Vehicle Bluetooth key positioning method and system
By employing hierarchical filtering and adaptive learning mechanisms, combined with recurrent neural networks, the problem of inaccurate positioning of vehicle Bluetooth keys in complex environments was solved, achieving stable and accurate positioning results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-03-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The vehicle Bluetooth key's positioning is inaccurate in complex environments, affected by multiple interferences and instantaneous jumps, resulting in unstable positioning and inability to function properly.
A hierarchical filtering method is adopted, including a first RSSI filter to correct instantaneous jumps and a second RSSI filter to filter out environmental noise. Combined with an adaptive learning mechanism and a recurrent neural network, accurate positioning is achieved through a radio propagation model and a positioning algorithm.
It significantly improves the accuracy and stability of Bluetooth key positioning, reduces the impact of environmental noise and non-line-of-sight errors, and enables the system to dynamically adapt to complex signal environments, reduce misjudgments, and generate smooth positioning trajectories.
Smart Images
Figure CN121751082A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle communication technology, and in particular to a vehicle Bluetooth key positioning method and system. Background Technology
[0002] The vehicle Bluetooth key is a key component for enabling features such as keyless entry, automatic welcome, and precise vehicle location. The accuracy and stability of the positioning determines the user experience. The positioning of the vehicle Bluetooth key is determined by analyzing the RSSI sampling information received from the key to give the relative position of the key and the vehicle.
[0003] However, in complex real-world situations (underground parking garages, crowded environments, and rapid user movement), vehicle Bluetooth keys may experience inaccurate or unreliable positioning, or fail to function properly. This is because the RSSI signal is susceptible to interference from various signals, and a single filter cannot effectively filter out interference at different levels. The RSSI data of the Bluetooth key is the input condition for accurate location estimation. When a user takes the key out of their pocket, walks through a crowd, or passes by a metal plate, the RSSI may suddenly fluctuate. In addition, interference from Bluetooth signals from computers, mobile phones, and smartwatches can also cause the RSSI to continuously jitter.
[0004] Existing technologies typically perform single-level filtering on RSSI, rather than multi-level filtering. If the filtering bandwidth is too large, RSSI transitions cannot be filtered out; if the filtering bandwidth is too small, the user's movement response will be filtered out, leading to position drift and causing the user to search frantically for their car in an underground parking garage.
[0005] Therefore, a vehicle Bluetooth positioning method is needed that can eliminate multiple interferences between RSSIs in layers and adaptively correct instantaneous jumps to achieve accurate positioning of the vehicle Bluetooth key. Summary of the Invention
[0006] This invention provides a vehicle Bluetooth key positioning method and system, which uses layered processing of RSSI dual interference and adaptive correction of instantaneous jumps to achieve accurate positioning of the vehicle Bluetooth key. The first aspect of this invention provides a vehicle Bluetooth key positioning method, the method comprising: Step S1: Configure a Bluetooth signal controller and multiple Bluetooth key antennas on the vehicle. The Bluetooth signal controller acts as a slave device, pairs with the Bluetooth key, and communicates with the distributed Bluetooth key antennas. Step S2: The Bluetooth key antenna acquires the RSSI value of the Bluetooth key in real time in listening mode and reports it to the Bluetooth signal controller; Step S3: Perform layered filtering on the RSSI values. First, use the first RSSI filter to correct instantaneous jumps, and then use the second RSSI filter to filter out environmental noise. Step S4: Based on the filtered RSSI value, calculate the distance using the radio propagation model, and apply a positioning algorithm to determine the location of the Bluetooth key; Step S5: Determine that the Bluetooth key is in the preset area based on the calculated Bluetooth key position, and trigger vehicle function control.
[0007] As a further aspect of the present invention, the correction of the first RSSI filter in step S3 includes: calculating the instantaneous difference value of the RSSI value, that is, the difference between the current value and the value at the previous moment, and comparing the instantaneous difference value with a dynamically configurable reference threshold. When the absolute value of the instantaneous difference exceeds the reference threshold, it is determined to be an RSSI mutation and classified according to the sign of the second-order difference: if the second-order difference is negative and the instantaneous difference is negative, it is classified as a decay mutation; if the second-order difference is positive and the instantaneous difference is positive, it is classified as an enhancement mutation.
[0008] As a further aspect of the present invention, when RSSI undergoes a sudden change, it is corrected to limit it to a preset reasonable range of variation. The correction process employs an adaptive learning mechanism, which dynamically optimizes the reference threshold through historical signal pattern analysis. The specific implementation of the adaptive learning mechanism includes: using a sliding window technique to collect RSSI sequences from the most recent 100 time points, and calculating the mean, variance, and mutation frequency of the sequences in real time. When the variance suddenly increases beyond the safe range, a threshold recalibration is triggered. The recalibration uses a gradient descent algorithm to minimize the signal transition error.
