UWB using method

Abstract

The invention belongs to the technical field of wireless communication positioning, and provides a UWB (Ultra Wideband) use method, which comprises the following steps of: performing signal detection on a space around a vehicle body by using a UWB antenna array deployed on an electric vehicle so as to obtain dynamic signal data related to a user or an object; based on a preset scene triggering condition, activating a corresponding target detection mode from a plurality of detection modes; under the target detection mode, the dynamic signal data are analyzed and processed to recognize the operation intention of a user or judge the safety state of the environment where the vehicle is located; according to an operation intention or an environment safety state, a corresponding vehicle control instruction is generated and executed, through scene fusion of UWB and multi-source sensing information, conversion from passive response to active intelligent interaction is achieved, multiple functions such as non-inductive unlocking, natural interaction, accurate vehicle searching and graded anti-theft are achieved through one system, and the system is suitable for popularization and application. The user experience and the vehicle safety are remarkably improved, and meanwhile the system complexity and the false triggering rate are reduced.

Description

Technical Field

[0001] This invention belongs to the field of wireless communication positioning technology, specifically a method for using UWB. Background Technology

[0002] With the increasing popularity of electric two-wheelers, users are demanding more intelligent, safer, and more convenient vehicles. Currently, most electric vehicles on the market use traditional physical keys, remote controls, or simple proximity unlocking solutions based on Bluetooth. For theft prevention, they primarily rely on vibration sensors to trigger alarms, but these solutions are prone to false alarms due to environmental interference (such as wind or passing vehicles), resulting in a poor user experience. In terms of user interaction, the functions are limited, usually requiring users to manually operate the key, remote control, or mobile app to perform functions such as finding the vehicle and unlocking it, which is not convenient enough.

[0003] While existing technologies employ UWB technology for precise positioning, they are mostly limited to simple distance detection or key replacement. They fail to fully utilize UWB's high precision, high resolution, and strong anti-interference capabilities, and fail to deeply integrate it with other vehicle sensors (such as Bluetooth and IMU) to achieve scenario-based interaction and security protection, moving from passive response to proactive intelligent perception. For example, how to distinguish between unintentional and intentional approach by the vehicle owner, how to determine the rider's riding status without contact to replace fragile seat sensors, how to provide a safe and convenient way to answer calls, and how to effectively filter interference from pets and other sources for precise theft prevention are all problems that existing technologies have not adequately addressed.

[0004] Therefore, the present invention provides a method for using UWB. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for using UWB to solve at least one technical problem existing in the prior art.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: A method for using UWB includes the following steps: Step S10: Use the UWB antenna array deployed on the electric vehicle to detect signals in the space around the vehicle to obtain dynamic signal data related to the user or object. Step S20: Based on at least one preset scene triggering condition, activate the corresponding target detection mode from multiple predefined UWB detection modes; Step S30: In the target detection mode, the dynamic signal data is analyzed and processed to identify the user's operating intention or to determine the safety status of the vehicle's environment. Step S40: Generate and execute the corresponding vehicle control command according to the operation intention or the environmental safety status.

[0007] In a further technical solution of the present invention, the environmental perception step in step S10 further includes synchronously acquiring auxiliary information from non-UWB perception modules. The auxiliary information includes at least: detecting the connection status and approximate distance information of authorized mobile terminal devices through a Bluetooth module, and acquiring the vehicle's own motion state information through one or more of the vehicle's built-in inertial measurement unit, wheel speed sensor, and vibration sensor. The intent and state parsing step in step S30 is performed based on the fusion processing result of the original signal data and the auxiliary information.

[0008] In a further technical solution of the present invention, the UWB detection mode in step S20 includes gesture trajectory mode, physiological micro-motion mode and near field sensing mode; The activation conditions for the gesture trajectory or near-field sensing mode are both that an authorized mobile terminal is detected to have entered the first preset distance and that the vehicle can be unlocked. In gesture trajectory mode, step S30 matches the user's hand trajectory captured by UWB with the pre-stored unlock template. If the matching degree meets the standard and the terminal is located at the second preset distance, the unlocking intention is determined to be established. In near-field sensing mode, step S30 determines whether the user coordinates calculated by UWB are located in the preset area in front of the seat cushion. If so and the terminal is connected, the unlocking intention is determined to be valid. Step S40 controls the vehicle to unlock according to the unlocking intention.

[0009] In a further technical solution of the present invention, another triggering condition for the activation of the gesture trajectory mode in step S20 is that the vehicle central control system enters a specific interactive enable state. In gesture trajectory mode, the user's operation intention command parsed in step S30 is the control intention of the vehicle device. The parsing process includes: recognizing the user's head nodding / shaking motion or the sliding trajectory of the hand in a specific direction captured by the UWB antenna array, and mapping it to the confirmation / rejection of the vehicle voice command or the switching command of the instrument interface option; step S40 then controls the vehicle audio, instrument or corresponding functional module to perform the corresponding operation.

[0010] In a further technical solution of the present invention, another triggering condition for the activation of the gesture trajectory mode in step S20 is that the vehicle enters the vehicle search assistance state, which is triggered by an authorized mobile terminal device entering a third preset distance range that is farther than the first preset distance. In gesture trajectory mode, the user's operation intention command parsed in step S30 is a vehicle search trigger intention. The parsing process includes: recognizing the preset vehicle search gesture action captured by the UWB antenna array; and step S40 controls the vehicle's audio-visual system to issue a location prompt signal.

