Automobile digital key positioning method, direction identification method and computer program product

Abstract

The application discloses a kind of automobile digital key positioning method, direction identification method and computer program product.The positioning method includes first determining the direction area where automobile digital key is located, then according to the direction area determined, select the positioning anchor station that signal propagation path between digital key is reliable to form anchor station subset, and based on the signal strength measurement value of anchor station subset, digital key is positioned calculation.Through the direction statistical model of conditional probability and marginal probability based on RSSI sorting result, direction identification is carried out according to maximum posterior probability or through machine learning model, direction identification is carried out.The application needs only to sort real-time signal strength in line time, and direction identification can be completed according to maximum posterior probability, with small amount of calculation, suitable for real-time operation on vehicle controller;Sorting result and signal strength difference feature are only dependent on the relative relationship of the signal strength of each anchor station, not affected by the difference of key transmitting power, without calibration for terminal one by one, with good generalization ability.

Description

Technical Field

[0001] This invention relates to the field of wireless communication positioning technology, and more specifically, to a car digital key positioning method, a direction recognition method, and a computer program product. Background Technology

[0002] Car digital keys are a technology that uses smartphones, smartwatches, and other devices as car keys. The most basic function of a digital key is to unlock and lock the car doors seamlessly based on the key's distance from the vehicle. When the key approaches the vehicle and gets close enough (e.g., 2 meters), the doors automatically unlock. When the key moves away from the vehicle and gets far enough (e.g., 5 meters), the doors automatically lock. The entire unlocking and locking process requires no intervention from the owner, thus improving the user experience. This process typically relies on near-field communication (NFC) technologies, such as Bluetooth and ultra-wideband (UWB). Utilizing the received signal strength of wireless communication systems for ranging / positioning is a common technique; for example, Apple's iBeacon system is based on Bluetooth for positioning. Besides Bluetooth, wireless communication technologies such as WiFi and Zigbee are also widely used for indoor / outdoor positioning. Furthermore, in recent years, Bluetooth has been increasingly used for car digital key positioning.

[0003] All the above techniques utilize the Received Signal Strength Indicator (RSSI) of the wireless received signal to estimate distance. The method of estimating distance using RSSI is based on the understanding that the greater the distance between the communicating parties, the lower the received signal power. The fundamental understanding of RSSI-based ranging is that the greater the distance between the communicating parties, the lower the signal power received by the receiver. Under the free-space propagation model, path loss increases linearly with the distance between the transceiver. Therefore, the distance between the transceiver can be estimated based on the currently measured RSSI.

[0004] In practical applications, RSSI-based positioning accuracy is not ideal. The relationship between the signal strength received by the receiver antenna and the transmitter power, the distance between the transmitter and receiver, etc., is given by Friis's formula:

[0005] (1)

[0006] Where P t For transmission power, P r For received power, G t For the transmit antenna gain, G r For receiving antenna gain, It is the wavelength of the carrier signal. Equation (1) shows that the received power of the receiver antenna is related not only to the transmit power and the distance d between the transceiver, but also to the antenna gain and the frequency of the carrier signal. The received power given by the free space propagation formula is the result under ideal conditions. In actual systems, the received power is also affected by various factors such as multipath propagation, obstruction, large-scale fading, small-scale fading, surrounding moving targets, thermal noise and interference. The relationship between RSSI and distance is not an ideal monotonic function. Even if the transceiver remains stationary and the distance remains constant, the received signal strength will continue to fluctuate randomly. Especially in the scenario of car digital keys, the metal body of the car body has a strong blocking effect on wireless signals: when the key is located on one side of the vehicle, the anchor station on the same side of the vehicle and the key are line-of-sight propagation, and the RSSI of the key decreases significantly with increasing distance, so the ranging result is relatively reliable; however, the anchor station on the other side of the vehicle is completely blocked by the metal body of the car body, and there is no direct path. The received signal is a superposition of multipath signals such as diffraction and reflection, and the signal is significantly weaker and fluctuates violently. Within a distance of several meters, the RSSI does not even decrease with increasing distance. The same RSSI value may correspond to multiple different distances, which leads to serious ranging ambiguity.