[0009] As a further aspect of the present invention, the dynamic correction of the first RSSI filter further includes: real-time detection based on signal change patterns, and derivation of correction values through time series analysis; For attenuation abrupt changes, the correction value is derived using an exponential smoothing model. The mean RSSI of the most recent 10 time points is used as the benchmark, and a smoothing factor is applied for weighted averaging. The smoothing factor is dynamically adjusted according to the magnitude of the abrupt change. For enhanced mutations, the amplitude limiting filter principle is used to limit the current RSSI value to within the historical maximum value.
[0010] As a further aspect of the present invention, in step S3, the second RSSI filter is an unscented Kalman filter, whose process noise covariance and observation noise covariance are dynamically optimized through an online learning algorithm. The optimization process includes: The process noise covariance is estimated based on the state prediction error and weighted by the forgetting factor mechanism, which is used to weight the historical and current values; the observation noise covariance is calculated by the observation residuals and weighted by the smoothing factor mechanism, wherein the observation residuals are the difference between the actual RSSI value and the predicted value; When multiple parameters are adjusted collaboratively during the optimization process, the trace ratio of the two covariance matrices is constrained to be within a preset range. If the ratio exceeds the limit, re-initialization is triggered.
[0011] As a further aspect of the present invention, the positioning algorithm integrates a probabilistic model and a geometric model. The probabilistic model uses a Naive Bayes classifier, and the geometric model is based on the trilateration principle and introduces an error correction factor to handle non-line-of-sight errors.
[0012] As a further aspect of the present invention, the hierarchical filtering process also includes a post-verification step, which verifies the effect by comparing the statistical characteristics of the signals before and after filtering, and performs three levels of verification: the first level calculates the variance ratio before and after filtering, and if it is lower than the threshold, the second level is triggered; the second level uses a Gaussian mixture model to calculate the KL divergence between the filtered distribution and the ideal distribution; the third level calculates the mean absolute error and the root mean square error.
[0013] As a further aspect of the present invention, the online learning algorithm adopts a recurrent neural network structure, with the current and historical RSSI sequences as input and the optimized noise covariance parameter as output. The network structure includes an input layer, hidden layers, and a linear regression output layer, and is trained using real vehicle multi-scenario data; Overfitting is controlled during the training process using five-fold cross-validation and regularization.
[0014] As a further aspect of the present invention, the recurrent neural network includes a gated recurrent unit for capturing the time dependence of RSSI values; The gating loop unit controls the information flow through update gates and reset gates. The update gate adjusts the degree of retention of historical information, and the reset gate adjusts the degree of influence of the current input. When a transient disturbance occurs in the signal, the reset threshold is lowered to focus on the current change; When the signal stabilizes, the update threshold is increased to reinforce the memory.
[0015] Another aspect of the present invention provides a vehicle Bluetooth key positioning system, comprising: The Bluetooth signal controller includes a layered processing module, a distance calculation module, and a positioning calculation module, which are used to filter and correct the received RSSI values. The signal acquisition module is used to acquire the RSSI value of the Bluetooth key in real time via the Bluetooth key antenna; Bluetooth key antennas are distributed on the B-pillars and inside the bumper of the vehicle, providing omnidirectional signal coverage. The function triggering module is used to determine whether the Bluetooth key is in a preset area based on the calculated Bluetooth key location and to trigger vehicle function control. The layered processing module is used to perform layered processing on the RSSI value, first correcting instantaneous jumps through the first filtering unit, and then filtering out environmental noise through the second filtering unit; the distance calculation module is used to calculate the distance between the Bluetooth key and the vehicle based on the filtered RSSI value using a radio propagation model; and the positioning calculation module is used to calculate the position of the Bluetooth key by combining a positioning algorithm.
