Method and apparatus for sensor selection for positioning and tracking

Abstract

Methods and apparatuses for sensor selection for positioning and tracking are provided. A method performed at an access point (101, 101a, 101b, 101c, 901, 902, 903) includes: determining a detectability (302) of the access point (101, 101a, 101b, 101c, 901, 902, 903) for positioning a target device (102) based on channel state information of a radio link between the target device (102) and the access point (101, 101a, 101b, 101c, 901, 902, 903); and determining whether the access point (101, 101a, 101b, 101c, 901, 902, 903) will be used for location estimation (304) of the target device (102) based on the detectability. A positioning server may further eliminate access points (101, 101a, 101b, 101c, 901, 902, 903) in bad positions from the location estimation by using a predefined weighted kernel method.

Description

Technical Field

[0001] Embodiments of the present disclosure generally relate to positioning, and more particularly, to methods, apparatuses, and computer-readable storage media for sensor selection for indoor positioning and tracking. Background Art

[0002] Precise positioning opens up a series of new possibilities for mobile services. Consumers will benefit from personalized context information and services, as well as new services such as navigation. It will also create new marketing opportunities, meaning that appropriate services and information can be provided based on the user's current location or future location. Emerging location-based services (LBS) include social networking, people-finding tools, marketing campaigns, asset tracking, etc. In addition, precise positioning can not only simplify people's lives, but also have a huge impact by helping firefighters, police officers, soldiers, and medical staff save lives and perform specific tasks.

[0003] From this perspective, the Indoor Location Alliance (ILA), consisting of 22 member companies (now ILA has expanded to 95 member companies), has been launched to promote innovation and market adoption of high-precision indoor positioning and related services.

[0004] However, when implementing high-precision indoor positioning, there are various difficulties. Due to the unreliability and obstacles in the indoor environment, standard methods for outdoor positioning, including the Global Positioning System (GPS), are not easily used.

[0005] One solution is that cellular carriers provide a unified continuous positioning system as a communication system. Therefore, Wi-Fi or WLAN-based positioning services are ideal ways to cover shopping malls, airports, and other large buildings (such as large exhibition areas).

[0006] To improve positioning accuracy, a large number of access points (APs) are usually deployed. Increasing the number of APs helps to distinguish more obvious locations. However, how to improve the accuracy of indoor positioning, especially how to select sensors in real time for precise positioning, has become a difficult problem. This is because it is challenging to design effective algorithms to overcome the exhaustive search of all possible subsets of sensors to optimize performance. Summary of the Invention

[0007] The present disclosure will solve the above problems by presenting an effective solution for sensor selection for precise positioning and tracking, for example, in indoor positioning and tracking. Other features and advantages of the embodiments of the present disclosure will also be understood from the following description of specific embodiments when read in conjunction with the accompanying drawings, which illustrate the principles of the embodiments of the present disclosure by way of example.

[0008] According to a first aspect of the present disclosure, a method performed at an access point is provided. The method includes: determining a detectability of an access point for locating a target device based on channel state information of a radio link between the target device and the access point; and determining whether the access point will be used for location estimation of the target device based on the detectability.

[0009] In some embodiments, the detectability includes the detectability of an access point for locating a target device in indoor positioning and tracking.

[0010] In some embodiments, if the detectability is lower than a predetermined threshold, it may be determined that the access point will not be used for location estimation of the target device.

[0011] In some embodiments, determining the detectability may include: training a deep neural network using a set of channel state information, corresponding estimated angles, and ground truth; and predicting the detectability by using the trained deep neural network. In some embodiments, the deep neural network may include a convolutional neural network, a long short-term memory layer, and a fully connected layer, wherein the long short-term memory layer is coupled between the convolutional neural network and the fully connected layer.

[0012] In some embodiments, the method may further include: sampling channel state information of the radio link within a predetermined time period.

[0013] In some embodiments, the method may further include: obtaining information about an adjustment to the radio link. The determination of the detectability may be further based on the information about the adjustment. In some embodiments, the adjustment to the radio link may include a change in the frequency band and / or power of the radio link.