[0007] See Figure 1 The vehicle has a Bluetooth node installed on each of the left and right B-pillars to receive wireless signals from the mobile phone and measure the received power. Figure 2 and Figure 3 Give the following when the phone is along Figure 1 The relationship between the received signal power and distance of the Bluetooth nodes / anchor stations on the left and right sides when the CRRC moves in a straight line on the left side. Figure 1 It can be seen that the Bluetooth node on the left B-pillar communicates with the phone via line-of-sight, while the connection between the Bluetooth node on the right B-pillar and the phone is completely obstructed by the metal body of the car. Figure 2 As can be seen, when the key is on the left side of the car, the signal strength received by the positioning anchor station on the left side of the car shows a clear trend of decreasing signal strength with increasing distance. Data fitting of the test results yields the distance-RSSI curve. The fluctuations in the test results shown in the figure are caused by small-scale fading. Figure 3 It can be seen that within a distance range of 2-7 meters, the relationship between the signal strength received by the anchor station on the right side of the vehicle and the distance does not show a trend of RSSI weakening with increasing distance. For example, when RSSI = -76dBm, corresponding to five different distances within the range of 2.4-5.6 meters, this leads to ambiguity in the ranging results, resulting in uncertainty in the positioning results. According to Figure 1 At this point, the connection between the phone key and the anchor point on the right side of the car is completely blocked by the metal body. Because there is no direct path, the received signal is a superposition of diffracted and reflected signals, which makes the signal significantly weaker and causes violent fluctuations. Figure 4This further demonstrates the relationship between distance and RSSI in scenarios where there is obstruction between the phone key and the anchor station. Figure 4 As the mobile phone key moves along the left rear of the car, the relationship between the strength of the mobile phone's wireless signal and distance is as follows: Figure 5 As shown. By Figure 5 It can be seen that, with Figure 5 Similarly, at this time, RSSI does not show a trend of gradually weakening with increasing distance, and the accurate distance cannot be estimated based on RSSI.

[0008] Therefore, existing positioning methods typically use the measurement results of all anchor stations indiscriminately for position calculation. Unreliable measurement results from obscured anchor stations can significantly reduce positioning accuracy. Some improved methods attempt to identify line-of-sight anchor stations and correct the ranging results of non-line-of-sight anchor stations. However, the identification of line-of-sight versus non-line-of-sight anchor stations depends on the ranging results themselves, and the reliability of identification is difficult to guarantee when the signal fluctuates severely. Some methods use deep learning models to directly process RSSI signals. Although they have strong fitting capabilities, the number of model parameters and computational load are large, which is not friendly to low-computing-power microcontrollers on the vehicle. Furthermore, they require a large amount of data collection and calibration for different types of terminals, resulting in high portability complexity.

[0009] Since vehicle body obstruction can render the measurement results of some anchor stations unusable, if the approximate direction of the digital key relative to the target object can be known during the positioning process, a set of anchor stations with more reliable measurement results can be selected to participate in the position calculation, thereby avoiding the ranging error caused by the obstructed anchor stations. Summary of the Invention

[0010] To address the problems existing in the prior art, the present invention aims to provide a method for locating a car digital key, a method for recognizing direction, and a computer program product, which performs direction recognition of the digital key before locating the digital key, thereby solving the problem of inaccurate direct positioning.

[0011] To achieve the above objectives, in a first aspect, the present invention provides a method for locating a vehicle using a digital key, which divides the horizontal plane where the vehicle is located into a finite number of different directions and deploys multiple positioning anchor stations; the positioning anchor stations are used to receive wireless signals sent by the digital key and measure the received signal strength of the wireless signals; the method includes the following steps:

[0012] First, determine the directional area where the car digital key is located. Then, based on the determined directional area, select a subset of positioning anchor stations from the plurality of positioning anchor stations that have a reliable signal propagation path with the digital key. Based on the signal strength measurement values ​​of the anchor station subset, perform positioning calculation on the digital key to obtain the position information of the digital key relative to the vehicle.

[0013] Secondly, the present invention provides a method for locating a car digital key, used to determine the directional area where the car digital key is located in the car digital key location method described above, comprising the following steps:

[0014] During the offline phase, signal strength measurement samples of the multiple positioning anchor stations receiving the wireless signal of the digital key are collected in each directional area. The signal strength measurement values ​​of each positioning anchor station in each sample are sorted according to the signal strength magnitude to obtain the sorting result corresponding to the sample. The conditional probability of each sorting result appearing in each directional area and the marginal probability of appearing in the entire sample space are calculated. Based on the samples in each directional area and their sorting results, conditional probability and marginal probability, a directional estimation model is established.

[0015] During the online phase, multiple real-time signal strength measurements are obtained from the wireless signals of the digital key received by the multiple positioning anchor stations at the same time. The multiple real-time signal strength measurements are sorted according to the signal strength to obtain the current sorting result. The current sorting result is input into the direction estimation model, and the direction region where the digital key is currently located is determined from the multiple direction regions according to the maximum a posteriori probability criterion.

[0016] The aforementioned direction recognition step is one of the core steps before digital key positioning. By offline statistical analysis of the conditional probabilities of each sorting result appearing in different directional regions and establishing a direction estimation model, the direction recognition can be completed online simply by sorting the real-time signal strength and applying the maximum a posteriori probability criterion. The computational load is small, making it suitable for real-time operation on low-computing-power controllers at the vehicle end.

[0017] In some implementations, sorting the signal strength measurements of each positioning anchor station in each sample according to the magnitude of the signal strength to obtain the sorting result corresponding to the sample includes: arranging multiple signal strength measurements in each sample in descending order or ascending order to obtain an arrangement sequence composed of the corresponding positioning anchor station identifiers, and using the arrangement sequence as the sorting result corresponding to the sample; calculating the conditional probability of each sorting result occurring in each directional region includes: for each directional region, counting the number of times each sorting result occurs in that directional region, and using the ratio of the number of occurrences to the total number of samples in each directional region as the conditional probability of the sorting result occurring in that directional region.