[0016] The beneficial effects of this invention are as follows: This invention significantly reduces the impact of environmental noise and non-line-of-sight errors on positioning results through hierarchical filtering and fusion algorithms, thereby improving positioning accuracy. The adaptive learning mechanism and recurrent neural network enable the system to dynamically adapt to complex signal environments. A three-level verification process ensures the reliability of the filtering effect, resulting in a final RSSI that is both accurate and highly stable. Furthermore, because the output of the probabilistic model is inherently continuous, it is fused with the discrete coordinates calculated by the geometric model to generate a smoother, less abrupt positioning trajectory, thus improving the accuracy and stability of Bluetooth key positioning. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of the vehicle Bluetooth key positioning method in an embodiment of the present invention. Figure 2 This is a flowchart of RSSI filtering for a vehicle Bluetooth key in an embodiment of the present invention; Figure 3 This is a flowchart of the Bluetooth key positioning method based on filtered RSSI in an embodiment of the present invention; Figure 4 This is a layout diagram of the vehicle Bluetooth signal controller and antenna in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the communication connection between the Bluetooth key, the Bluetooth signal controller, and the antenna in an embodiment of the present invention. Figure 6 This is a diagram illustrating the RSSI data processing effect in an embodiment of the present invention. Figure 7 This is a connection diagram of the vehicle Bluetooth key positioning system module in an embodiment of the present invention.
[0019] The attached figures are labeled as follows: 1. Bluetooth signal controller; 2. Bluetooth key antenna; 3. Signal acquisition module; 4. Layered processing module; 5. Distance calculation module; 6. Positioning calculation module; 7. Function triggering module. Detailed Implementation
[0020] This invention provides a vehicle Bluetooth key positioning method and system. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.
[0021] like Figure 1 As shown, this embodiment of the invention provides a vehicle Bluetooth key positioning method, including: Step S1: A Bluetooth signal controller 1 and multiple Bluetooth key antennas 2 are configured on the vehicle. The Bluetooth signal controller 1 acts as a slave device, pairing with the Bluetooth key, and simultaneously communicates with the distributed Bluetooth key antennas 2 via LIN; Figure 4 As shown, multiple Bluetooth key antennas 2 are distributed in different locations of the vehicle, including but not limited to the front, rear and side doors. The Bluetooth signal controller 1 is located on the center console and connected to it. This arrangement ensures maximum signal coverage, thereby improving positioning accuracy.
[0022] Step S2: Bluetooth key antenna 2 collects the RSSI value of Bluetooth key in real time in listening mode and uploads the collected data to Bluetooth signal controller 1. Bluetooth signal controller 1 maintains data interaction with Bluetooth key antenna 2 through communication protocols (such as UART or SPI) and acts as a slave device to Bluetooth key. Bluetooth key can be a user's mobile phone or a key configured with a Bluetooth module. Bluetooth signal controller 1 transmits the received RSSI value to hierarchical processing module 4 for RSSI filtering.
[0023] Step S3: As Figure 2 As shown, the RSSI value is subjected to hierarchical filtering. First, the RSSI value is corrected for instantaneous jumps by the first filtering unit. The correction process of the first filtering unit includes calculating the instantaneous difference value of the RSSI value. That is, the RSSI value at the current time. RSSI value compared to the previous time step The differences are as follows: , By comparing the instantaneous difference value with a dynamically adjusted reference threshold, an RSSI mutation is identified when the absolute value of the instantaneous difference value exceeds the reference threshold, requiring further analysis of the second-order difference value. The mutation type is determined, and then, based on real-time detection of signal mutation patterns, correction values are derived through time series analysis, where: , If both the second-order difference and the instantaneous difference are negative, it is classified as a decaying abrupt change; if both are positive, it is classified as an enhancing abrupt change. For decaying abrupt changes, an exponential smoothing model is used to derive the correction value, with the mean RSSI of the most recent 10 time points as the benchmark, and a smoothing factor is applied. (The default value is 0.2) A weighted average is calculated, and the smoothing factor is based on the magnitude of the mutation. Dynamic adjustment: , The magnitude of the mutation is obtained by comparing the current data point with the baseline value. The magnitude of the mutation = |current data point value - baseline value|, where the baseline value needs to represent the normal or expected level of the signal.
[0024] For enhanced mutations, an amplitude-limiting filter principle is used to adjust the current RSSI value. Limit to historical maximum Within, that is , Furthermore, when RSSI exhibits abrupt changes, an adaptive learning mechanism is used to limit it to a pre-defined reasonable range of variation. This adaptive learning mechanism employs a sliding window technique to collect RSSI sequences from the most recent 100 time points and calculates the mean of the sequences in real time. ,variance and mutation frequency When the variance suddenly increases beyond the safe range, a threshold recalibration is triggered. The recalibration uses a gradient descent algorithm to minimize the signal jump error. The optimization objective function is: , in The threshold parameter to be optimized. For window size, This is the actual value. These are predicted values.