[0011] In a further technical solution of the present invention, the triggering condition for the activation of the physiological micro-motion mode in step S20 is that the vehicle is in a protected state and no authorized mobile terminal device is detected within a preset safety range; in the physiological micro-motion mode, the environmental safety status assessment in step S30 includes the determination of the presence of living beings, the process of which is: performing time-frequency analysis on the reflected signal received by the UWB antenna array, detecting whether there are periodic micro-motion signal features representing breathing or heartbeat in the frequency range of 0.1Hz to 0.5Hz, thereby determining whether there are unauthorized living beings around the vehicle.

[0012] In a further technical solution of the present invention, the environmental safety status assessment in step S30 also includes comprehensive risk assessment and classification. The process is as follows: based on the determination of the presence of a living body, the vibration duration and intensity information from the vehicle vibration sensor, the vehicle position change information from the displacement sensor, and the vehicle tilting state information from the attitude sensor are integrated, and the risk is accumulated and calculated according to the preset risk factor scoring model. The risk is divided into multiple levels according to the total accumulated score.

[0013] In a further technical solution of the present invention, the control execution step in step S40 specifically includes a graded response mechanism: triggering composite protection actions that are progressively enhanced from low to high according to different risk levels; wherein, the low-level response action includes logging and APP message push; the intermediate-level response action adds vehicle audible and visual warnings on the basis of the low-level action; and the high-level response action adds locking the vehicle drive motor and activating satellite positioning tracking and reporting functions on the basis of the intermediate-level action.

[0014] In a further technical solution of the present invention, step S30 includes a step of detecting and repairing outliers in the gesture trajectory data before recognizing the motion trajectory. Specifically, it includes: based on independent coordinate sequences from multiple UWB antennas, suspicious outliers are initially marked through local statistical feature analysis, kinematic continuity test and signal quality assessment. The detection results from multiple antennas are spatiotemporally aligned and cross-validated to identify real outliers and assign confidence labels. Real outliers are repaired by interpolation using consistent data from multiple antennas or adjacent normal points, and then clean 3D gesture trajectory data is obtained through physical constraint optimization and adaptive filtering for subsequent intent recognition.

[0015] In a further technical solution of the present invention, the process of intention recognition based on the clean 3D gesture trajectory data specifically includes: A confidence-weighted dynamic time warping algorithm is used to match the trajectory to be identified with a pre-stored standard gesture template trajectory. Each frame coordinate of the trajectory is assigned a weight obtained by fusing signal quality and repair confidence, and low-confidence continuous regions are allowed to be skipped in the matching path. The system integrates multi-scale matching results and performs secondary verification using Bluetooth proximity context information. Finally, it outputs the recognized gesture command through threshold decision.

[0016] The beneficial effects of this invention are as follows: Through the collaboration of UWB and Bluetooth, a shift has been achieved from requiring active user operation to the system proactively sensing and responding. Users can unlock their vehicles and locate them using natural gestures without needing to take out their phones or keys, experiencing a smooth and seamless experience. Utilizing UWB's centimeter-level positioning and micro-motion detection capabilities, combined with multi-sensor information fusion algorithms, the probability of false triggers (such as false locking or false alarms) is greatly reduced, and the accuracy of intent recognition and status judgment is improved. A single UWB hardware system supports multiple core scenarios, from keyless entry, intelligent vehicle location, in-vehicle interaction to active anti-theft, replacing traditional seat cushion sensors, some millimeter-wave radars, and simple vibration alarms, reducing system complexity and cost. The innovative liveness detection and multi-factor risk assessment model can distinguish between minor interference and real threats, achieving a tiered response from reminders and warnings to physical intervention. This avoids annoying false alarms and provides powerful active anti-theft capabilities, translating natural body language such as head movements and gestures into control commands (such as answering calls or changing songs), making interactions during driving safer and more convenient, and reducing the risk of drivers being distracted by physical buttons. Attached Figure Description

[0017] Figure 1 This is a flowchart of the steps of the UWB usage method in an embodiment of the present invention; Figure 2 This is a logical schematic diagram of the UWB usage method in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the steps of performing DTW dynamic time warping and matching on gesture trajectories in the UWB usage method of this invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described in a non-limiting manner below with reference to the accompanying drawings and specific embodiments.

[0019] Example 1, as Figures 1-2 As shown, a method for using UWB in this embodiment includes the following steps: Step S10: Use the UWB antenna array deployed on the electric vehicle to detect signals in the space around the vehicle to obtain dynamic signal data related to the user or object.

[0020] The implementation process of step S10 is as follows: Multiple UWB antennas are pre-installed at different locations on the electric vehicle body (such as the front, rear, and sides) to form an antenna array, wherein the antennas are connected to the UWB core module via cables; The UWB module periodically emits extremely short pulses (nanosecond level) of radio waves through an antenna array. When the pulses encounter users (such as hands, heads, bodies) or objects around the vehicle, they are reflected, and the receiving antenna captures these reflected signals. The UWB module processes the transmitted and received signals to calculate the raw dynamic signal data, which includes: Distance information: Based on time-of-flight (ToF), determine the distance between the user / object and the antenna; Angle / direction information: Estimate the direction of the user / object relative to the antennas by using the phase difference (such as angle of arrival AoA) of signals received from multiple antennas; Signal characteristic changes: Due to the slight movements of the user / object (such as gestures, breathing, heartbeat), the phase and amplitude of the reflected signal will undergo subtle, regular or irregular changes. It should be noted that dynamic signal data is a set of real-time, continuous dynamic data streams that describe changes in the space surrounding the vehicle.

[0021] Step S20: Based on at least one preset scene triggering condition, activate the corresponding target detection mode from multiple predefined UWB detection modes.