[0014] In some embodiments, the method may further include: determining a weight of the access point based on the determined detectability. In some embodiments, the method may additionally include: sending the weight to a location server.

[0015] According to a second aspect of the present disclosure, a method performed at a location server is provided. The method includes: determining weights of at least two access points, wherein the weights of the access points are associated with the detectability of the access points for locating a target device; constructing respective weighted kernel functions for each of the at least two access points; calculating an inner product of the weight of a corresponding access point and the weighted kernel function constructed for the corresponding access point for each of the at least two access points; constructing an estimation error function by using the inner products for the at least two access points; determining weight parameters of the corresponding weighted kernel functions to optimize the estimation error function with a constraint condition of L1 regularization; and selecting one or more access points to be used for location estimation of the target device from the at least two access points according to the determined weight parameters.

[0016] In some embodiments, weights may be received from respective access points of at least two access points.

[0017] In some embodiments, if a determined weight parameter of a weighted kernel function constructed for an access point is non - zero, it may be determined that the access point will be used for location estimation.

[0018] In some embodiments, the weighted kernel function may include a Gaussian function.

[0019] In some embodiments, the optimization of the estimation error function is further constrained by the range of distances that the target device may move.

[0020] In some embodiments, the method may further include: receiving channel state information of a radio link between the target device and a specific access point among at least two access points.

[0021] In some embodiments, the method may further include: using the received channel state information to estimate the location of the target device.

[0022] In some embodiments, the method may further include: based on the channel state information, determining the detectability of a specific access point for locating the target device; and based on the determined detectability, determining whether the specific access point will be used for the location estimation of the target device.

[0023] In some embodiments, determining the detectability may include: training a deep neural network with a set of channel state information, corresponding estimated angles, and ground truth; and predicting the detectability by using the trained deep neural network.

[0024] In some embodiments, the method may further include: receiving information about an adjustment to a radio link between the target device and a specific access point. The determination of the detectability of the specific access point may be further based on the adjustment.

[0025] In some embodiments, the method may further include: determining the weight of a specific access point based on the determined detectability.

[0026] According to a third aspect of the present disclosure, an apparatus includes: at least one processor; and at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processor, perform the method according to the first aspect.

[0027] According to a fourth aspect of the present disclosure, an apparatus includes: at least one processor; and at least one memory including computer program code, the at least one memory and the computer program code being configured to, with the at least one processor, perform the method according to the second aspect.

[0028] According to a fifth aspect of the present disclosure, there is provided a computer-readable storage medium having instructions stored thereon, which, when executed by at least one processor, cause the at least one processor to perform the method according to the first aspect.

[0029] According to a fifth aspect of the present disclosure, there is provided a computer-readable storage medium having instructions stored thereon, which, when executed by at least one processor, cause the at least one processor to perform the method according to the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Some exemplary embodiments will now be described with reference to the drawings, in which:

[0031] Figure 1 shows an overview of a networking system according to an embodiment of the present disclosure;

[0032] Figure 2 shows a process for sensor selection according to an embodiment of the present disclosure;

[0033] Figure 3 is a flowchart depicting a process for sensor selection according to an embodiment of the present disclosure;

[0034] Figure 4 is another flowchart depicting a process for sensor selection according to an embodiment of the present disclosure;

[0035] Figure 5A is a graph showing the CSI (Channel State Information) distribution for a target with a strong line of sight (LOS);

[0036] Figure 5B is a graph showing the CSI distribution for a target passing through an obstacle;

[0037] Figure 6 shows an exemplary architecture of a deep neural network for determining detectability from CSI according to an embodiment of the present disclosure;

[0038] Figure 7 shows an estimation error model based on angle positioning;

[0039] Figures 8A - 8C show a sensor selection process according to an embodiment of the present disclosure;

[0040] Figure 9 shows a scenario of sensor selection using distance constraints;

[0041] Figure 10 shows an exemplary access point distribution and map of a residence where indoor positioning and tracking can be performed;

[0042] Figure 11Shows the estimation results based on all access points;

[0043] Figure 12 Shows the estimation results based on the sensor selection scheme according to an embodiment of the present disclosure;