[0025] After correction by the first filtering unit, the RSSI value enters the second filtering unit for environmental noise filtering. The second filtering unit uses an unscented Kalman filter, and its process noise covariance is... and observation noise covariance The optimization process utilizes an online learning algorithm for dynamic optimization. This optimization includes process noise covariance optimization based on state prediction error estimation. When multiple parameters are adjusted collaboratively during optimization, the trace ratio of the two covariance matrices is constrained to a preset range. If the ratio exceeds the limit, reinitialization is triggered. The online learning algorithm employs a recurrent neural network (RNN) structure. The input is the current and historical RSSI sequences (length L=10), and the output is the optimized noise covariance parameters. The network structure includes an input layer, hidden layers, and a linear regression output layer. Training uses real-world multi-scenario data. The RNN includes a gated recurrent unit (GRU) to capture the temporal dependence of RSSI values. The GRU internally updates the gate... ) and reset door ( The system controls the information flow, updates the gate to adjust the retention level of historical information, and resets the gate to adjust the impact of the current input. When a transient disturbance occurs in the signal, the reset gate is lowered to focus on the current change; when the signal stabilizes, the update gate is raised to reinforce the memory.
[0026] The tiered processing also includes a post-validation stage, which verifies the effectiveness by comparing the statistical characteristics of the signals before and after filtering. Three levels of validation are performed: the first level calculates the variance ratio before and after filtering; if it is below a threshold, the second level is triggered, using a Gaussian mixture model (GMM) to calculate the KL divergence between the filtered distribution and the ideal distribution; if the KL divergence exceeds a threshold, the third level is triggered, calculating the mean absolute error (MAE) and root mean square error (RMSE). , , in, This is the RSSI after hierarchical filtering. This is the true RSSI obtained through calibration at a known distance.
[0027] like Figure 6 As shown in the graph of RSSI data processing, the red line represents the original RSSI, which exhibits significant fluctuations, while the black line represents the processed RSSI, which is smoother and more stable. This embodiment significantly reduces the impact of environmental noise and non-line-of-sight errors on the positioning results through hierarchical filtering and fusion algorithms, thereby improving positioning accuracy. The adaptive learning mechanism and recurrent neural network enable the system to dynamically adapt to complex signal environments, enhancing the system's robustness. The three-level verification design further ensures the reliability of the filtering effect and reduces the possibility of misjudgment, resulting in a final processed RSSI that is both accurate and highly stable.
[0028] Step S4: As Figure 3 As shown, based on the filtered RSSI value, the distance is calculated using a radio propagation model, and a positioning algorithm is applied to determine the location of the Bluetooth key. The radio propagation model employs a logarithmic distance path loss model. , in, To keep the Bluetooth key at a distance The RSSI value received at the location, For reference distance The receiving power at a distance of (typically 1 meter) This is the path loss index (typically 2.0-4.0 in a vehicle environment). A Gaussian random variable with a mean of 0 and a standard deviation of 0. (Reflects environmental noise).
[0029] This model considers the signal strength attenuation with distance and takes into account the influence of environmental factors such as obstacles and reflectors on the signal, introducing an error correction factor during the calculation process. To handle non-line-of-sight errors and ensure the accuracy of distance calculations, the final corrected distance is: , The error correction factor is obtained through actual vehicle calibration, specifically as follows: , in For the number of calibration points, These are measured values. These are the model's predicted values.
[0030] Subsequently, the precise location of the Bluetooth key is calculated by the location solution module 6, which integrates a probabilistic model and a geometric model. The probabilistic model employs a Naive Bayes classifier to initially determine the region to which the Bluetooth key belongs (e.g., inside the vehicle, outside the vehicle, near the driver's side, etc.). Let the region set be... , The feature vector is a vector composed of RSSI values obtained from M antennas: , The Bluetooth key belongs to the area. The posterior probability is: , in For prior probability, Let RSSI be the likelihood function. Assuming that each RSSI value is conditionally independent, then: .
[0031] The geometric model is based on the trilateration principle. It uses the RSSI values collected by multiple Bluetooth key antennas 2 to calculate the distance between the Bluetooth key and each Bluetooth key antenna 2, thereby determining the specific location of the Bluetooth key. Taking three antennas as an example, let their coordinates be (…). () () The measured distances were respectively Then the Bluetooth key coordinates ( This can be obtained by solving the following system of equations: , Due to measurement errors, the system of equations may not have an exact solution. In this case, the least squares method or centroid algorithm is used to find the optimal solution, thereby determining the precise location of the Bluetooth key.