[0018] In some implementations, inputting the current sorting result into the directional statistical model and determining the directional region where the digital key is currently located from the plurality of directional regions based on the maximum a posteriori probability criterion includes: for each directional region, determining the conditional probability of the current sorting result appearing in that directional region. and the marginal probability of the current sorting result Calculate the posterior probability that the digital key is located in the region of that direction; the formula for calculating the posterior probability is: m = A, B, C, ..., L, where, , N is the total number of samples in all directions. For sorting results in the m direction The frequency of occurrence, A, B, C, ..., L represent the various directions; the prior probability P(m) is the probability of the key being in a certain direction; the direction region with the highest posterior probability is determined as the direction region where the digital key is currently located.

[0019] In some implementations, the prior probability of the digital key being located in each directional region is equal.

[0020] Thirdly, the present invention provides another method for locating a car digital key, used to determine the directional region where the car digital key is located in the car digital key location method described above, and to determine the current directional region of the digital key through a machine learning model, including the following steps:

[0021] During the offline phase, signal strength measurement samples are collected for each directional region. The signal strength measurement values ​​of each positioning anchor station in each sample are subtracted pairwise to obtain a feature vector composed of multiple signal strength differences. The directional region to which the sample belongs is used as a label to form a training sample. The training sample is used to train a multi-class machine learning model to obtain a directional classification model.

[0022] During the online phase, the real-time feature vector is obtained by subtracting multiple real-time signal strength measurements pairwise. The real-time feature vector is then input into the direction classification model to obtain the probability that the digital key is in each directional region. The directional region with the highest probability is determined as the directional region where the digital key is currently located.

[0023] In the above technical solutions, the signal strength difference characteristics depend only on the relative relationship of the signal strength of each anchor station, are not affected by the difference in digital key transmission power, do not require calibration for different types of terminals, and have good generalization and portability.

[0024] In some implementations, the multi-class machine learning model includes at least one of decision trees, multilayer perceptrons, support vector machines, k-nearest neighbors classifiers, and Naive Bayes classifiers.

[0025] In some embodiments, the wireless signal is a Bluetooth Low Energy signal, and the positioning anchor station is a Bluetooth anchor station; the digital key includes at least one of a smartphone, a smart wearable device, or a physical Bluetooth key.

[0026] In some embodiments, at least some of the plurality of positioning anchor stations are arranged on the outer side of the vehicle body, including at least two locations in the left front region, right front region, left rear region and right rear region of the vehicle.

[0027] Finally, the present invention provides a computer program product, which, when executed by a processor, implements the steps of the car digital key positioning method or the car digital key direction recognition method as described above.

[0028] Compared with the prior art, the present invention has the following technical effects:

[0029] (1) This invention utilizes the relative strength of the signals received by multiple Bluetooth anchor stations at the same time to identify the directional region of the digital key: since the anchor station closer to the digital key and propagating at its line of sight receives a stronger signal, while the anchor station blocked by the target object receives a weaker signal with more drastic fluctuations, the ranking of the signal strength of each anchor station can stably reflect the direction of the digital key; by offline statistical analysis of the conditional probability of each ranking result appearing in different directional regions and the marginal probability of appearing in the entire sample space, and establishing a directional statistical model, online, only the real-time signal strength needs to be ranked and the direction identification can be completed according to the maximum a posteriori probability criterion, which has a small computational load and is suitable for real-time operation on a low-computing-power controller at the vehicle end. Experiments have shown that the probability of correctly identifying a certain direction using this method is as high as 87.1%.

[0030] (2) The sorting results and signal strength difference features depend only on the relative relationship of the signal strength of each anchor station, so that the direction identification method of the present invention is not affected by the difference in digital key transmission power, and does not need to be calibrated for different types of terminals one by one, and has good generalization ability and portability.

[0031] (3) Based on the direction recognition results, the positioning algorithm can eliminate anchor stations that are occluded or whose measurement results are unreliable, and only select anchor stations with line-of-sight propagation and reliable measurement results for position calculation, thereby making the present invention significantly improve the positioning accuracy in occluded scenarios. Attached Figure Description

[0032] The invention, its features and advantages will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0033] Figure 1 This is an illustration of a scenario where a mobile phone is located on the left side of a vehicle. The stars in the image indicate anchor points.

[0034] Figure 2 The relationship between the received signal power of the Bluetooth node on the left B-pillar and the distance between the transceiver and the transceiver is shown.

[0035] Figure 3The relationship between the received signal power of the Bluetooth node on the right B-pillar and the distance between the transceiver and the transceiver is shown.

[0036] Figure 4 This is an illustration of a scenario where the phone key is located at the left rear of the vehicle. The stars in the image indicate anchor points.

[0037] Figure 5 In order to be in Figure 4 The relationship between the signal power received by the Bluetooth node at the front bumper and the distance between the transceiver in the scenario shown.

[0038] Figure 6 This is a schematic diagram of a six-anchor station arrangement for a Bluetooth key according to an embodiment of the present invention. Stars in the diagram indicate the anchor station locations.