[0044] Figure 13 Shows Figure 12 the measurement cumulative error distribution; and

[0045] Figure 14 Displays a simplified block diagram of a device according to an embodiment of the present disclosure. Detailed Description of the Invention

[0046] Embodiments of the present disclosure are described in detail with reference to the accompanying drawings. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and thus implement the present disclosure, rather than imposing any limitation on the scope of the present disclosure. References to features, advantages, or similar language throughout the specification do not imply that all features and advantages achievable by the present disclosure should or exist in any single embodiment of the present disclosure. On the contrary, language referring to features and advantages is understood to mean that a particular feature, advantage, or characteristic described in connection with an embodiment is included in at least one embodiment of the present disclosure.

[0047] In addition, the described features, advantages, and characteristics of the present disclosure may be combined in any suitable manner in one or more embodiments. Those skilled in the relevant art will recognize that the present disclosure may be practiced without one or more specific features or advantages of a particular embodiment. In other instances, additional features and advantages may be recognized in certain embodiments, which may not exist in all embodiments of the present disclosure.

[0048] Figure 1 is a block diagram showing an overview of a network system according to an embodiment of the present disclosure. Referring to Figure 1 , network configuration 100, such as a mesh network configuration, includes one or more access points (APs), access nodes, or access links 101a - 101c (collectively 101), which may provide a stable wireless communication network or wireless access network, such as a wireless local area network (WLAN) (e.g., Wi-Fi) network, etc. For example, access points may be deployed to support high-speed Internet access in a smart campus or a large shopping mall. Terminal device 102 may communicate with the access points via a wireless local area network (WLAN) (e.g., Wi-Fi) radio and / or a radio capable of wireless communication. Although three access points are shown, more access points may be coupled to the positioning server 103. Alternatively, network configuration 100 including one or more access points (APs) 101a - 101c (collectively 101) may provide a wireless local communication network, such as or a UWB (ultra-wideband) network, etc.

[0049] The access node can be communicatively coupled to the positioning server 103 via the network 110. The network 110 can be any type of network, such as a local area network (LAN), a wide area network (WAN) such as the Internet, a cellular network, or a combination thereof, wired or wireless. All access points in the access point 101 can share the information of the terminal device 102 for obtaining channel state information on the network 110. As will be understood, by directly or indirectly connecting the AP 101 and the positioning server 103 and / or any one of many other devices to the network 110, the AP 101 can communicate with each other, communicate with the positioning server 103, etc., so as to implement various functions for positioning, such as sending data, instructions, information, etc. to each other, and / or receiving data, instructions, information, etc. from each other.

[0050] As used herein, the terms "data", "instructions", "information" and similar terms can be used interchangeably to refer to data that can be transmitted, received and / or stored according to embodiments of the present invention. Therefore, the use of any such term should not be construed as limiting the spirit and scope of the embodiments of the present invention.

[0051] The access point 101 can be any type of network device that can be configured to provide wireless access to a target device 102 via a radio link, based on technologies such as radio frequency (RF), infrared (IrDA), or any of many different wireless network technologies, including WLAN technology, Worldwide Interoperability for Microwave Access (WiMAX) technology, and / or Wireless Personal Area Network (WPAN) technology, Bluetooth (BT), Ultra Wideband (UWB), Code Division Multiple Access (CDMA)-based wireless cellular communication technology, High Rate Packet Data (HRPD), Universal Terrestrial Radio Access Network (UTRAN), Long Term Evolution (LTE), LTE-Advanced (LTE-A), Fifth Generation (5G) cellular systems, etc. It should be understood that the illustrated embodiments are non-limiting, and as will be readily understood by those skilled in the art, any number of various wireless devices and telecommunication systems can be employed. The positioning server 103 can be any type of server, such as a Web or cloud server, application server, backend server, edge server, base station, personal computer, the target device 102, the access point 101, or a combination thereof. The access point 101 and the positioning server 103 can be configured to support positioning, such as indoor positioning based on a Wi-Fi platform using the Angle of Arrival (AOA) method. The access point 101 can be configured to send positioning-related information of a target to the positioning server 103. For example, the positioning-related information can include channel state information of the communication with the target. The positioning server 103 can be configured to collect positioning-related information of a target (such as the terminal device 102) from the access point 101 and estimate the position of the target. In some embodiments, the positioning server can be installed in the same entity as the access point. In some embodiments, the functions of the positioning server can be distributed among multiple entities, including the access point and the server. Figure 1 The system can also be used for outdoor positioning, for example, in IoT (Internet of Things) solutions.