[0032] In this embodiment, the initial range is first defined by a probabilistic model, and one or several regions with the highest probability are selected as the priority search space for the next geometric solution. This can effectively avoid the geometric model from obtaining an obviously unreasonable position due to signal fluctuations when solving in the global range. Then, the geometric model is used for precise positioning. Since the probabilistic model has eliminated a large number of interference regions, the calculation of the geometric model is more focused, and its results are more reliable.
[0033] Meanwhile, the position information calculated by the geometric model can also feed back to the probabilistic model. If the coordinates calculated by the geometry remain stable in the "external unlocking zone" for a period of time, but the probabilistic model gives a high probability of "inside the vehicle", it will trigger a re-evaluation or adaptive adjustment, making the judgment of the probabilistic model more consistent with the current geometric solution. This solves the problem that the geometric positioning based solely on RSSI in complex signal scenarios will produce large errors due to multipath effects. Moreover, the probabilistic model is based on statistical laws and is not easily affected by the jump of a single signal. Furthermore, by narrowing the solution range in advance through the probabilistic model, it avoids the geometric model from performing a large number of unnecessary calculations globally, making it particularly suitable for devices with limited computing power, such as automotive MCUs.
[0034] Meanwhile, since the output of the probability model is continuous, it can be fused and filtered with the discrete coordinates calculated by the geometric model to generate a smoother and less abrupt positioning trajectory, effectively preventing the door locks or lights from frequently opening and closing due to small fluctuations in position estimation.
[0035] Step S5: The function trigger module 7 determines whether the Bluetooth key is in a preset area (such as the welcome area, unlocking area, in-vehicle area, non-in-vehicle area, etc.) based on the calculated Bluetooth key position and triggers the corresponding vehicle function control. For example, when the Bluetooth key is in the welcome area (usually 3-5 meters around the vehicle), the welcome light is triggered to light up; when it is in the unlocking area (usually 1-2 meters), the door is triggered to unlock; when it is in the in-vehicle area, the engine can be started. The function trigger module 7 sends commands to the vehicle control system (such as the body control module BCM, powertrain control module PCM) through the vehicle CAN bus or LIN bus to execute the corresponding function.
[0036] In practical applications, when a user approaches a vehicle with a Bluetooth key (such as a smartphone or smart key fob), the Bluetooth key antenna 2 collects RSSI values in real time and uploads them to the Bluetooth signal controller 1. The Bluetooth signal controller 1 transmits the RSSI values to the hierarchical processing module 4 for instantaneous jump correction and environmental noise filtering. The processed RSSI values are used to calculate the distance between the Bluetooth key and the vehicle, and combined with the positioning algorithm to calculate the specific location of the Bluetooth key. When the Bluetooth key enters the preset area, the function triggering module 7 automatically triggers vehicle function control, such as unlocking the door or starting the engine. The entire process does not require manual operation by the user, improving the user experience. In addition, the system continuously optimizes the filtering effect and positioning accuracy through an adaptive learning mechanism and online learning algorithm to adapt to the complex scenario requirements in different environments.
[0037] This invention also provides a vehicle Bluetooth key positioning system to implement the above-mentioned Bluetooth key positioning method, such as... Figure 7 As shown, the system includes: Bluetooth signal controller 1 includes a layered processing module 4, a distance calculation module 5, and a positioning calculation module 6, which are used to filter and correct the received RSSI values; Signal acquisition module 3 is used to acquire the RSSI value of the Bluetooth key in real time via the Bluetooth key antenna; Bluetooth key antenna 2 is distributed in the B-pillar and inside the bumper of the vehicle, providing omnidirectional signal coverage; The function trigger module 7 is used to determine whether the Bluetooth key is in a preset area based on the calculated Bluetooth key position and to trigger vehicle function control.
[0038] The layered processing module 4 is used to perform layered processing on the RSSI value. First, the instantaneous jump is corrected by the first filtering unit, and then the environmental noise is filtered out by the second filtering unit. The distance calculation module 5 is used to calculate the distance between the Bluetooth key and the vehicle based on the filtered RSSI value and the radio propagation model. The positioning calculation module 6 is used to calculate the position of the Bluetooth key by combining the positioning algorithm.
[0039] like Figure 5As shown, the signal acquisition module 3 is connected to multiple Bluetooth key antennas 2. The number of Bluetooth key antennas 2 is M and M is greater than or equal to 3. These antennas are distributed at different locations in the vehicle and establish a data interaction channel with the Bluetooth signal controller 1 through a communication protocol. The Bluetooth signal controller 1 acts as a slave device and pairs with the Bluetooth key, and transmits the acquired RSSI value to the hierarchical processing module 4. The hierarchical processing module 4 includes a first filtering unit and a second filtering unit, which are used to correct instantaneous jumps and filter out environmental noise, respectively. The RSSI value after hierarchical processing is transmitted to the distance calculation module 5. This module calculates the distance between the Bluetooth key and the vehicle based on the radio propagation model. Subsequently, the positioning solution module 6 combines the probability model and the geometric model to solve the specific position of the Bluetooth key. Finally, the function triggering module 7 determines whether the Bluetooth key is in the preset area based on the calculated position and triggers the corresponding vehicle function control.