[0039] Figure 7 This is a schematic diagram of a scenario in one embodiment of the present invention where the positioning plane is divided into 12 directions.

[0040] Figure 8 This is a flowchart illustrating a digital key direction recognition method according to an embodiment of the present invention.

[0041] Figure 9 This is a schematic diagram of orientation recognition based on machine learning in one embodiment of the present invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but these are not intended to limit the scope of the invention.

[0043] The core idea of ​​this application is as follows: When using multiple Bluetooth anchor stations to locate a digital key based on signal strength, the relative strength of the signal strength measurements of each anchor station at the same time is first used to determine the directional area of ​​the digital key relative to the target object. Then, based on the directional area, a group of anchor stations with reliable signal propagation paths (e.g., line-of-sight propagation with the digital key) are selected to participate in the positioning calculation, thereby avoiding the contamination of the positioning results by unreliable measurement results from obstructed anchor stations. The following explanation uses Bluetooth Low Energy (BLE) signals as the wireless signal and Bluetooth anchor stations as the positioning anchor stations. Those skilled in the art will understand that the method of this application is also applicable to digital keys in the form of physical Bluetooth keys (FOB), and to positioning systems based on other short-range wireless communication systems besides Bluetooth (e.g., Wi-Fi, ZigBee, etc.). It can also be applied to positioning scenarios outside the automotive industry, such as indoor service robots and asset positioning in warehousing and logistics. This application does not limit these applications.

[0044] In some embodiments of this application, a digital key direction recognition method is provided. This method is executed by a target object deployed with multiple Bluetooth anchor stations. The multiple Bluetooth anchor stations are used to receive wireless signals transmitted by the digital key and measure the received signal strength. The method includes: dividing the plane surrounding the target object into multiple directional regions; in an offline phase, collecting signal strength measurement values ​​of the received digital key wireless signals from multiple Bluetooth anchor stations in each directional region; sorting the signal strength measurement values ​​of each Bluetooth anchor station in each sample according to signal strength magnitude to obtain the sorting result corresponding to the sample; calculating the conditional probability of each sorting result appearing in each directional region and the marginal probability of its appearance in the entire sample space; and based on each direction... A direction estimation model is established based on the region's samples and their ranking results, conditional probability, and marginal probability. During the online phase, multiple real-time signal strength measurements are obtained from multiple Bluetooth anchor stations simultaneously receiving the digital key's wireless signal. These measurements are then ranked according to signal strength to obtain the current ranking result. This ranking result is input into the direction estimation model, and the current direction region of the digital key is determined from multiple direction regions based on the maximum a posteriori probability criterion. Based on the determined direction region, Bluetooth anchor stations with reliable signal propagation paths to the digital key are selected from among the multiple Bluetooth anchor stations to form an anchor station subset. The digital key's location is calculated based on the signal strength measurements of this anchor station subset to obtain the digital key's position information relative to the target object.

[0045] In the specific implementation process, the Bluetooth anchor station deployed on the target object refers to a communication node installed on the target object that has the ability to receive wireless signals and measure signal strength.

[0046] Example 1

[0047] Taking vehicles as an example, car digital key systems typically install multiple positioning anchors on the vehicle. (See...) Figure 6This embodiment presents an arrangement scheme comprising six Bluetooth anchor stations: one on each of the left and right B-pillars, one on each of the front and rear bumpers, and one each on the front and rear ceiling inside the vehicle. It is understood that the number of directions to be divided is related to the desired positioning accuracy. The method for dividing directions is not unique; one possible approach is to draw corresponding marking lines along the desired directions, using the vehicle's two-dimensional geometric center as the origin. The number and placement of anchor stations can be adjusted according to vehicle model structure, cost, and positioning accuracy requirements. For example, four, five, or more anchor stations can be used. External anchor stations are preferably placed around the vehicle's perimeter in the left front, right front, left rear, and right rear areas to ensure that anchor stations in all directions around the vehicle can form a line-of-sight propagation path with the digital key. Each anchor station communicates with the vehicle-side controller (e.g., the body controller or a dedicated digital key controller). The vehicle-side controller periodically collects the RSSI measurements obtained from each anchor station and executes the direction recognition and positioning method of this application. Digital keys can be smart wearable devices such as smartphones and smartwatches with digital key applications installed, or they can be physical Bluetooth keys. Digital keys broadcast or send Bluetooth signals after establishing a connection with anchor stations according to an agreed period. Each anchor station receives the Bluetooth signal and measures the RSSI.

[0048] observe Figure 6 It is known that regardless of the key's orientation on the vehicle, some anchor stations will always be obstructed. To improve positioning accuracy, the positioning algorithm requires measurement results from a set of anchor stations with truly usable RSSI to calculate the key's position. In this embodiment, we determine the direction based on the RSSI obtained by each anchor station at a certain moment, and utilize the relative relationship between the RSSI of each anchor station. For example, if the mobile phone key is located on the left side of the vehicle, the RSSI of the left anchor station should generally be stronger than that of the right anchor station. The division of directions can be based on specific application requirements and the arrangement of anchor stations. For situations where positioning accuracy requirements are not high, it may only be necessary to determine whether the key is on the left, right, front, or rear of the vehicle. In this invention, we divide the horizontal direction into 12 directions, A, B, C, ..., L, with an angle of 30 degrees between each direction, as shown below. Figure 7 As shown.

