dynamics into a numerical value that characterizes the user's psychological intention, which is easily used by subsequent decision-making systems.

[0134] In some embodiments, the calculation of decisiveness can adopt the following formula:

[0135]

[0136] where R represents decisiveness, with a theoretical value range of -1 to 1, but the actual effective interval is closer to 0 to 1. The higher the R value, the more decisive the leaving behavior.

[0137] S norm represents the normalized absolute value of the slope, reflecting the instantaneous speed of the user's leaving.

[0138] A norm represents the normalized acceleration, reflecting the trend of the user's leaving "speed", which is obtained by calculating the rate of change of the slope of the Bluetooth signal time series (i.e. the second derivative) and normalizing it; a positive value indicates that the user is accelerating away, a negative value indicates deceleration, and a value close to zero indicates a constant speed.

[0139] V norm represents the normalized slope variance, reflecting the stability of the user's leaving process.

[0140] C path represents the path curvature proxy, which is an index for indirectly estimating the degree of curvature of the motion trajectory. It can be approximately calculated by analyzing the frequency of sign changes or zero-crossing rate of the slope sequence within the sliding window. A straight leaving action should keep the slope sign unchanged (consistently negative), so the value of C path is close to 0. A roundabout or hesitation action will cause the slope sign to change frequently, and the value of C path is higher.

[0141] tanh(·) represents the hyperbolic tangent function, which is a sigmoid activation function with an output range between -1 and 1. Similar to the logistic function, it provides a non-linear saturation effect, i.e., the output value changes tend to flatten when the input value is very large or very small. It is used here to evaluate the contribution of "speed" and "acceleration" to decisiveness, respectively.

[0142] α represents the dynamic weight factor, which is a core adaptive parameter with a value range between 0 and 1. α is used to dynamically adjust the relative importance of S norm and A norm in the calculation of decisiveness. Its value is determined by the stability of the signal V norm .

[0143] k s ,k a both represent sensitivity coefficients, which are used to adjust the sensitivity of decisiveness to changes in speed and acceleration, respectively.

[0144] λ,C p both are penalty factors, λ is the basic stability penalty factor, and C p is specifically used to penalize the curvature of the path.

[0145] S thresh is the judgment threshold of speed.

[0146] V thresh is the judgment threshold of variance, which determines the critical point at which the dynamic weight α changes significantly.

[0147] β represents the α transition sharpness coefficient, which is used to control the speed of the smooth transition of the dynamic weight α from 1 to 0. The larger β is, the steeper the transition is.

[0148] The above formula builds a highly nonlinear and self-adjusting decision-making model by introducing a dynamic weight mechanism and multi-dimensional kinematic characteristics. Its beneficial effects are reflected in the following aspects:

[0149] First, it realizes the leap from "static combination" to "dynamic arbitration", which embodies significant non-obviousness. The core innovation of the formula lies in the design of the dynamic weight α. The calculation logic of α is: when the signal is stable (V norm <V thresh ), α tends to 1, and at this time the judgment of decisiveness mainly depends on the more reliable speed information (S norm ); when the signal is unstable (V norm >V thresh ), α tends to 0, and the system will automatically reduce the trust in the possibly distorted speed information and instead rely more on the acceleration information (A normThis kind of adaptive arbitration mechanism simulates the complex behavior of high-level agents adjusting their decision-making basis in the face of uncertain information, far beyond the simple weighted sum.

[0150] Secondly, predictive (acceleration) and geometric (path curvature) features are introduced to build a more complete behavioral portrait. Instead of focusing only on "how fast are you now" (velocity), the formula now looks at "will you be going faster or slower next" (acceleration) and "are you taking a straight line or a curved path" (path curvature). Acceleration is a powerful indicator of intent, a user who is accelerating away is clearly more committed than one who is maintaining a constant speed or decelerating. Path curvature effectively filters out non-evasive behaviors such as "circling for items". This multi-dimensional, high-order feature fusion allows the system to understand the "narrative" of user behavior, rather than just the "instantaneous".