[0040] The RSSI value is instantaneously abruptly corrected by the first filtering unit. The correction process of the first filtering unit includes calculating the instantaneous difference of the RSSI value, that is, the difference between the RSSI value at the current time and the RSSI value at the previous time, and comparing the instantaneous difference with a dynamically adjusted reference threshold. When the absolute value of the instantaneous difference exceeds the reference threshold, it is determined to be an RSSI abrupt change. At this time, it is necessary to further analyze the sign of the second-order difference to classify the type of abrupt change: if the second-order difference is negative and the instantaneous difference is negative, it is classified as a decaying abrupt change; if the second-order difference is positive and the instantaneous difference is positive, it is classified as an enhancing abrupt change. For decaying abrupt changes, the correction value is derived using an exponential smoothing model. The average RSSI value at the most recent 10 time points is used as the benchmark, and a smoothing factor is applied for weighted averaging. The smoothing factor is dynamically adjusted according to the magnitude of the abrupt change. For enhancing abrupt changes, the amplitude limiting filtering principle is used to limit the current RSSI value to within the historical maximum value.
[0041] Furthermore, when RSSI undergoes a sudden change, an adaptive learning mechanism is used to limit it to a preset reasonable range of variation. The adaptive learning mechanism uses a sliding window technique to collect RSSI sequences from the most recent 100 time points, and calculates the mean, variance, and frequency of change of the sequences in real time. When the variance suddenly increases beyond the safe range, a threshold recalibration is triggered. The recalibration uses a gradient descent algorithm to minimize the signal jump error.
[0042] After correction by the first filtering unit, the RSSI value enters the second filtering unit for environmental noise filtering. The second filtering unit uses an unscented Kalman filter, and its process noise covariance and observation noise covariance are dynamically optimized through an online learning algorithm. The optimization process includes process noise covariance optimization based on state prediction error estimation, using a forgetting factor mechanism to weight historical and current values; the observation noise covariance is calculated through observation residuals, using a smoothing factor mechanism for weighting, where the observation residual is the difference between the actual RSSI value and the predicted value. When multiple parameters are adjusted collaboratively during the optimization process, the trace ratio of the two covariance matrices is constrained to within a preset range. If the ratio exceeds the limit, re-initialization is triggered. The learning algorithm employs a recurrent neural network structure. The input consists of current and historical RSSI sequences, and the output is the optimized noise covariance parameter. The network structure includes an input layer, hidden layers, and a linear regression output layer. Training is conducted using real-vehicle multi-scenario data. The training process uses five-fold cross-validation and regularization to control overfitting. The recurrent neural network contains a gated recurrent unit to capture the time dependence of RSSI values. Inside the gated recurrent unit, information flow is controlled through update and reset gates. The update gate adjusts the degree of retention of historical information, while the reset gate adjusts the degree of influence of the current input. When a transient disturbance occurs in the signal, the reset gate is lowered to focus on the current change; when the signal is stable, the update gate is raised to strengthen memory.
[0043] After processing by the hierarchical processing module 4, the RSSI value is transmitted to the distance calculation module 5. The distance calculation module 5 calculates the distance between the Bluetooth key and the vehicle based on the radio propagation model. The radio propagation model considers the signal strength attenuation with distance and takes into account the influence of environmental factors such as obstacles and reflective surfaces on the signal. An error correction factor is introduced during the calculation process to handle non-line-of-sight errors and ensure the accuracy of the distance calculation. Subsequently, the positioning solution module 6 combines a probabilistic model and a geometric model to solve the specific location of the Bluetooth key. The probabilistic model uses a Naive Bayes classifier, and the geometric model is based on the trilateration principle. The trilateration principle uses the RSSI values collected by multiple Bluetooth key antennas 2 to calculate the distance between the Bluetooth key and each antenna, thereby determining the specific location of the Bluetooth key. The hierarchical processing also includes a post-verification stage, which verifies the effect by comparing the statistical characteristics of the signals before and after filtering. Three levels of verification are performed: the first level calculates the variance ratio before and after filtering, and if it is lower than the threshold, the second level is triggered; the second level uses a Gaussian mixture model to calculate the KL divergence between the filtered distribution and the ideal distribution; the third level calculates the mean absolute error and the root mean square error.