[0049] See Figure 8 The method proposed in this embodiment includes the following three steps:

[0050] 1. Data Acquisition: Assume that the RSSI measurements of the six anchor stations at a certain moment are R1, R2, R3, R4, R5, and R6, respectively. Collect a certain number of samples along each direction m (m = A, B, C, ..., L), with the same number of samples collected in each direction.

[0051] 2. Modeling process (offline process):

[0052] 1) Select a sample in a certain direction and sort the 6 RSSIs of the sample (R1 R2 R3 R4 R5 R6). For example, they can be sorted in descending order. For example, the sorting result of a sample is s. i = R3 ≥ R1 ≥R4 ≥ R6 ≥ R5 ≥ R2, i = 1, 2, 3,....

[0053] 2) Repeat the sorting by size process in the previous step for all samples in a certain direction, and count the number of times all possible sorting results occur.

[0054] 3) Perform the above two steps on the samples collected in each direction.

[0055] 4) After the above three steps, the possible sorting results s in each direction are obtained. i This constitutes the entire sample space S, i.e., s i ∈S.

[0056] At this point, we have obtained the number of occurrences of each sorting result in each direction, and further obtained information about s. i Two probabilities:

[0057] Conditional probability: Assuming the total number of samples in all directions is N, the ranking result s in direction m (m=A, B, C, ..., L) is... i If the number of occurrences is n(m), then the conditional probability is:

[0058] (2)

[0059] Marginal probability: sorting result s i The probability of s occurring in the entire sample space. i The total probability of occurrence in all directions is:

[0060] (3)

[0061] Following the method in step 4), for each sorting result s in the sample space S... i Calculate the conditional probability and marginal probability for each direction.

[0062] For each sorting result in the sample space S, its conditional probability and marginal probability in each directional region are calculated, thereby establishing a directional estimation model. In practical implementation, a lookup table-like data structure can be designed to associate each possible sorting result with its frequency and conditional probability in each directional region. This lookup table constitutes the directional statistical model and is stored in the memory of the vehicle-end controller.

[0063] It should be noted that the sorting results only utilize the relative magnitude relationship between the RSSIs of each anchor station, and do not depend on the absolute value of the RSSI. Therefore, the differences in the digital key's transmission power and the differences in the radio frequency performance of different terminal models will not change the directional pattern reflected by the sorting results, thus saving the workload of calibration for different terminals.

[0064] 3. Directional reasoning process (online process):

[0065] During key positioning, an inference process is needed to determine the current key's orientation relative to the vehicle. After the vehicle acquires the measured values ​​(R1 R2 R3 R4 R5 R6) in real time, the six RSSIs are first sorted to obtain the sorting result s. j According to the model s j Based on the relevant information, the orientation of the key is inferred using the maximum a posteriori probability criterion. Specifically:

[0066] For each direction, calculate the value when the sorting result is s. j The posterior probability P(m|s) at time t j ), m = A, B, C,..., L, specifically:

[0067] Posterior probability , m=A,B,..., L (4)

[0068] In the above formula, the prior probability P(m) is the probability of the key being in a certain direction. In most application scenarios, it can be assumed that the direction the key is in is equally probable, that is, P(m) is a constant, such as equal to 1 / 12.

[0069] Selecting the maximum posterior probability P(m|s) j The direction corresponding to ) is taken as the final direction recognition result. The maximum posterior probability P(m|s) is calculated according to formula (4). j ), and find the maximum conditional probability P(s) j |m) is equivalent; we only need to know in which direction the current sorting result s is. j The condition with the highest probability is found by searching the current sorted result s. j Direction recognition can be completed by identifying the region with the highest conditional probability. This eliminates the need for floating-point probability calculations, resulting in minimal computational overhead and making it ideal for real-time execution on vehicle-mounted microcontrollers with limited computing power.

[0070] The subsequent positioning algorithm selects a set of reasonable anchor stations based on the direction recognition results, and uses the measurement results of these anchor stations to calculate the location of the key.

[0071] Based on the above methods, the performance of the direction recognition algorithm was tested on a real vehicle, and the resulting confusion matrix is ​​as follows:

[0072] Table 1. Confusion matrix for orientation recognition based on maximum a posteriori probability criterion

[0073]

[0074] Each row in the table corresponds to the probability of a true direction being correctly identified and misidentified as another direction. Taking direction B as an example, the probability of direction B being correctly identified is 87.1%, and the probabilities of being misidentified as direction A, C, D, and F are 2.2%, 8.1%, 2.5%, and 0.1%, respectively. The elements on the diagonal of the table represent the probability of each direction being correctly identified.