[0151] Finally, through the composite design of the exponential penalty term, a more rigorous and comprehensive "confidence veto" mechanism is established. The final penalty term The penalties for stationarity (V norm ) and path straightness (|C path |) are coupled together. This means that no matter how strong the "motion intent" calculated by speed and acceleration is, if the entire process is unstable or the path is curved, the final decisiveness score will be exponentially suppressed. This "veto" mechanism ensures that the system's decision is extremely conservative and safe, and only when all evidence points to a clear, stable, and direct leaving behavior will the system give a high decisiveness judgment, greatly improving the reliability in complex scenarios.

[0152] S1093, set a time threshold according to the decisiveness, wherein the time threshold is inversely proportional to the decisiveness, so that a high decisiveness corresponds to a shorter time threshold, and a low decisiveness corresponds to a longer time threshold.

[0153] This step is the core link to achieve adaptive adjustment of the lock delay time, aiming to make the system's response strategy highly match the user's behavioral intent. After calculating the decisiveness score of the leaving behavior, the system will use a pre-set mapping function to dynamically set the time threshold for triggering the lock instruction. The core feature of this mapping function is that the generated time threshold is strictly inversely proportional to the input decisiveness score.

[0154] In implementation, the system can set a maximum delay time (e.g. 5 seconds) and a minimum delay time (e.g. 1 second). The mapping function can linearly or nonlinearly interpolate the final time threshold according to the decisiveness score (assuming it has been normalized to the interval of 0 to 1). For example, when the decisiveness score is the highest, indicating that the user's departure intention is extremely clear, the system can set the time threshold to the minimum value to achieve a quick and decisive locking response. Conversely, when the decisiveness score is the lowest, indicating that the user's departure behavior is very ambiguous and may only be a short stay at the side of the car, the system can set the time threshold to the maximum value, thereby providing sufficient buffer time for the user to return or change their intention, effectively avoiding false locking.

[0155] Through this dynamic adjustment mechanism, the system realizes the transition from a "one-size-fits-all" fixed delay to an "individualized" intelligent delay.

[0156] In some embodiments, based on step 109, after the user state is converted from the approaching state or the in-car state to the away state, and before the locking instruction is triggered, the method further comprises:

[0157] S201, starting the countdown of the time threshold when the locking control system enters the locking preparation state.

[0158] This step is the formal conversion link from intention judgment to execution delay. When the user state inside the finite state machine exactly transfers from the "approaching state" or "in-car state" to the "away state", the system does not immediately enter a passive waiting period, but actively switches its control logic to a clearly defined locking preparation state.

[0159] Entering this locking preparation state marks the initialization of a potential locking process. Synchronously with this state switching action, the system activates a high-precision internal timer module. The timer module will be loaded with a specific countdown duration, which is the preset time threshold. The time threshold can be a variable value dynamically calculated according to the "decisiveness", or a fixed safety time.

[0160] Once the timer is started, it begins to count down at the system clock frequency, and the current remaining time or elapsed time is the "timing value". This locking preparation state is an active monitoring phase, during which the system not only maintains the operation of the timer, but also continuously performs subsequent data acquisition and intention analysis tasks to provide real-time basis for the final locking decision or cancellation decision.

[0161] S202, during the duration of the locking preparation state, calculating a return intention index for quantifying the user's return possibility based on the latest acquired signal time series data.

[0162] This step aims to detect the real-time and fine intention reversal of the user's post-behavior. During the whole process of the system being in the lock preparation state and counting down, the Bluetooth module of the target vehicle will not stop working, but will continue to collect the time series data of the Bluetooth signal from the mobile terminal at a preset frequency.