[0044] Finally, the function triggering module 7 determines whether the Bluetooth key is in the preset area based on the calculated Bluetooth key position and triggers the corresponding vehicle function control. The preset area can be a specific range around the vehicle, such as the door unlocking area or the start area. When the Bluetooth key is in the preset area, the function triggering module 7 sends a command to the vehicle control system to execute the corresponding function, such as unlocking the door or starting the engine. The entire system operation is coordinated by the Bluetooth signal controller 1 to ensure smooth data interaction and logical connection between the modules. The Bluetooth signal controller 1 is also responsible for monitoring the system operation status and triggering an alarm or restart mechanism in case of abnormality.
[0045] In the above process, the distributed arrangement of the Bluetooth key antenna 2 is as follows: Figure 3 As shown, multiple Bluetooth key antennas 2 are distributed in different locations on the vehicle, such as the front, rear, and near the side doors. This arrangement ensures maximum signal coverage, thereby improving positioning accuracy. The Bluetooth key antennas 2 establish a data interaction channel with the Bluetooth signal controller 1 through a communication protocol to ensure stable transmission of RSSI values. The Bluetooth signal controller 1 acts as a slave device, paired with the Bluetooth key, and transmits the collected RSSI values to the hierarchical processing module 4.
[0046] Through the above steps, the system can effectively address RSSI signal fluctuations in complex scenarios. For example, in underground parking garages or densely built-up areas, sudden changes in RSSI values caused by signal reflection and obstruction can be suppressed by the instantaneous jump correction of the first filtering unit. In densely populated areas or where interference from other Bluetooth devices is present, the environmental noise filtering of the second filtering unit can significantly improve signal stability. Furthermore, through adaptive learning mechanisms and online learning algorithms, the system can continuously optimize filtering effects and positioning accuracy to adapt to the needs of complex scenarios in different environments.
[0047] In summary, this invention significantly reduces the impact of environmental noise and non-line-of-sight errors on positioning results through hierarchical filtering and fusion algorithms, thereby improving positioning accuracy. The adaptive learning mechanism and recurrent neural network enable the system to dynamically adapt to complex signal environments. A three-level verification process ensures the reliability of the filtering effect and reduces the possibility of misjudgment, resulting in a final RSSI that is both accurate and highly stable. Furthermore, because the output of the probabilistic model is inherently continuous, it can be fused and filtered with the discrete coordinates calculated by the geometric model, generating a smoother, less abrupt positioning trajectory. This effectively prevents frequent opening and closing of door locks or headlights due to minor fluctuations in position estimation.
[0048] All content not described in detail in this specification is prior art known to those skilled in the art, and the model parameters of each electrical appliance are not specifically limited; conventional equipment can be used. Electrical control components not mentioned in this technical solution are not shown in the figures because they are prior art, and will not be described further here.
[0049] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for locating a vehicle via Bluetooth key, characterized in that, The method includes: Step S1: Configure a Bluetooth signal controller and multiple Bluetooth key antennas on the vehicle. The Bluetooth signal controller acts as a slave device, pairs with the Bluetooth key, and communicates with the distributed Bluetooth key antennas. Step S2: The Bluetooth key antenna acquires the RSSI value of the Bluetooth key in real time in listening mode and reports it to the Bluetooth signal controller; Step S3: Perform layered filtering on the RSSI values. First, use the first RSSI filter to correct instantaneous jumps, and then use the second RSSI filter to filter out environmental noise. Step S4: Based on the filtered RSSI value, calculate the distance using the radio propagation model, and apply a positioning algorithm to determine the location of the Bluetooth key; Step S5: Determine that the Bluetooth key is in the preset area based on the calculated Bluetooth key position, and trigger vehicle function control.
2. The vehicle Bluetooth key positioning method according to claim 1, characterized in that, The correction of the first RSSI filter in step S3 includes: Calculate the instantaneous difference of the RSSI value, that is, the difference between the current value and the value at the previous moment, and compare the instantaneous difference with a dynamically configurable reference threshold; When the absolute value of the instantaneous difference exceeds the reference threshold, it is determined to be an RSSI mutation and classified according to the sign of the second-order difference: if the second-order difference is negative and the instantaneous difference is negative, it is classified as a decay mutation; if the second-order difference is positive and the instantaneous difference is positive, it is classified as an enhancement mutation.