[0075] After obtaining the current directional area of ​​the digital key, the subsequent positioning algorithm selects a set of reasonable anchor stations based on the directional area: according to the correspondence between the directional area and the anchor station placement, anchor stations facing that directional area and with line-of-sight propagation to the digital key are selected to form a subset of anchor stations, while anchor stations obscured by the vehicle body and with non-line-of-sight propagation to the digital key are removed. For example, when the direction recognition result is the left side of the vehicle, the left B-pillar anchor station and its adjacent front or rear anchor stations with line-of-sight propagation to the digital key are selected for positioning, while anchor stations such as the right B-pillar obscured by the metal body are removed. Subsequently, based on the RSSI measurement values ​​of each anchor station in the anchor station subset, the position of the digital key relative to the target object is calculated according to a certain positioning algorithm. Since the anchor stations involved in the calculation are all with line-of-sight propagation and a good monotonic relationship between RSSI and distance, the accuracy and stability of the positioning results are significantly improved.

[0076] The above technical solution utilizes the relative strength ranking of RSSI of each anchor station at the same time for direction identification. The direction statistical model can be established simply by offline statistical lookup table, and online inference only requires one ranking and one lookup, resulting in low computational load and good real-time performance. Based on the direction identification results, obstructed anchor stations are eliminated, and only line-of-sight anchor stations are selected for positioning calculation, effectively avoiding the contamination of positioning results by the drastic fluctuations in RSSI of non-line-of-sight anchor stations, and significantly improving positioning accuracy in obstructed scenarios. At the same time, the ranking results are not affected by the differences in terminal transmission power, eliminating the need for individual calibration for different terminals, and exhibiting good generalization.

[0077] Example 2

[0078] This embodiment provides a digital key positioning and direction recognition method, and the execution process is similar to that of Embodiment 1. However, data acquisition can include static data acquisition and dynamic data acquisition. During static data acquisition, a series of points are marked on a line at certain intervals, and the mobile phone key is placed on each point, with the vehicle end collecting data; during dynamic data acquisition, the mobile phone key moves along the direction of the line.

[0079] In some preferred embodiments of this application, the plane containing the target object is divided into multiple directional regions around the target object, including: using the geometric position of the target object on the plane as a reference point, the circumferential angle around the reference point is evenly divided into N directional regions, where N is an integer greater than or equal to 2, and the value of N is determined according to the positioning accuracy requirements and the arrangement of multiple Bluetooth anchor stations.

[0080] In practical implementation, the reference point can be the geometric center of the two-dimensional planar contour of the target object, or other reference positions on the target object, such as the midpoint of the rear axle of a vehicle. The direction region can be a fan-shaped area with the reference point as the vertex and adjacent marker lines as the boundaries, with equal angles between adjacent marker lines, all being 360 degrees / N. The larger the value of N, the finer the granularity of direction recognition, and the more accurate the guidance for subsequent anchor selection, but the workload of offline data collection and modeling also increases accordingly, and the requirements for the number and density of anchors are higher. The smaller the value of N, the coarser the direction recognition, but the simpler the implementation. In actual implementation, N can be selected between 4 and 12 according to the positioning accuracy requirements. For example, N=4 can be used when it is only necessary to distinguish between the front, rear, left, and right sides of the vehicle to select anchors on the same side, and N=12 can be used when it is necessary to accurately select a single anchor. The direction region can be numbered using letters, numbers, or angle ranges, and this application does not impose any limitations.

[0081] By adopting the above technical solution, the directional region is evenly divided and the number of directional regions is flexibly set, so that the directional recognition granularity matches the positioning accuracy requirements and the anchor station layout scheme, thus controlling the offline modeling cost while ensuring the effectiveness of directional recognition.

[0082] In some preferred embodiments of this application, the signal strength measurement value is the received signal strength indication value. Before sorting the signal strength measurement values ​​according to their signal strength magnitude, the method further includes: preprocessing the received signal strength indication values ​​measured by each Bluetooth anchor station. The preprocessing includes at least one of filtering and smoothing, outlier removal, and missing value compensation. The same preprocessing rules can be used in both offline modeling and online inference stages to ensure consistency between the statistical model and real-time data. By preprocessing the RSSI, the interference of random signal fluctuations and abnormal jumps on the sorting results is suppressed, improving the stability of the sorting results and the accuracy of direction recognition.

[0083] In some preferred embodiments of this application, the signal strength measurements of each Bluetooth anchor station in each sample are sorted according to the magnitude of the signal strength to obtain the sorting result corresponding to the sample. This includes: arranging multiple signal strength measurements in each sample in descending order or ascending order to obtain an arrangement sequence composed of the corresponding Bluetooth anchor station identifiers, and using the arrangement sequence as the sorting result corresponding to the sample.

[0084] Example 3

[0085] This embodiment provides another digital key orientation recognition method to improve the accuracy of subsequent key positioning. Based on the orientation recognition ideas proposed in Embodiments 1 and 2, and also based on a set of RSSI results obtained from vehicle-end anchor station measurements, we can also choose a machine learning-based orientation recognition algorithm, which is essentially a multi-classification problem. The specific steps are as follows:

[0086] The direction division and data collection are exactly the same as those given in Example 1. A certain number of samples are collected along each direction to serve as the dataset needed for machine learning training and testing.