[0022] The implementation process of step S20 is as follows: Within the vehicle's central control system, multiple scenario trigger conditions are pre-programmed, typically related to vehicle status and non-UWB external signals, such as: Condition A: Bluetooth detects that the paired mobile phone has entered within 10 meters, and the vehicle is in a locked state; Condition B: The vehicle is armed and Bluetooth does not detect any paired devices within 5 meters; Condition C: The in-vehicle voice assistant is activated, or the music playback interface is opened; Meanwhile, multiple UWB detection modes are predefined, each corresponding to a set of UWB module operating parameters and data processing focus, for example: Mode X (Gesture Trajectory Mode): Focuses on high-frequency sampling to accurately track the continuous movement path of the hand for gesture recognition; Mode Y (Physiological Micro-motion Mode): Focuses on analyzing the low-frequency periodic components of the signal, used to detect micro-motions such as breathing and heartbeat; Mode Z (Near Field Sensing Mode): Focuses on the accuracy of spatial positioning and the stability of static coordinates. It obtains the three-dimensional coordinates of the target in real time through existing multi-antenna array joint calculation (such as TDOA+AOA fusion algorithm) and combines the signal strength (RSSI) to determine the area for contactless unlocking and presence detection. The system continuously monitors the vehicle status. Once the current status meets a certain scenario trigger condition (such as condition A), the system will issue a command to control the UWB module to switch from standby or the previous mode to the target detection mode bound to the scenario trigger condition (such as switching to mode X).

[0023] Step S30: In the target detection mode, the dynamic signal data is analyzed and processed to identify the user's operating intention or to determine the safety status of the vehicle's environment.

[0024] The implementation process of step S30 is as follows: In the target detection mode activated in step S20, in-depth analysis is performed on the dynamic signal data acquired in S10, specifically including: In gesture trajectory mode, the spatial coordinate sequence of the hand is extracted from the data stream, smoothed and filtered, and then compared with several preset templates (such as "waving" and "drawing circles") to calculate the similarity. In physiological micro-movement mode, time-frequency analysis (such as short-time Fourier transform) is performed on the signal to look for stable spectral peaks in the range of 0.1-0.5Hz, so as to determine whether there are signs of live respiration. The deep analysis results are mapped to predefined intents or status labels. For example, the user's operational intent is identified: the user wants to unlock the vehicle, the user wants to change the song, the user wants to find the car, etc.; the safety status of the vehicle's environment is determined: the vehicle is surrounded by suspicious living beings, the vehicle is in a safe and stationary state, or the vehicle is at risk of being moved.

[0025] Step S40: Generate and execute the corresponding vehicle control command according to the operation intention or the environmental safety status.

[0026] The implementation process of step S40 is as follows: Based on the output of step S30, a preset "action mapping table" is queried to generate one or more specific, low-level vehicle control commands. For example, if the intention is to unlock, the command may be: close the main relay or open the seat lock; if the status is high-risk vehicle theft, the command may be: trigger a 120-decibel alarm, lock the motor controller, or send GPS coordinates to the cloud. The central control unit sends these instructions to the corresponding actuators on the vehicle (such as lock controller, motor controller, alarm, instrument panel, and lighting system) via the corresponding bus (such as CAN, LIN) or I / O port. The actuator receives and executes instructions to change the physical state of the vehicle (such as unlocking, honking the horn, turning on the lights, or locking the motor).

[0027] Example 2: This example provides a UWB system hardware architecture, specifically including the following parts: Central Control Unit (MCU): As the core processor of the system, it is responsible for running control logic, processing sensor data, executing communication protocols and issuing control commands. In this embodiment, the central control unit integrates a 4G communication module and a Bluetooth module (supporting BLE, especially HID protocol), which are used for remote data interaction and near-field connection with user's personal devices (such as mobile phones and smartwatches), respectively. UWB module: It communicates with the central control unit via CAN bus (or alternatively, via 485 bus) for high-speed and reliable data communication. The UWB module is connected to multiple UWB antennas. UWB antenna array: Multiple UWB antennas are strategically arranged on the vehicle body. For example, one (ANT1) is placed inside the dashboard at the front of the vehicle for frontal close-range interaction and posture recognition; one (ANT2, ANT3) is placed on each side of the vehicle body (such as under the foot pedals or on the side of the seat) for side proximity detection and liveness detection; and one (ANT4) is placed at the rear of the vehicle (such as at the taillights) for rear proximity detection. This forms a multi-angle, multi-baseline signal coverage of the space around the vehicle, providing a hardware foundation for accurate positioning and posture recognition. Vehicle status sensor group: including but not limited to inertial measurement unit (IMU) (for detecting vehicle tilt angle and acceleration), wheel speed sensor (for detecting wheel movement status), GPS / BeiDou module (for detecting displacement), vibration sensor, the sensors are connected to the central control unit through corresponding interfaces (such as ADC, SPI, I2C); Actuators include motor controllers (for controlling motor locking), audible and visual alarms (buzzers, LED lights), instrument panel HMI, seat lock controllers, vehicle power management systems, etc., and are controlled by the central control unit.

[0028] Example 3, based on the system hardware architecture provided in Example 2, also provides a method for using UWB, specifically including the following steps: S201: Continuous environmental awareness. The system continuously runs low-power Bluetooth scanning in the background to listen for broadcast signals from bound authorized devices (such as the owner's mobile phone) and estimate their distance through RSSI or more advanced AoA technology. At the same time, vehicle status sensor data is periodically read. S202: Determine the scene triggering conditions. The central control unit matches the currently acquired perception information (such as Bluetooth device distance, vehicle arming status, central control interface status, etc.) with multiple pre-set scene triggering rules. S203: Activate the corresponding UWB detection mode. If a certain scene is matched, send a command to the UWB module through the CAN bus to configure its working parameters (such as activating a specific antenna, setting the detection frame rate, and switching the recognition algorithm) and enter the corresponding detection mode (such as gesture trajectory recognition mode or life micro-motion detection mode). S204: Collects and analyzes UWB data. In the active mode, the UWB module sends pulse signals and receives reflected signals through the antenna array, calculates the original data streams such as distance, phase, and signal strength, and uploads them to the central control unit. S205: Multi-source information fusion and decision-making. The central control unit runs the corresponding recognition algorithm (such as performing DTW dynamic time warping matching on the gesture trajectory and extracting the breathing frequency by performing time-frequency analysis on the reflected signal), and makes a comprehensive judgment by combining the current Bluetooth connection status, whether the vehicle is turning, and other information. S206: Execute control response, issue instructions to the corresponding execution mechanism based on the decision result, complete the interaction or protection action, and then return to the continuous sensing state according to the scenario exit conditions.