[0163] The data processing unit of the system will use a sliding time window to analyze the latest acquired Bluetooth signal time series data in this stage, and calculate a key quantitative indicator, the return intention index, based on this. The return intention index is a specially designed comprehensive score for measuring the user's possibility of changing his mind and returning to the vehicle after leaving. Its calculation can be based on the sign and amplitude of the slope of the Bluetooth signal time series: a slope that changes from negative to positive and continues to increase in amplitude will make a great positive contribution to the return intention index, because it directly corresponds to the signal strength changing from weakening to strengthening, which is the strongest physical evidence of the user's return. The return intention index will be updated periodically, forming a time series reflecting the strength of the user's return intention over time.

[0164] In some embodiments, step S202 can specifically include the following steps:

[0165] S2021, synchronize and align the latest acquired Bluetooth signal time series data with the inertial measurement unit data from the mobile terminal; and perform joint time-frequency domain analysis on the Bluetooth signal time series data to extract its spectral feature vector X; at the same time, extract the motion singularity feature vector Y indicating the physical motion mutation event from the synchronized IMU data.

[0166] This step aims to extract structured features with high information density and suitable for the input of advanced artificial intelligence models from raw, multi-modal sensor data.

[0167] Firstly, the system performs strict data synchronization and alignment operation to ensure that each frame of the latest acquired Bluetooth signal time series data can be accurately matched with the inertial measurement data from the mobile terminal at the same time on the same high-precision time reference, which is the cornerstone of subsequent multi-modal information fusion.

[0168] Subsequently, for the synchronized Bluetooth signal time series data, the system does not use traditional time domain analysis, but performs joint time-frequency domain analysis. By applying techniques such as short-time Fourier transform (STFT) or continuous wavelet transform (CWT), the system can convert one-dimensional signal strength changes into two-dimensional time-frequency spectrograms, revealing the frequency components hidden in the signal fluctuations and their evolution over time.

[0169] From this time-frequency spectrogram, the system extracts a high-dimensional spectral feature vector X, which can contain the energy distribution of specific frequency bands, spectral entropy, spectral centroid, and a series of features that can finely depict the unique "fingerprint" of different user movement patterns (such as walking at a constant speed, turning around, pausing) on the wireless channel.

[0170] At the same time, the system processes the inertial measurement data synchronized with it, focusing on detecting non-stationary, high-transient events in physical motion, i.e., motion singularities. By calculating the derivative of acceleration and performing peak detection, or analyzing the abrupt changes in gyroscope angular velocity, the system can accurately capture key actions such as sudden stopping, turning around, or changing direction of the user, and encode the occurrence time, intensity, duration, etc. of these events into a motion singularity feature vector Y.

[0171] S2022, the spectral feature vector X and the motion singularity feature vector Y are input into a pre-trained dual-flow attention deep learning model in parallel; the model internally calculates the dynamic influence weight of the motion singularity event on the Bluetooth spectral features through cross-attention mechanism, and outputs a preliminary return intention probability value.

[0172] This step is the core link of intelligent fusion and intention inference of two heterogeneous features using a deep learning model. The system sends the spectral feature vector X and the motion singularity feature vector Y generated in the previous stage into two independent input streams of a pre-trained dual-flow attention deep learning model. The dual-flow architecture of the model allows it to parallelly encode the features of the two modalities, for example, using a one-dimensional convolutional neural network (1D-CNN) to capture the local time-frequency patterns in the spectral feature vector X, while using a recurrent neural network (RNN) or self-attention layer to understand the event sequence represented by the motion singularity feature vector Y.

[0173] These two information streams are not simply spliced at a later stage, but are deeply interacted through a core cross-attention mechanism. In this mechanism, the motion singularity feature vector Y from the inertial measurement unit (IMU) data is used as "query" to "probe" the "key" and "value" space composed of the spectral feature vector X from the Bluetooth signal. Its physical meaning is that the model actively "asks": when a clear physical turning event (from the motion singularity feature vector Y) occurs, which part of the spectral features of the Bluetooth signal (from the spectral feature vector X) is most strongly associated with this event? Through the calculated dynamic influence weight, the model can greatly enhance the expression of spectral features that are highly related to physical actions, while suppressing irrelevant noise fluctuations.