[0052] From a positioning perspective, the terminal device can also be referred to as the target device. It refers to any terminal device that can access a wireless communication network and receive services therefrom. By way of example and not limitation, the target device can refer to a User Equipment (UE) or other suitable device. The UE can be, for example, a user station, a portable user station, a Mobile Station (MS), or an Access Terminal (AT), or any combination thereof. The target device can include, but is not limited to, a portable computer, an image capture device such as a digital camera, a game terminal device, a music storage and playback device, a mobile communication device, a mobile phone, a cellular phone, a smart phone, a navigation device, a tablet computer, a wearable device, a smart watch, a fitness band, a remote monitoring band, a Personal Digital Assistant (PDA), a vehicle, an Internet of Things device, a sensor device, etc., or any combination thereof.

[0053] As described above, in general, sensor selection is a difficult problem because it is challenging to design an effective algorithm to overcome the exhaustive search of all possible subsets of sensors to optimize performance.

[0054] The present disclosure provides a predefined weighted kernel method to simplify sensor selection. Figure 2 A process for selecting sensors to obtain a given target location according to the present disclosure is shown. Based on the detectability of each AP, a weight range (e.g., denoted as w1, w2, w3) is assigned to each AP, as shown in 220-1, 220-2, and 220-3 respectively. Detectability indicates the observation quality of the AP regarding the given target location estimation. Machine learning can be used to predict the detectability of a specific AP (e.g., X1, X2, X3 of the corresponding AP) based on the channel state information 210 of the communication channel between the given target and the specific AP. In machine learning, a map can be created based on the detectability of the AP and the probability of detectability. APs with detectability below a threshold can be removed. Thus, the impact of APs with poor channel states on positioning can be mitigated.

[0055] Additionally, as shown in 230, a weighted kernel can be constructed to mitigate the impact of APs at poor locations. Furthermore, an estimation error (which can be based on angle positioning) can be introduced as a loss function and LASSO (Least Absolute Shrinkage and Selection Operator) regularization to eliminate APs at poor locations, as shown in 240. An L1 distance constraint can also be added to the loss function.

[0056] Now refer to Figure 3 , which shows a flowchart of a method 300 according to some embodiments of the present disclosure. The method 300 can be implemented at an access point (e.g., a Wi-Fi or WLAN access point, such as the AP 101 shown in Figure 1 ). As shown in Figure 3 , the method 300 can include: at block 302, determining the detectability of an access point for positioning a target device based on channel state information (CSI) data of a radio link between the target device and the access point; and at block 304, determining whether the access point will be used for the location estimation of the target device based on the detectability. If the detectability of the access point is below a predetermined threshold, it can be determined that the access point will not be used for the location estimation of the target device. Thus, the impact of access points with poor channel states on the location estimation in positioning can be mitigated. Therefore, the number of access points in positioning and the computational complexity can be significantly reduced. In an alternative solution, the method 300 can determine the detectability of an access point for indoor positioning based on CSI data, measured received signal strength indicator (RSSI) data, and / or received channel power indicator (RCPI) data of a radio link between the target device and the access point.

[0057] In some embodiments, method 300 may further include determining (e.g., assigning) a weight to an access point at block 306 based on the determined detectability, where an access point with higher detectability is assigned a higher weight. The weight may be determined in response to a determination that the access point will be used for location estimation of a target device. In one example, an access point with low detectability may be assigned a lower weight. Thus, an access point with high detectability may be assigned a higher weight.

[0058] In some embodiments, method 300 may further include: at block 308, sending the weight from the access point to a location server. The access point may sample channel state information of the radio link over a predetermined time period. Channel state information (CSI) is information about the channel attributes of a radio link. It describes the factors that attenuate the signal in each transmission path, such as scattering, fading, multipath fading, or shadow fading, as well as the power attenuation over distance. CSI may additionally reflect the observation quality of the access point for precisely locating a target.