[0029] Example 4, a UWB usage method described in this embodiment of the invention, mainly for keyless gesture unlocking scenarios, specifically includes the following steps: S301: Bluetooth proximity trigger. When the owner approaches the vehicle with a bound mobile phone within about 5 meters, the vehicle's Bluetooth module and the mobile phone automatically complete the security authentication connection. The central control unit determines that the trigger conditions for the "keyless unlocking" scenario are met (authorized device enters the first preset distance). S302: Activate UWB gesture recognition mode. The central control unit instructs the UWB module to enter high-precision gesture trajectory recognition mode, focusing on activating the antennas at the front of the vehicle (ANT1) and possibly on the sides to cover the area where the user is standing in front of the side of the vehicle. S303: Gesture acquisition and preprocessing. The UWB module continuously captures the spatial position changes of the user's hand, forming a three-dimensional trajectory point sequence. The central control unit performs filtering and normalization processing on the trajectory. S304: Gesture matching compares the processed trajectory with the unlocking gesture template pre-recorded and stored by the vehicle owner (e.g., "quickly push forward twice"). The matching algorithm must simultaneously satisfy the requirements that the trajectory shape similarity and the movement speed / amplitude are within a certain threshold. S305: Fusion Verification. During gesture matching, the system simultaneously verifies: 1) whether the mobile phone's Bluetooth connection is stable and at close range (e.g., <1.5 meters); 2) whether the vehicle is stationary (wheel speed zero) and unarmed or in a state awaiting unlocking. If gesture matching is successful and fusion verification passes, it is recognized as a valid "unlocking intent." S306: Unlocking is performed. The central control unit sends commands to the power management system and the under-seat lock controller, the vehicle automatically powers on, the under-seat locks unlock, the welcome screen illuminates on the instrument panel, and the keyless unlocking process is completed.

[0030] In addition to the gesture unlocking mentioned above, this embodiment also supports contactless unlocking logic based on UWB electronic fences, namely near-field sensing mode, the specific process of which is as follows: S311: Area monitoring. The system presets a riding preparation area in the vehicle coordinate system. The riding preparation area is defined as a three-dimensional space area with the geometric center of the vehicle seat as the origin, extending 0 to 60cm towards the front of the vehicle and 0 to 40cm towards both sides of the vehicle body, and with a height between 0.8 meters and 1.8 meters. S312: Coordinate Calculation and Judgment. After the Bluetooth connection is established, the UWB module calculates the three-dimensional coordinates (x, y, z) of the user's center of gravity in real time, and the intent analysis module continuously determines whether the coordinates fall within the riding preparation area. S313: Intent Confirmation. The system sets a dwell time threshold (e.g., 1.5 seconds). The system only confirms the user's driving intent if the user's coordinates continuously fall within this area for more than the dwell time threshold, and the Bluetooth signal strength (RSSI) is greater than a preset signal threshold (e.g., RSSI value > -50dBm). S314: Execution. Once the intent is confirmed, the system skips the gesture matching step and executes the vehicle power-on and unlocking operation in step S306. It should be noted that the near-field sensing mode utilizes the centimeter-level positioning characteristics of UWB to distinguish between passing by (coordinates outside the area) and preparing to ride (coordinates within the area), achieving seamless unlocking while reducing accidental touches.

[0031] Example 5: This example describes a UWB usage method, primarily for intelligent live anti-theft scenarios, and specifically includes the following steps: S401: Arming mode is activated. The vehicle owner can set the vehicle to anti-theft mode via the app, key, or by pressing and holding the P button. At this time, the vehicle is powered off, and the central control unit records the arming time. Among them, entering the armed state can also be when UWB detects a human body beyond the seat cushion range and the central control determines that the vehicle is in a stopped state. S402: Check for authorized devices. The system continuously monitors the surrounding area for authorized Bluetooth devices. If an authorized device is detected near the vehicle (within 3 meters) within a certain period after arming (e.g., 30 seconds) or during continuous monitoring, the system assumes the vehicle owner is still active and will not activate advanced anti-theft detection, or will only activate low-sensitivity detection. S403: Trigger liveness detection. When the system confirms that no authorized device is within the preset security range (e.g., within 5 meters), it determines that the "Advanced Anti-theft" trigger condition is met. S404: Activate UWB vital sign detection mode. The central control unit instructs the UWB module to switch to vital sign detection mode. In this mode, the UWB module transmits pulses at a specific frame rate and receives reflected signals with high sensitivity. Through algorithms, it separates the micro-Doppler effect caused by chest cavity fluctuations or the periodic changes in signals caused by body movement. S405: Liveness detection and risk assessment. The system analyzes UWB signals within a preset duration (e.g., 10 seconds). If a significant energy peak is detected in the frequency range of 0.15-0.3Hz (respiratory frequency band) or a frequency band related to heart rate, it is determined that "a live person is present." Simultaneously, the system reads data from the IMU (to determine if the vehicle has been lifted or violently shaken) and vibration sensors. S406: Risk Factor Scoring and Classification, with real-time scoring by the system (example): Detection of a continuous live signal for more than 8 seconds: +30 points; Detection of brief vibration (<2 seconds): +10 points; Detection of continuous vibration (≥2 seconds): +20 points; GPS detection of displacement >200 meters: +40 points; IMU detection of a vehicle tipping over: +30 points; Current time is nighttime (18:00-6:00): +5 points; as shown in Table 1 below; Table 1: Risk Factor Scoring and Classification Based on the total score, the risk levels are divided into: safe (<30 points), suspicious (30-60 points), high risk (60-85 points), and confirmed car theft (≥86 points), as shown in Table 2 below; Table 2: Risk Level Classification S407: Tiered Response Execution: Suspicious Level: A reminder message containing the time and risk type is pushed to the car owner's APP via the 4G network, and the vehicle's taillights flash slowly; High-risk level: Based on the suspicious level response, add an urgent alarm sound from the vehicle horn, display a warning message on the dashboard (if it can be woken up), and send a strong reminder notification to the APP; Theft Confirmation: Based on the aforementioned response, the central control unit sends a command to the motor controller to lock the motor, making the vehicle unable to be pushed or ridden; at the same time, it uploads real-time GPS location information to the cloud monitoring platform via the 4G module, and can trigger the process of automatically dialing the reserved phone number.