[0174] After the deep fusion through the cross-attention layer, the integrated feature information is sent to the output layer of the model, and a preliminary return intention probability value is finally generated through an activation function (such as Sigmoid).

[0175] S2023, real-time evaluation of the cognitive uncertainty of the preliminary return intention probability value output by the deep learning model; and according to the size of the cognitive uncertainty, a dynamic adjustment of an environmental noise adaptation factor is performed to calibrate the preliminary return intention probability value, and finally a return intention index is output, which not only reflects the return possibility but also contains the model's confidence information.

[0176] This step aims to overcome the "overconfidence" problem that the deep learning model may produce when facing unknown or ambiguous inputs, and to provide a guarantee for the reliability of the final output. Instead of directly adopting the preliminary return intention probability value output by the model, the system first quantifies the uncertainty of the model output.

[0177] Specifically, the system will real-time evaluate the cognitive uncertainty of the model output, which reflects the model's "knowledge" deficiency due to insufficient training data or encountering out-of-distribution samples. An effective implementation method is to use the Monte Carlo Dropout technique, which runs the same input through the model multiple times (such as dozens of times) during the inference phase, and randomly "drops" a portion of neurons during each forward propagation. By calculating the variance of the dozens of output results, the system can obtain a quantitative measure of cognitive uncertainty.

[0178] Subsequently, the system dynamically adjusts an environmental noise adaptation factor according to the size of the cognitive uncertainty. The mechanism of the environmental noise adaptation factor is as follows: when the cognitive uncertainty is low, it indicates that the model is very confident in its judgment, and the factor has little effect on the preliminary return intention probability value; on the contrary, when the cognitive uncertainty is very high, it indicates that the model has encountered a difficult-to-understand scenario, and the factor will significantly pull or shrink the preliminary return intention probability value to a neutral and uncertain value (such as 0.5).

[0179] By performing mathematical operations (such as weighting or scaling) on the preliminary return intention probability value and the dynamically adjusted environmental noise adaptation factor, the system completes the calibration process and finally outputs a return intention index that not only contains the model's prediction of the user's return possibility but also embeds the model's confidence level in the prediction after careful calibration.

[0180] S203, determining whether the timer value reaches the time threshold.

[0181] This step is the time dimension judgment core of the decision-making process in the lock preparation state. After the timer starts from entering the lock preparation state, the central control unit of the system will query the current timing value of the timer in each processing cycle. The timing value can be the remaining time in the countdown mode or the elapsed time in the positive counting mode.

[0182] The system will compare the real-time timing value with the initial set time threshold in a deterministic numerical comparison. If countdown is used, the judgment condition is whether the timing value is less than or equal to zero; if positive counting is used, the judgment condition is whether the timing value is greater than or equal to the time threshold. This judgment operation is continuous and constitutes a key precondition in the lock decision-making process. The judgment itself does not directly trigger any lock command, but serves as a time-gated signal, and its result ("reach" or "not reach") will be passed to the subsequent combination logic judgment unit to determine whether all conditions for executing the lock operation are met at the precise moment when the time window is closed.

[0183] S204, if the timing value reaches the time threshold and the return intention index does not exceed the preset cancellation threshold since entering the lock preparation state, trigger the lock command.

[0184] This step is the composite condition decision link of the final lock operation. When the timing value of the timer has reached the time threshold in the previous step, the system will immediately perform the verification of the second condition. The system will query an internal state flag or backtrack the historical sequence of the return intention index recorded during the entire lock preparation state duration. This verification aims to confirm whether the peak value of the calculated return intention index remains below another preset numerical threshold, i.e., the "cancellation threshold," from the moment of entering the "lock preparation state" to the last moment of timing. The cancellation threshold represents the critical point at which the system recognizes the user's return intention as sufficiently clear. Only when both the "time condition" (the timing value reaches the preset time threshold) and the "intention condition" (the return intention index never exceeds the cancellation threshold) are met, the system will finally determine that the user's departure intention is continuous and unshaken.