[0059] In some embodiments, method 300 may further include: obtaining information about an adjustment to the radio link. The determination of detectability may be further based on the adjustment information. The adjustment may include a change in the frequency band and / or power of the radio link.

[0060] Now refer Figure 4 , which shows a flowchart of method 400 according to an embodiment of the present disclosure. Method 400 may be implemented at a location server (e.g., location server 103 as shown Figure 4 . Alternatively, method 400 may be implemented in one or more access points 101, in which case the access points share the determined weights. As shown Figure 4 , method 400 may include: at block 402, determining (e.g., obtaining) the weights of at least two access points, where the weight of an access point is associated with the detectability of the access point for locating a target device; at block 404, constructing a respective weighted kernel function for each of the at least two access points; at block 406, calculating the inner product of the weight of a corresponding access point and the weighted kernel function constructed for the corresponding access point for each of the at least two access points; at block 408, constructing an estimation error function by using the inner products for the at least two access points; at block 410, determining the weight parameters of the corresponding weighted kernel function to optimize the estimation error function with L1 regularization constraints; and at block 412, selecting, based on the determined weight parameters, one or more access points from the at least two access points for location estimation of the target device.

[0061] In some embodiments, the positioning server may additionally utilize the received channel state information to determine the detectability of a particular access point and further determine whether the particular access point will be used to locate the target device based on the determined detectability, in a manner similar to that shown in Figure 3 boxes 302 and 304. Additionally, the positioning server may determine the weight of a particular access point based on the determined detectability, in a manner similar to that shown in Figure 3 box 308. This may be advantageous for power efficiency, for example, in cases where the access point has limited computing power and power. For example, at block 402, the positioning server may receive CSI data and other data, such as RSSI, RCPI, and / or radio link adjustment data, from at least two access points and may determine (e.g., assign) the weights of the corresponding access points based on the determined detectability.

[0062] In another alternative, at block 402, the positioning server may receive weights from the corresponding access points of at least two access points. In this regard, one or more access points 101 may receive CSI data and other data, such as RSSI, RCPI, and / or radio link adjustment data, from at least two access points and may determine (e.g., assign) the weights of the corresponding access points based on the determined detectability.

[0063] In some embodiments, if the determined weight parameter of the weighted kernel function constructed for an access point is non - zero, it may be determined that the access point will be used for position estimation. In some embodiments, if the determined weight parameter of the weighted kernel function constructed for an access point is close to zero, it may be determined that the access point will not be used for position estimation.

[0064] In some embodiments, method 400 may further include receiving channel state information of the radio link between the target device and a particular access point among at least two access points. The received channel state information may be used to estimate the position of the target device.

[0065] The detectability of an access point may be determined by machine - learning techniques. Figure 5A And 5B show that there are significant differences when a Wi - Fi access point can detect a target with various detectabilities. In practice, one can easily determine whether an access point is of high quality based on the CSI distribution map and / or related information. For example, it is obvious that Figure 5A the CSI distribution in Figure 5B is stable, for example, in terms of sub - carriers and phases; while the CSI distribution in

[0066] In the past, the detectability of sensors could be determined by statistical methods, which did not work well in practice. A new paradigm based on deep neural networks (DNNs) emerged to alleviate the problem of manual tuning. By using DNNs, higher confidence can be achieved in the solution of the present disclosure to remove very poor sensors based on detectability. Figure 6 FIG. Figure 6 depicts the proposed DNN architecture 600, which mainly consists of a feature extractor part and a quality regressor part. The feature extractor part can be a convolutional neural network (CNN) 601, which is configured to learn from the raw channel state information (CSI). Instead of learning from the estimated angle of arrival, the raw CSI is preferred. The regressor part 602 consists of a recurrent network, which may include a long short-term memory (LSTM) layer 602 and a fully connected layer 603. Compared with directly using a fully connected layer, the LSTM layer 602 is more suitable for learning long dependencies over time. It should be understood that other DNNs with the ability to extract features and data dependencies can also be used to determine detectability.