[0032] Example 6 describes a UWB usage method, primarily targeting in-vehicle gesture interaction scenarios, which includes the following steps: When the user is riding, the in-vehicle voice assistant can be activated by a voice wake-up word (such as "Hello, Xiaoxin"); The voice assistant announces: "Gesture control is enabled." The central control unit then sends a command to the UWB module via the CAN bus to activate the gesture recognition mode facing the driver area (mainly using the ANT1 at the front of the vehicle). When a user makes a "wave to the right" gesture in front of the dashboard, UWB captures the trajectory and identifies it as a "next track" command; The central control unit sends the command to the in-vehicle music playback module to perform the song switching operation. If there is no subsequent gesture within 30 seconds, the system will automatically exit the mode to save energy.

[0033] Example 7: This example describes a UWB usage method, primarily for vehicle location scenarios, and includes the following steps: After receiving the command to enter vehicle location assistance mode, the vehicle activates the UWB dynamic attitude detection mode and enters the listening state. The UWB module detects the driver's specific posture and movement (the preset "waving back and forth" movement). After the central control unit confirms this, it controls the vehicle to sound the horn twice and the hazard lights to flash, helping the driver to quickly locate the vehicle.

[0034] Example 8, a UWB usage method described in this example, mainly addresses the problem that outliers affect the accuracy of operation intent recognition during DTW dynamic time warping matching of gesture trajectories in the process of recognizing user operation intent. It also includes the following steps: Step 1: Deploy four UWB antennas at the front, sides, and rear of the vehicle to acquire hand-reflected signals in real time. Calculate the time of flight and angle of arrival to convert multiple raw signals into a 3D spatial coordinate sequence with a unified time reference. In step one, the process of obtaining the 3D spatial coordinate sequence is as follows: Fixed deployment of UWB antennas at four locations: front (directly forward), left side, right side, and rear (directly backward) of the vehicle, ensuring that the antenna spacing is ≥ 1 / 3 of the horizontal / vertical axis distance of the device, and that the antenna transmitting / receiving surfaces face the gesture interaction area (preset to be within 0.5-3m of the device); debug the operating parameters of the UWB module, setting the carrier frequency to 3.5-6.5GHz (ultra-wideband), signal sampling rate ≥ 100Hz, and transmit power ≤ 10dBm, ensuring that the four antennas are synchronously in transceiver mode to receive UWB echo signals reflected from the hand in real time; Four antennas independently acquire raw Time of Flight (ToF) and Angle of Arrival (AoA) data of hand-reflected signals. ToF records the time difference from signal transmission to reception, and AoA acquires the azimuth (horizontal direction) and elevation (vertical direction) respectively. At the same time, the acquisition timestamp of each frame of signal is recorded (accurate to the microsecond level), forming four independent raw signal streams. Based on the joint calculation algorithm of ToF and AoA, a 3D spatial coordinate calculation model is established: with the geometric center of the device as the origin of the coordinate system, the front-to-back direction of the vehicle body is the X-axis, the left-to-right direction is the Y-axis, and the vertical direction to the ground is the Z-axis. The straight-line distance from the antenna to the hand is calculated by ToF, d=c×ToF / 2 (c is the speed of light). Combined with the azimuth angle θ and elevation angle φ of AoA, the 3D coordinates of the hand under the single antenna view are calculated as (x,y,z)=(d×cosφ×cosθ,d×cosφ×sinθ,d×sinφ). The acquisition timestamps of the original signals from the four antennas are extracted. Using the system clock of the device's main controller as a unified time reference, the calculated coordinates of the four antennas are time-synchronized and calibrated. Coordinates with timestamp deviations ≤50μs are frame-aligned, and coordinates with deviations >50μs are frame-filled using linear interpolation. Finally, a 3D spatial coordinate sequence Pi(x,y,z,t) with completely synchronized time axes for each of the four antennas is generated, where i=1,2,3,4 corresponds to the four antennas.