[0185] At this time, the system will generate and send a lock command to the body control module of the target vehicle, completing the entire intelligent lock process.

[0186] S205, if the return intention index exceeds the cancellation threshold at any time before the timing value reaches the time threshold, exit the lock preparation state and cancel this lock operation.

[0187] This step is the key interrupt logic to achieve instant response to user intention change and operation undo. During the countdown of the lock preparation state, the system continuously calculates the return intention index and compares it with the preset undo threshold in real time. The priority of this comparison operation is extremely high. Once the system detects that the value of the currently calculated return intention index has exceeded the undo threshold at any time point before the timer reaches the preset time threshold, an interrupt event will be triggered immediately.

[0188] The interrupt event will enforce a series of operations: first, the system will immediately exit the current user state from the "lock preparation state" and may reset it to the "approaching state" according to the strength of the return intention index; second, the running timer will be immediately stopped and reset; finally, and most importantly, the lock process that has been initialized and is in a waiting execution state will be completely undone.

[0189] Through this mechanism, the system ensures that the user's change of mind to return the vehicle can be responded extremely sensitively and timely, avoiding the embarrassing experience of the vehicle being locked while the user has not walked far.

[0190] Figure 2 It is a technical path diagram of the method of the embodiment, and the core technical concept of the embodiment of the application embodies a systematic solution of multi-dimensional perception fusion and intelligent decision-making, and the technical idea can be summarized as a four-layer architecture of "double-source perception-dynamic analysis-intelligent arbitration-precise control".

[0191] The first layer is a double-source perception architecture. The embodiment discards the limitation of the traditional scheme of simply relying on the instantaneous strength of the Bluetooth signal, and innovatively constructs a dual perception system of Bluetooth signals and inertial measurement data. The Bluetooth signal reflects the relative position relationship between the user and the vehicle, and the inertial measurement data reveals the absolute motion state of the mobile terminal itself, and the two independently observe the same user behavior from different physical dimensions, laying a solid foundation for subsequent cross-validation.

[0192] The second layer is a dynamic analysis mechanism. Compared with the traditional static threshold judgment, the embodiment introduces a dynamic feature extraction technology based on time series. By calculating the slope and variance of the Bluetooth signal, the system can deeply understand the trend and stability of the user behavior, upgrading the "point" judgment to the analysis of the "process". At the same time, the introduction of the finite state machine enables the system to have the ability to remember and model complex user behavior trajectories.

[0193] The third layer: intelligent arbitration mechanism. The core innovation of the embodiment is to design a complete conflict detection and arbitration logic. When the results of double-source perception are contradictory, the system does not simply select one of them, but dynamically evaluates the credibility of each data source based on the logical consistency of the physical world and the characteristics of the sensor, and realizes intelligent decision-making. This arbitration mechanism effectively solves the misjudgment problem of a single sensor in a complex environment.

[0194] The fourth layer: precise control strategy. In the final instruction triggering link, the embodiment realizes the leap from coarse control to fine control. For the locking operation, the system not only requires state conversion, but also introduces a dynamic time threshold and a return intention monitoring mechanism to ensure that the lock is only executed when the user has a stable intention to leave. This multi-protection mechanism fundamentally solves the core pain point of frequent false locking in the prior art.

[0195] In summary, the embodiment realizes the deep understanding and accurate response of the user's true intention by constructing a complete closed-loop control system from perception to decision-making to execution, which represents the technical paradigm shift of the keyless entry system from simple signal processing to intelligent behavior analysis.

[0196] The method provided by the above embodiment can be executed by a vehicle lock control system including an electronic device. The electronic device in the embodiment of the application is described from the perspective of hardware processing below. Please refer to Figure 3 , which is a schematic diagram of an entity device structure of the electronic device in the embodiment of the application.

[0197] It should be noted that Figure 3 The structure of the electronic device shown is only an example and should not impose any limitation on the functions and use range of the embodiment of the application.