[0087] Feature engineering: We construct dataset features by pairwise subtraction of the six RSSI values ​​of the samples (R1 R2 R3 R4 R5 R6). This feature data reflects the orientation of the mobile phone key and eliminates the impact of different transmitter power on machine learning performance, thus saving the calibration workload of different mobile phone keys. The feature data constructed in this embodiment is as follows:

[0088] R1 - R2, R1 - R3, R1 - R4, R1 - R5, R1 - R6

[0089] R2 - R3, R2 - R4, R2 - R5, R2 - R6

[0090] R3 - R4, R3 - R5, R3 - R6

[0091] R4 - R5, R4 - R6

[0092] R5 - R6

[0093] Model training: After constructing feature data in various directions and labeling the feature data, it can be used as the final dataset for machine learning. The machine learning algorithm used for classification can be one of decision trees, multilayer perceptrons, support vector machines, kNN, Naive Bayes, etc. The choice of algorithm and the adjustment of algorithm parameters depend on the trade-off between the embedded computing power and performance of the final inference code and the complexity of porting the trained model to the vehicle.

[0094] Inference process: The feature vectors obtained by pairwise subtraction of real-time RSSI are input into the trained orientation classification model. The model outputs the probability that the digital key is located in each orientation region. The orientation region with the highest probability is taken as the orientation recognition result, such as... Figure 9As shown. During the localization process, the machine learning-based inference code runs on the vehicle's MCU or the more powerful A-core, depending on the design of different vehicles.

[0095] It should be noted that the direction recognition method based on sorting statistics and the direction recognition method based on machine learning can be used selectively, or they can be used in parallel and their outputs can be fused for judgment. This application does not limit this.

[0096] Using the above technical solution, the machine learning multi-classification model with pairwise RSSI differences as features can also identify the direction based on the relative relationship of signal strength of each anchor station. It can eliminate the influence of terminal transmission power differences, eliminate terminal calibration work, and the model selection is flexible and can be adapted to vehicle-end platforms with different computing power levels.

[0097] In practical implementation, the above models are all mature multi-class classification models: Decision trees complete classification by hierarchical judgment of feature thresholds, and the inference process involves several comparisons, requiring no floating-point operations, making them the simplest to port; Naive Bayes calculates the posterior probability of each class based on the feature conditional independence assumption, consistent with the statistical ideas of this application, and has fewer model parameters; the k-nearest neighbor classifier does not require explicit training, directly using training samples as references for nearest neighbor voting; Support Vector Machines handle nonlinear classification boundaries through kernel function mapping; Multilayer Perceptrons fit complex feature-class relationships through multilayer nonlinear transformations, offering the highest potential accuracy but with relatively high computational cost. During implementation, multiple models can be trained on the training set, and the direction recognition accuracy, inference time, and memory usage can be compared on the test set to select the model with the best overall performance for the target platform.

[0098] Bluetooth Low Energy (BLE) is currently the most widely used short-range communication technology in the automotive digital key market, widely supported by mobile phones, smartwatches, and other terminals. Vehicle-side Bluetooth anchor stations are low-cost and easy to integrate. The method in this application does not rely on the channel detection and ranging capabilities of Bluetooth 6.0, and is also applicable to Bluetooth versions that only support traditional RSSI measurements, demonstrating good compatibility. Besides Bluetooth, any short-range communication system that uses measured signal strength for ranging and positioning, such as Wi-Fi and ZigBee, can use the method in this application for direction identification and anchor station selection. In addition to automotive digital key scenarios, the method in this application can also be applied to other multi-anchor station positioning scenarios with obstructions, such as indoor robot positioning, warehouse asset tracking, and venue personnel positioning. By adopting the above technical solution, the application scope of this application expands from automotive Bluetooth digital keys to various multi-anchor station positioning systems based on signal strength, demonstrating broad applicability.

[0099] In some embodiments of this application, a computer program product is also provided. When executed by a processor, this program or instruction implements the steps of the digital key orientation recognition method of the first aspect of this application, including sample collection, sorting statistics, and model training (in a machine learning implementation) in the offline modeling stage, and sorting, posterior probability decision, anchor station selection, and positioning calculation in the online stage. The electronic device where the processor is located can be a mobile electronic device or a fixed electronic device. In the automotive digital key scenario, the electronic device where the processor is located is preferably a vehicle-side controller deployed on the vehicle, such as a body controller, digital key controller, or vehicle host, which connects to each Bluetooth anchor station through a communication interface, collects RSSI measurement values, and executes the method of this application.

[0100] Those skilled in the art should understand that variations can be implemented by combining existing technology with the above embodiments, which will not be elaborated here. Such variations do not affect the essence of the present invention, and will not be elaborated here either.