[0035] Step 2: Based on the independent 3D coordinate sequence of each antenna, perform single-antenna anomaly detection in parallel: calculate local statistical features to identify abrupt change points, use kinematic constraints to check the continuity of velocity / acceleration, combine UWB signal quality indicators to evaluate the credibility of each point, and initially mark suspicious outliers; In step two, the process of marking suspected outliers is as follows: A sliding time window is used to segment the 3D coordinate sequence of a single antenna. The window length is set to 5-10 frames (adaptive to the sampling rate, such as 50-100ms at a sampling rate of 100Hz). The step size is 1 frame to ensure that the coordinates of each frame are contained in at least one window, resulting in several local coordinate windows Wk={Pk1,Pk2,...,Pkn} (n is the number of coordinate frames in the window). For each sliding window, calculate local statistical features of the coordinate sequence, including the mean, variance, and standard deviation of the X / Y / Z coordinates, and the first difference of the three-axis coordinates (the change in coordinates between adjacent frames), to quantify the degree of fluctuation of the local trajectory; if the absolute value of the three-axis difference of a certain frame coordinate exceeds 3 times the standard deviation of the difference within the window, it is initially marked as a statistical abrupt change point; The hand movement speed is calculated by the time difference between the coordinates of adjacent frames, and a maximum speed threshold vmax for the gesture movement is set (usually 2m / s, which can be adjusted according to the application scenario). If the speed of a certain frame is greater than vmax, it is marked as a speed anomaly. The acceleration is calculated by the time difference between the speeds of adjacent frames, and a maximum acceleration threshold amax is set (usually 10m / s²). If the acceleration of a certain frame is greater than amax, it is marked as an acceleration anomaly. Extract the original UWB reflected signal quality indicators corresponding to each frame coordinate, the core of which include: Received Signal Strength (RSSI), Signal-to-Interference Ratio (SIR), and Time of Arrival Error (TRE); set quality thresholds: RSSI ≥ -80dBm, SIR ≥ 10dB, and TRE ≤ 10ns are considered acceptable for the signal. Conversely, if any of the signal indicators corresponding to a frame coordinate is lower than the threshold, it is marked as a signal quality anomaly point, and a confidence score Sq (0-1 point, the worse the indicator, the lower the score) is assigned according to the degree of indicator deviation. Establish multi-dimensional anomaly fusion rules: If the coordinates of a frame simultaneously satisfy the conditions of statistical mutation point + kinematic anomaly point, or signal quality anomaly point and the confidence score is <0.5, then it is marked as a suspicious outlier point; After each antenna completes full-sequence detection, it outputs its own 3D coordinate sequence and corresponding labeling results, including normal coordinate points and suspicious outliers, forming four independent "coordinate-anomaly labeling" intermediate results.

[0036] Step 3: Spatiotemporally align the detection results from the four antennas, perform cross-validation, confirm the real outliers, and output the set of outliers with spatial location and confidence labels; In step three, the process of outputting the set of outliers with spatial location and confidence labels is as follows: Using the unified time reference from step one as the core, the coordinate-anomaly marker results of the four antennas are precisely aligned at the frame level to ensure that the four coordinate points (including markers) at the same timestamp correspond one-to-one, forming a spatiotemporally aligned four-dimensional detection matrix: , among which, (L i (t) represents the labeling result of the i-th antenna at time t, where 0 represents a normal value and 1 represents a suspicious outlier. For four coordinate points at the same timestamp, calculate the mean spatial Euclidean distance Dmean and the deviation Dstd of the four coordinates. Set a spatial consistency threshold Dth (generally 0.1m, meaning the hand coordinate deviation calculated by the four antennas at the same time does not exceed 10cm). If Dmean < Dth and Dstd < 0.05m, the four coordinates are considered to be in the same spatial position. If Dmean ≥ Dth, the single coordinate point with excessive deviation is marked separately and regarded as a "spatial inconsistency point". Based on the spatiotemporally aligned four-dimensional detection matrix, a multi-antenna cross-validation rule is established, with the core principles being majority rule and spatial location assistance. ① If at the same timestamp, ≥3 antennas mark the coordinate point as a suspected outlier, and the spatial positions of the four coordinates are consistent, it is directly confirmed as a real outlier; ② If, at the same timestamp, two antennas are marked as suspicious outliers, and the anomaly type of both points is "signal quality anomaly" (such as being simultaneously blocked), and the signal quality scores of the other two antennas are both ≥0.8, they are determined to be false suspicious points and the markings are removed; ③ If ≤1 antennas are marked as suspicious outliers at the same timestamp, regardless of whether their spatial locations are consistent, they are all determined to be false suspicious points and the marking is canceled; ④ If the suspected outlier is a "spatial inconsistency point" and the signal quality score of that antenna is <0.3, it is confirmed as a real outlier (coordinate deviation caused by severe obstruction of a single antenna). For confirmed outliers, calculate the spatial location confidence score Sp and the antenna consistency confidence score Sc, and obtain the final confidence score Sf through weighted fusion: ① Sp is calculated based on the spatial deviation of the four coordinates, the smaller the deviation, the higher Sp (0-1 points); ② Sc is calculated based on the number of antennas marked as suspicious outliers, the more antennas, the higher Sc. The weighting is as follows: ; Integrate the verification results across all timestamps and output a set of labeled true outliers. , of which O m Let x, y, z, t be the spatial coordinates and timestamp of the real outlier, and T be the timestamp of the real outlier. o The anomaly type is (single antenna blockage / sudden motion change / signal interference).