[0198] As Figure 3 shown, the electronic device includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 402 or programs loaded from a storage portion 408 to a random access memory (RAM) 403, such as performing the method described in the above embodiment. In the RAM 403, various programs and data required for system operation are also stored. The CPU 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0199] The following components are connected to the I / O interface 405: an input section 406 including an audio input device, a push button switch, and the like; an output section 407 including a Liquid Crystal Display (LCD), and an audio output device, a lamp, and the like; a storage section 408 including a hard disk and the like; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication section 409 performs a communication process via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as necessary. A removable recording medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like, is attached to the drive 410 as necessary so that a computer program read therefrom can be installed into the storage section 408 as necessary.

[0200] In particular, the processes described above with reference to the flow charts can be implemented as a computer software program in accordance with embodiments of the present application. For example, an embodiment of the present application includes a computer program product comprising a computer program carried on a computer readable medium, the computer program containing computer programs for executing the methods shown in the flow charts. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 409, and / or installed from the removable recording medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the present application are performed.

[0201] Note that specific examples of the computer readable storage medium can include but are not limited to one or more of a conduit with one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read only memory (ROM), an erasable programmable read only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, the computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0202] The computer program product of the present application can be a computer program embodied on a non-transitory computer readable medium. When the program is executed by a computer, it can achieve the objectives of the present application as discussed above. Accordingly, the present application is not limited to any specific combination of hardware circuitry and software but can be implemented in any desired way.

[0203] In particular, the electronic device of the embodiment includes a processor and a memory coupled with the one or more processors, the memory configured to store computer program code comprising computer instructions to be invoked by the one or more processors to cause the electronic device to perform the method provided by the above embodiment.

[0204] As another aspect, the present application also provides a computer readable storage medium, which can be included in the electronic device described in the above embodiment, or can exist separately without being assembled into the electronic device. The storage medium carries one or more computer programs, which, when executed by a processor of the electronic device, cause the electronic device to implement the method provided in the above embodiment.

[0205] The above-described embodiments are merely intended for describing the technical solutions of the present application, not to limit the present application; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

[0206] In the above embodiments, according to the context, the term "when" can be interpreted as meaning "if" or "after" or "in response to determining" or "in response to detecting". Similarly, according to the context, the phrase "on determining" or "if detecting (the stated condition or event)" can be interpreted as meaning "if determining" or "in response to determining" or "on detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)".

[0207] Those skilled in the art can understand that all or part of the processes in the above-mentioned method embodiments can be implemented by a computer program instructing relevant hardware to complete, the program can be stored in a computer readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The aforementioned storage medium includes ROM or random storage memory RAM, magnetic disc or optical disc and various storage code medium.

Claims

1. A method for controlling a car lock based on trajectory prediction and intention recognition, characterized in that, The method comprises: continuously collecting, by a Bluetooth module of a target vehicle, a Bluetooth signal feature parameter from a mobile terminal at a preset frequency to form a Bluetooth signal time series data; receiving, from the mobile terminal, real-time collected inertial measurement data; calculating, based on the Bluetooth signal time series data, a dynamic characteristic value for representing a movement process of the mobile terminal relative to the target vehicle, the dynamic characteristic value comprising a slope and a variance of the Bluetooth signal time series; and generating, based on the inertial measurement data, a first judgment result for representing an absolute motion state of the mobile terminal itself; inputting the dynamic characteristic value and the first judgment result into a preset finite state machine, and generating, based on the dynamic characteristic value, a second judgment result about a user state, the user state comprising a far-away state, a close-in state and an in-vehicle state; comparing the first judgment result and the second judgment result to obtain a comparison result; if the comparison result meets a preset conflict condition, selecting the first judgment result or the second judgment result as an effective output according to a preset arbitration logic; if the comparison result does not meet the preset conflict condition, using the second judgment result as the effective output; updating a current user state of the mobile terminal according to the effective output; when the updated user state of the finite state machine is converted from the far-away state to the close-in state, triggering an unlocking instruction to make the target vehicle in an unlocked state; when the updated user state of the finite state machine is converted from the close-in state or the in-vehicle state to the far-away state, and the far-away state lasts stably for more than a preset time threshold, triggering a locking instruction to make the target vehicle in a locked state.