[0101] The preferred embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above. Systems and structures not described in detail should be understood as being implemented in a conventional manner in the art. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention using the methods and techniques disclosed above, or modify them into equivalent embodiments with equivalent changes, without departing from the scope of the present invention. This does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for locating a vehicle using a digital key, comprising dividing the horizontal plane where the vehicle is located into a finite number of different directions and deploying multiple positioning anchor stations; wherein the positioning anchor stations are used to receive wireless signals transmitted by the digital key and measure the received signal strength of the wireless signals; characterized in that: Includes the following steps: First, determine the directional area where the car digital key is located. Then, based on the determined directional area, select a subset of positioning anchor stations from the plurality of positioning anchor stations that have a reliable signal propagation path with the digital key. Based on the signal strength measurement values ​​of the anchor station subset, perform positioning calculation on the digital key to obtain the position information of the digital key relative to the vehicle.

2. A method for recognizing the orientation of a car digital key, characterized in that: The method for determining the directional region where the vehicle digital key is located in the vehicle digital key positioning method as described in claim 1 includes the following steps: During the offline phase, signal strength measurement samples of the multiple positioning anchor stations receiving the wireless signal of the digital key are collected in each directional area. The signal strength measurement values ​​of each positioning anchor station in each sample are sorted according to the signal strength magnitude to obtain the sorting result corresponding to the sample. The conditional probability of each sorting result appearing in each directional area and the marginal probability of appearing in the entire sample space are calculated. Based on the samples in each directional area and their sorting results, conditional probability and marginal probability, a directional estimation model is established. During the online phase, multiple real-time signal strength measurements are obtained from the wireless signals of the digital key received by the multiple positioning anchor stations at the same time. The multiple real-time signal strength measurements are sorted according to the signal strength to obtain the current sorting result. The current sorting result is input into the direction estimation model, and the direction region where the digital key is currently located is determined from the multiple direction regions according to the maximum a posteriori probability criterion.

3. The method for recognizing the direction of a car digital key according to claim 2, characterized in that, The step of sorting the signal strength measurements of each positioning anchor station in each sample according to the signal strength magnitude to obtain the sorting result corresponding to the sample includes: arranging multiple signal strength measurements in each sample in descending order or ascending order to obtain an arrangement sequence composed of the corresponding positioning anchor station identifiers, and using the arrangement sequence as the sorting result corresponding to the sample; the step of calculating the conditional probability of each sorting result occurring in each directional region includes: for each directional region, calculating the number of times each sorting result occurs in that directional region, and using the ratio of the number of occurrences to the total number of samples in each directional region as the conditional probability of the sorting result occurring in that directional region.

4. The method for recognizing the direction of a car digital key according to claim 2, characterized in that, The step of inputting the current sorting result into the direction estimation model and determining the current direction region of the digital key from the multiple direction regions based on the maximum a posteriori probability criterion includes: for each direction region, determining the conditional probability of the current sorting result appearing in that direction region based on the conditional probability of the current sorting result appearing in that direction region. and the marginal probability of the current sorting result Calculate the posterior probability that the digital key is located in the region of that direction; the formula for calculating the posterior probability is: , m = A, B, C, ..., L, where, , N is the total number of samples in all directions. For sorting results in the m direction The frequency of occurrence, A, B, C, ..., L represent the various directions; the prior probability P(m) is the probability of the key being in a certain direction; the direction region with the highest posterior probability is determined as the direction region where the digital key is currently located.

5. The method for recognizing the direction of a car digital key according to claim 4, characterized in that, The prior probability of the digital key being located in each directional region is equal.

6. A method for recognizing the direction of a car digital key, characterized in that: To determine the directional region where the car digital key is located in the car digital key positioning method as described in claim 1, and to determine the current directional region of the digital key through a machine learning model, the method includes the following steps: During the offline phase, signal strength measurement samples are collected for each directional region. The signal strength measurement values ​​of each positioning anchor station in each sample are subtracted pairwise to obtain a feature vector composed of multiple signal strength differences. The directional region to which the sample belongs is used as a label to form a training sample. The training sample is used to train a multi-class machine learning model to obtain a directional classification model. During the online phase, the real-time feature vector is obtained by subtracting multiple real-time signal strength measurements pairwise. The real-time feature vector is then input into the direction classification model to obtain the probability that the digital key is in each directional region. The directional region with the highest probability is determined as the directional region where the digital key is currently located.

7. The method for recognizing the direction of a car digital key according to claim 6, characterized in that, The multi-class machine learning model includes at least one of decision tree, multilayer perceptron, support vector machine, k-nearest neighbor classifier and Naive Bayes classifier.

8. A method for recognizing the direction of a car digital key according to any one of claims 2 to 7, characterized in that, The wireless signal is a low-power Bluetooth signal, and the positioning anchor station is a Bluetooth anchor station; the digital key includes at least one of a smartphone, a smart wearable device, or a physical Bluetooth key.

9. A method for recognizing the direction of a car digital key according to any one of claims 2 to 7, characterized in that, At least some of the multiple positioning anchor stations are arranged on the outer side of the vehicle body, including at least two locations in the left front area, right front area, left rear area and right rear area of ​​the vehicle.

10. A computer program product, characterized in that, When the computer program is executed by the processor, it implements the steps of the car digital key positioning method as described in claim 1 or the car digital key direction recognition method as described in any one of claims 2 to 9.

Images