[0037] Step 4: For real outliers, interpolation repair is performed using consistent data from multiple antennas or adjacent normal points to reconstruct the complete trajectory. Physical constraint optimization and adaptive filtering are then applied to obtain clean 3D gesture trajectory data. In step four, the process of obtaining clean 3D gesture trajectory data is as follows: For each real outlier, extract the multi-antenna consistent normal points in its temporal neighborhood: ① Using the outlier timestamp t as the center, take 2-3 frames forward and backward to form a time neighborhood [t-Δt, t+Δt]; ② Within the neighborhood, select coordinate points where all four antennas are marked as normal and their spatial positions are consistent (Dmean < Dth) to serve as the reference normal point set. p represents the number of baseline normal points; The interpolation method is selected based on the number of baseline normal points to ensure that the repaired coordinates conform to the trajectory continuity: ① If there are ≥2 reference normal points in the neighborhood, cubic spline interpolation (interpolation is performed on the X / Y / Z axes respectively) is used to fit the coordinate values ​​of the outlier points by the trajectory trend of the normal points, ensuring that the first and second derivatives of the interpolation points are continuous with the normal points before and after them; ② If there is only one reference normal point in the neighborhood, linear interpolation is used, and the coordinate values ​​are fitted with the time difference between the normal point and the outlier as the weight. ③ If there is no reference normal point in the neighborhood, the average value of multiple antennas is used for completion, and the average value of the coordinates of the same timestamp of the four antennas is taken as the repair value. All the repaired interpolation points and the original normal points are spliced ​​together in ascending order of timestamps to form a preliminary complete 3D gesture trajectory without outliers; Based on rigid body constraints and range of motion constraints of human hand gestures, the initial reconstructed trajectory is optimized as follows: ① Rigid body constraint: Ensure that the relative position of the hand coordinates in each frame conforms to the law of human joint movement during the hand gesture movement, and eliminate coordinate points that exceed the range of finger / wrist movement (such as Z-axis coordinates exceeding the 0-2m range); ② Motion smoothing constraint: For coordinate points with small jumps in the trajectory, local mean smoothing is used to make the velocity / acceleration changes of the trajectory conform to the law of natural motion and avoid "peak" jumps. An adaptive Kalman filter is used to denoise the initially reconstructed trajectory. Its core advantage is that it dynamically adjusts the filter parameters according to the motion state of the trajectory: ① When the trajectory is moving at a constant speed / low speed, the filter gain is increased to enhance the denoising effect; ② When the trajectory is moving at an acceleration / deceleration / fast speed, the filter gain is decreased to ensure the dynamic responsiveness of the trajectory. The trajectory after physical constraint optimization and adaptive filtering is the clean 3D gesture trajectory data.

[0038] Step 5: Based on the 3D gesture trajectory data, run the confidence-weighted improved DTW algorithm. Low-confidence areas are allowed to be skipped to avoid path distortion. Multi-scale matching results and contextual information such as Bluetooth proximity are fused, and the final gesture recognition command is output through threshold decision. In step five, the process of outputting the final gesture recognition command is as follows: Pre-collect standard 3D trajectories of various target gestures (such as up / down, left / right, forward / backward, circle, square, etc.), and perform the same processing (solution, denoising, optimization) as steps one to four on each standard trajectory to obtain standard template trajectories; add labels (such as "swipe up", "swipe down", "draw a circle clockwise") to each template trajectory, and classify and store them according to the trajectory length, movement direction, and feature points (such as trajectory extreme points, inflection points), and construct a template gesture trajectory library Tmodel={T1,T2,...,Tn} (n is the number of template gestures); Tclean extracts core features from clean 3D gesture trajectories, including: total trajectory length, range of motion along the X / Y / Z axes, extreme points (maximum / minimum coordinates), inflection points (coordinates where velocity direction changes), and time series trends of each axis coordinate, providing feature basis for multi-scale matching; Two core improvements are made to the traditional Dynamic Time Warping (DTW) algorithm to adapt to the recognition requirements of 3D gesture trajectories: ① Confidence weighting: Each frame coordinate of the clean trajectory is assigned a trajectory confidence score St, which is obtained by fusing the signal quality score Sq from step two and the repair confidence score Sf from step four (St=0.7×Sq+0.3×Sf, 0-1 points); ② Low confidence region skipping mechanism: A confidence threshold Sth=0.6 is set. If the consecutive frames St of a certain trajectory segment are less than Sth, the region is allowed to be skipped during the DTW matching process to avoid trajectory deviation in low confidence regions causing distortion of the overall matching path. The clean trajectory Tclean to be identified is matched with each standard template trajectory Tmodel in the template trajectory library using 3D spatial DTW: ① The X / Y / Z coordinates of the 3D trajectory are normalized (to eliminate the influence of dimensions) to obtain a normalized trajectory; ② The cumulative distance D between the trajectory to be identified and each template trajectory is calculated based on the improved DTW algorithm. The smaller the cumulative distance, the higher the trajectory similarity; ③ During the matching process, the matching constraints are strengthened for high-confidence frames (St≥0.6), requiring strict correspondence between frames; the constraints are relaxed for low-confidence frames (St<0.6), allowing skipping or partial matching. A multi-scale matching strategy is adopted, and the trajectory to be identified and the template trajectory are matched at "coarse scale (sliding window length of 10 frames)," "medium scale (window length of 5 frames)," and "fine scale (window length of 1 frame)" to obtain the cumulative matching distances D1, D2, and D3 at the three scales. The comprehensive matching distance Dsum = 0.2 × D1 + 0.3 × D2 + 0.5 × D3 is calculated by weighted fusion (fine scale has the highest weight to ensure local matching accuracy). The template trajectory with the smallest comprehensive matching distance is the preliminary matching result. Contextual information such as Bluetooth proximity is introduced to perform secondary verification on the initial matching results, improving the robustness of recognition: ① Extract hand proximity data from the Bluetooth module (if the distance between the hand and the device is within 0.5-3m, it is considered a valid proximity, assigned a value of B=1; otherwise, it is considered an invalid proximity, assigned a value of B=0); ② Establish a fusion decision model: calculate the comprehensive matching similarity S=1-Dsum / Dmax (Dmax is the maximum cumulative distance in the template library, S is 0-1 points), combine it with the Bluetooth proximity B, and obtain the final decision value F=S×B; ③ Set the recognition threshold Fth=0.7 (which can be adaptively adjusted according to the scenario). If F≥Fth, the initial matching result is confirmed as a valid recognition result; if F<Fth, it is judged as "unrecognized gesture"; based on the final decision result, output the gesture label corresponding to the template trajectory.