2. The method of claim 1, wherein, The method further comprises: extracting acceleration and angular velocity data from the inertial measurement data; calculating, based on the acceleration and angular velocity data, a real-time motion trajectory of the mobile terminal in a three-dimensional space by a pose solving algorithm; generating, according to a projection component of the real-time motion trajectory on a horizontal plane, the first judgment result for indicating that the mobile terminal is approaching the target vehicle, moving away from the target vehicle or lingering around the target vehicle.

3. The method according to claim 1 or 2, characterized in that, The arbitration logic comprises: when the preset conflict condition is that the first judgment result indicates that the mobile terminal is in a lingering state and the second judgment result indicates that the user state has been converted to the far-away state, selecting the first judgment result as the effective output; when the preset conflict condition is that the first judgment result indicates that the mobile terminal is in a stable approaching or stable moving away state, and the second judgment result indicates that the user state is an out-of-vehicle stationary state because the variance of the Bluetooth signal time series is continuously higher than a variance threshold, selecting the first judgment result as the effective output.

4. The method of claim 3, wherein, The arbitration logic is a dynamic weight arbitration logic, and execution of the dynamic weight arbitration logic comprises: calculating, in real time, a data reliability weight of the inertial measurement data and a data reliability weight of the Bluetooth signal time series data, wherein the data reliability weight of the inertial measurement data is calculated based on a signal noise level and a zero bias stability, and the data reliability weight of the Bluetooth signal time series data is calculated based on a signal-to-noise ratio and a multipath effect interference degree thereof; fusing the first judgment result and the second judgment result respectively with the corresponding data reliability weight to generate a final user state judgment after fusion; taking the final user state judgment as the valid output.

5. The method of claim 1, wherein, The user state further includes an outside-still state, and the method further includes: when detecting that a variance of the Bluetooth signal time series data continuously exceeds a preset variance threshold in the approaching state or the away state, determining that the current user state is the outside-still state, and delaying triggering of a lock control instruction.

6. The method of claim 1, wherein, The time threshold is a variable value calculated dynamically, and the dynamic calculation includes: when detecting that the user state is converted from the approaching state or the in-vehicle state to the away state, extracting a slope stability and a slope absolute value of the Bluetooth signal time series data before state conversion; quantitatively evaluating a decisiveness of a one-time leaving behavior based on the slope stability and the slope absolute value; setting the time threshold according to the decisiveness, wherein the time threshold is in an inverse relationship with the decisiveness, so that a high decisiveness corresponds to a shorter time threshold, and a low decisiveness corresponds to a longer time threshold.

7. The method of claim 1, wherein, After the user state is converted from the approaching state or the in-vehicle state to the away state, and before the lock instruction is triggered, the method further includes: starting a countdown of the time threshold when the lock control system enters a lock preparation state; during a duration of the lock preparation state, calculating a return intention index for quantifying a user return possibility based on the latest collected signal time series data; judging whether a timer value reaches the time threshold; if the timer value reaches the time threshold and the return intention index does not exceed a preset cancellation threshold since entering the lock preparation state, triggering the lock instruction; if the return intention index exceeds the cancellation threshold at any time before the timer value reaches the time threshold, exiting the lock preparation state and canceling this time of lock operation.

8. An electronic device, comprising: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is configured to store computer program codes including computer instructions, and the one or more processors are configured to invoke the computer instructions to enable the electronic device to perform the method in any one of claims 1-7.

9. A computer readable storage medium storing computer instructions, characterized in that, The computer instructions, when executed on an electronic device, enable the electronic device to perform the method in any one of claims 1-7.

10. A computer program product, characterised in that, The computer program product, when executed on an electronic device, enables the electronic device to perform the method in any one of claims 1-7.

Images