3. The vehicle Bluetooth key positioning method according to claim 2, characterized in that, When RSSI changes abruptly, correction is used to limit it to a preset reasonable range of variation. The correction process adopts an adaptive learning mechanism and dynamically optimizes the reference threshold through historical signal pattern analysis. The adaptive learning mechanism includes: using a sliding window technique to collect RSSI sequences from the most recent 100 time points, and calculating the mean, variance, and mutation frequency of the sequences in real time; When the variance suddenly increases beyond the safe range, a threshold recalibration is triggered. The recalibration uses a gradient descent algorithm to minimize the signal transition error.
4. The vehicle Bluetooth key positioning method according to claim 3, characterized in that, The dynamic correction of the first RSSI filter further includes: Real-time detection based on signal mutation patterns, and derivation of correction values through time series analysis; For attenuation abrupt changes, the correction value is derived using an exponential smoothing model. The mean RSSI of the most recent 10 time points is used as the benchmark, and a smoothing factor is applied for weighted averaging. The smoothing factor is dynamically adjusted according to the magnitude of the abrupt change. For enhanced mutations, the amplitude limiting filter principle is used to limit the current RSSI value to within the historical maximum value.
5. The vehicle Bluetooth key positioning method according to claim 1, characterized in that, In step S3, the second RSSI filter is an unscented Kalman filter, and its process noise covariance and observation noise covariance are dynamically optimized through an online learning algorithm. The optimization process includes: The process noise covariance is estimated based on the state prediction error and weighted by the forgetting factor mechanism, which is used to weight the historical and current values; the observation noise covariance is calculated by the observation residuals and weighted by the smoothing factor mechanism, wherein the observation residuals are the difference between the actual RSSI value and the predicted value; When multiple parameters are adjusted collaboratively during the optimization process, the trace ratio of the two covariance matrices is constrained to be within a preset range. If the ratio exceeds the limit, re-initialization is triggered.
6. The vehicle Bluetooth key positioning method according to claim 1, characterized in that, The localization algorithm in step S4 integrates a probabilistic model and a geometric model. The probabilistic model uses a Naive Bayes classifier, while the geometric model is based on the trilateration principle and introduces an error correction factor to handle non-line-of-sight errors.
7. The vehicle Bluetooth key positioning method according to claim 1, characterized in that, The hierarchical filtering process also includes a post-verification step, which verifies the effect by comparing the statistical characteristics of the signals before and after filtering, and performs three levels of verification: the first level calculates the variance ratio before and after filtering, and if it is lower than the threshold, the second level is triggered; the second level uses a Gaussian mixture model to calculate the KL divergence between the filtered distribution and the ideal distribution; the third level calculates the mean absolute error and the root mean square error.
8. The vehicle Bluetooth key positioning method according to claim 5, characterized in that, The online learning algorithm adopts a recurrent neural network structure, with the current and historical RSSI sequences as input and the optimized noise covariance parameter as output. The network structure includes an input layer, hidden layers, and a linear regression output layer, and is trained using real vehicle multi-scenario data; Overfitting is controlled during the training process using five-fold cross-validation and regularization.
9. The vehicle Bluetooth key positioning method according to claim 8, characterized in that, The recurrent neural network includes gated recurrent units to capture the time dependence of RSSI values; The gating loop unit controls the information flow through update gates and reset gates. The update gate adjusts the degree of retention of historical information, and the reset gate adjusts the degree of influence of the current input. When a transient disturbance occurs in the signal, the reset threshold is lowered to focus on the current change; When the signal stabilizes, the update threshold is increased to reinforce the memory.
10. A vehicle Bluetooth key positioning system, used to implement the vehicle Bluetooth key positioning method according to any one of claims 1 to 9, characterized in that, include: The Bluetooth signal controller includes a layered processing module, a distance calculation module, and a positioning calculation module, which are used to filter and correct the received RSSI values. The signal acquisition module is used to acquire the RSSI value of the Bluetooth key in real time via the Bluetooth key antenna; Bluetooth key antennas are distributed on the B-pillars and inside the bumper of the vehicle, providing omnidirectional signal coverage. The function triggering module is used to determine whether the Bluetooth key is in a preset area based on the calculated Bluetooth key location and to trigger vehicle function control. The layered processing module is used to perform layered processing on the RSSI value, first correcting instantaneous jumps through the first filtering unit, and then filtering out environmental noise through the second filtering unit; the distance calculation module is used to calculate the distance between the Bluetooth key and the vehicle based on the filtered RSSI value using a radio propagation model; and the positioning calculation module is used to calculate the position of the Bluetooth key by combining a positioning algorithm.