[0067] During the training process, the input is the raw CSI data, and the output can be the estimated angle. By comparing the estimated angle with the ground truth, a mapping between the observation quality and the raw CSI can be established. Then, through the trained DNN network, the detectability of the access point can be reliably predicted from the new CSI. With the help of machine learning, the problem of being highly sensitive to noise and outliers caused by manually adjusting parameters statistically can be alleviated. In addition to the raw CSI data, other data such as RSSI, RCPI, and / or radio link adjustment data can also be used in the training procedure.

[0068] Although a machine learning method is proposed to improve the detectability of APs as sensors, it is found that this machine learning method may still have a computational cost. Existing APs are designed for communication and can automatically adjust some parameters of the radio link, such as when a conflict occurs. In this case, the CSI deterioration may be caused by the adjustment rather than poor detectability. In some embodiments, the AP is further configured to report the adjustment of the radio link, such as changes in the frequency band and / or power of the radio link. Based on the adjustment, the detectability can be determined more accurately.

[0069] Based on the detectability of each AP, a weight (e.g., denoted as w i ) for each AP can be determined (e.g., assigned). The weight is the probability of the quantization of the detectability of the access node. The weight can be pre-given a range. Thus, the complexity of sensor selection can be significantly reduced. For example, access points with detectability below a defined threshold can be removed.

[0070] In the step shown in block 404, a kernel function can be constructed to mitigate the impact of bad access points based on detectability or corresponding weights. However, it cannot select access points at good locations. Therefore, the estimated error area as a loss function and LASSO (Least Absolute Shrinkage and Selection Operator) regularization are introduced to eliminate sensors at bad locations. Figure 7 An estimated error model based on angle positioning is presented, where the shaded area shows the estimated error area. Similar to Figure 1 , the estimated error of the best access point results in the smallest area. The sensor selection problem can be solved by borrowing machine learning techniques of LASSO, which significantly reduces the computational complexity. In this regard, the non-convex optimization problem can be transformed into a linear programming problem, which is beneficial for the implementation in practice.

[0071] Figures 8A - 8C show the process of sensor selection according to an embodiment. As shown in Figure 8A, multiple access points can be used as candidate sensors (represented as gray dots) for estimating the position of a target device (represented as a star) at a certain moment. As shown in Figure 8B, sensors with higher detectability are selected first, and these sensors are marked as white dots with bold borders. The gray dots with bold borders in Figure 8B represent sensors with low detectability and / or weights due to poor observation quality, and these sensors are still available. Second, sensors for estimating the position of the target device are selected based on the positions of the sensors marked as black dots. Through this process, sensor selection can be performed with higher accuracy and minimum power consumption.

[0072] If other constraints are introduced, other constraints can be introduced into the loss function with a regularization part. For example, the prediction of the movement of the target device can lead to a boundary for sensor selection based on the movement speed of the target device. Figure 9 A scenario of sensor selection using distance constraints in a mesh network configuration is shown. At a certain moment, a target device (such as a mobile phone) is located at 904 via access points 901, 902, and 903. After a period of time, the target device will be further located. According to the movement of the target device, the boundary of the possible positions of the target device can be determined. Therefore, positions outside the boundary (e.g., as shown by 904'-2 and 904'-3) can be excluded from the estimated error area. In one embodiment, a distance constraint can be added to the loss function to improve the positioning accuracy and power consumption.

[0073] More detailed embodiments of the solution of the present disclosure will be provided below, for example, in a mesh network configuration. In one example, 4 APs are deployed in a large villa for experiments, as Figure 10 shown, and this figure shows the positions of the APs and the reference directions. AP i(i = 1, 2, 3, or 4) The angle between the x-axis is marked for angle direction estimation. The target device (102) carried by a person (such as a mobile phone) is used as the target device for position determination and tracking in daily life.