[0039] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and various changes and modifications can be made without departing from the spirit and scope of the invention; all such changes and modifications will fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for using UWB, characterized in that: Includes the following steps: Step S10: Use the UWB antenna array deployed on the electric vehicle to detect signals in the space around the vehicle to obtain dynamic signal data related to the user or object. Step S20: Based on at least one preset scene triggering condition, activate the corresponding target detection mode from multiple predefined UWB detection modes; Step S30: In the target detection mode, the dynamic signal data is analyzed and processed to identify the user's operating intention or to determine the safety status of the vehicle's environment. Step S40: Generate and execute the corresponding vehicle control command according to the operation intention or the environmental safety status.

2. The UWB usage method according to claim 1, characterized in that: The environmental perception step in step S10 also includes synchronously acquiring auxiliary information from non-UWB perception modules. The auxiliary information includes at least: the connection status and approximate distance information of authorized mobile terminal devices detected by the Bluetooth module, and the vehicle's own motion state information acquired by one or more of the vehicle's built-in inertial measurement unit, wheel speed sensor, and vibration sensor. The intent and state parsing step in step S30 is performed based on the fusion processing result of the original signal data and the auxiliary information.

3. The UWB usage method according to claim 2, characterized in that: The UWB detection modes in step S20 include gesture trajectory mode, physiological micro-motion mode and near field sensing mode. The activation conditions for the gesture trajectory or near-field sensing mode are both that an authorized mobile terminal is detected to have entered the first preset distance and that the vehicle can be unlocked. In gesture trajectory mode, step S30 matches the user's hand trajectory captured by UWB with the pre-stored unlock template. If the matching degree meets the standard and the terminal is located at the second preset distance, the unlocking intention is determined to be established. In near-field sensing mode, step S30 determines whether the user coordinates calculated by UWB are located in the preset area in front of the seat cushion. If so and the terminal is connected, the unlocking intention is determined to be valid. Step S40 controls the vehicle to unlock according to the unlocking intention.

4. The UWB usage method according to claim 2, characterized in that: Another triggering condition for the activation of the gesture trajectory mode in step S20 is that the vehicle central control system enters a specific interactive enable state. In gesture trajectory mode, the user's operation intention command parsed in step S30 is the control intention of the vehicle device. The parsing process includes: recognizing the user's head nodding / shaking motion or the sliding trajectory of the hand in a specific direction captured by the UWB antenna array, and mapping it to the confirmation / rejection of the vehicle voice command or the switching command of the instrument interface option; step S40 then controls the vehicle audio, instrument or corresponding functional module to perform the corresponding operation.

5. The UWB usage method according to claim 2, characterized in that: Another triggering condition for the activation of the gesture trajectory mode in step S20 is that the vehicle enters the vehicle search assistance state, which is triggered by an authorized mobile terminal device entering a third preset distance range that is farther than the first preset distance. In the gesture trajectory mode, the user's operation intention command parsed in step S30 is the vehicle search trigger intention. The parsing process includes: recognizing the preset vehicle search gesture action captured by the UWB antenna array. Step S40 controls the vehicle's audio-visual system to issue a location alert signal.

6. The UWB usage method according to claim 2, characterized in that: The trigger condition for activating the physiological micro-motion mode in step S20 is that the vehicle is in a protected state and no authorized mobile terminal device is detected within the preset safety range. In the physiological micro-motion mode, the environmental safety status assessment in step S30 includes the determination of the presence of living beings. The process is as follows: perform time-frequency analysis on the reflected signal received by the UWB antenna array to detect whether there are periodic micro-motion signal features representing breathing or heartbeat in the frequency range of 0.1Hz to 0.5Hz, thereby determining whether there are unauthorized living beings around the vehicle.

7. The UWB usage method according to claim 6, characterized in that: The environmental safety status assessment in step S30 also includes comprehensive risk assessment and classification. The process is as follows: based on the determination of the presence of a living body, the vibration duration and intensity information from the vehicle vibration sensor, the vehicle position change information from the displacement sensor, and the vehicle tilting status information from the attitude sensor are integrated, and the risk is accumulated and calculated according to the preset risk factor scoring model. The risk is divided into multiple levels based on the total accumulated score.

8. The method of using UWB according to claim 7, characterized in that: The control execution step in step S40 specifically includes a graded response mechanism: triggering composite protection actions that are progressively enhanced from low to high according to different risk levels.

9. The UWB usage method according to claim 1, characterized in that: In step S30, before recognizing the motion trajectory, there is also a step of detecting and repairing outliers in the gesture trajectory data. Specifically, this includes: based on independent coordinate sequences from multiple UWB antennas, suspicious outliers are initially marked through local statistical feature analysis, kinematic continuity test and signal quality assessment. The detection results from multiple antennas are spatiotemporally aligned and cross-validated to identify real outliers and assign confidence labels. Real outliers are repaired by interpolation using consistent data from multiple antennas or adjacent normal points, and then clean 3D gesture trajectory data is obtained through physical constraint optimization and adaptive filtering for subsequent intent recognition.

10. The UWB usage method according to claim 9, characterized in that: The process of intention recognition based on the clean 3D gesture trajectory data specifically includes: A confidence-weighted dynamic time warping algorithm is used to match the trajectory to be identified with a pre-stored standard gesture template trajectory. Each frame coordinate of the trajectory is assigned a weight obtained by fusing signal quality and repair confidence, and low-confidence continuous regions are allowed to be skipped in the matching path. The system integrates multi-scale matching results and performs secondary verification using Bluetooth proximity context information. Finally, it outputs the recognized gesture command through threshold decision.

Images