[0074] CSI data of the radio link between one or more APs (denoted as AP i ) and the mobile phone can be obtained as sensor data (denoted as X i ) from one or more AP sides. Then, the CSI data can be analyzed through machine learning techniques to construct a map of the detectability and observation quality probability for the AP i . The analysis of the CSI data can be performed on the AP or the positioning server. In addition to the CSI data, other data such as RSSI, RCPI, and / or radio link adjustment data can also be analyzed through machine learning techniques.

[0075] Here, the sensor data X = {X i} is quantized and represented as Z = {z i} for the decision rule that is used to determine whether the AP will be used for positioning the mobile phone based on detectability. z i Is used by the quantization value X i Of the AP i . For example, the decision rule can be determined by using the kernel function k(·). Sensors with lower detectability are assigned lower weights. In Figure 10 In the shown example, the detectability of AP4 is very poor because there are multiple walls between the mobile phone and AP4, especially in the hall. The decision rule using the kernel function can be expressed as:

[0076] W(Z) = <w i (·), k(·)>, (1)

[0077] Where w i (·) is the weight assigned to the AP, and <, > represents the inner product operation.

[0078] The above decision rule according to Equation (1) does not consider the influence of the AP position. It is based on the detectability and observation quality of the AP. To remove APs at bad positions for positioning, L1 regularization can be introduced for the kernel weight parameters of sparse sensor selection. More specifically, the optimal weight parameter β = {β i} for all APs can be determined by minimizing the L1 regularization loss function. As described above, the estimated error area is introduced as the loss function. An example function is shown below:

[0079]

[0080] Where φ(·) is the error area function, is a weighted kernel function constructed for AP i and λ||β|| is a component of L1 regularization. In the above function, L-regularization for the kernel weight parameter encourages sparse weight selection.

[0081] Then, a decision rule for removing the AP at the bad location can be determined based on the optimal weight parameter. For example, if the optimal weight parameter β i is non-zero, the corresponding AP i can be selected for positioning. Otherwise, if the optimal weight parameter β i is zero, the corresponding AP i can be determined to be removed for positioning.

[0082] In the above embodiment, the Gaussian function k(x, y) = exp(-γ||x - y|| 2 ) is used as the kernel function because it is a radial basis function. However, it should be noted that when formulating the sensor selection problem as a weight selection problem, many other types of kernel functions can be utilized.

[0083] Figure 11 and Figure 12 shows Figure 10 the position estimation results (i.e., black dots) in the shown environment. Four APs (i.e., AP1 - AP4) are deployed in the villa, as Figure 10 shown. A user with a target device 102 (such as a mobile phone) walks from the hall to the living room and then stops in the dining room. The received signal can be measured individually on each AP, such as CSI data and / or other data (such as information about the adjustment of the radio link). The measured data, i.e., CSI data and / or other data, can be periodically sampled, for example, once every 50 milliseconds. A positioning server (such as Figure 1 the positioning server 103 in) or one or more APs collect the measurement data from the APs and make a decision on sensor selection. The villa map can be stored in the positioning server or stored in one or more APs. First, the detectability of each AP can be predicted based on training with a large amount of CSI data.

[0084] Figure 11 is the estimated position and tracking result that is always based on the measurement data from all APs. It is found that the estimation error is large and there are estimation missing in some areas due to bad observations and lack of AP selection. The estimation based on the proposed sensor selection scheme is as Figure 12 shown. The results show that AP1 and AP3 are always in use, while AP2 is active when the mobile phone passes through the living room. AP4 is inactive at all times. Compared with the positioning results (i.e., black dots) shown in Figure 11 Figure 12 ​It is shown that the proposed sensor selection scheme can significantly improve the estimation accuracy.

[0085] Figure 13 It shows Figure 12 the measured cumulative error distribution of the results in. Due to the large errors (which show that there are no positioning results in a region), there is no quantitative analysis for Figure 11 Compared with Figure 11 the results in Figure 12 it shows a significant improvement in the positioning results by using sensor selection.

[0086] Now refer to Figure 14 which shows a simplified block diagram of the apparatus 1400 that can be implemented in an access node (e.g., the access node 101 shown in Figure 1 ) or a positioning server (e.g., the positioning server 103 shown in Figure 1 ) or implemented as an access node.

[0087] The apparatus 1400 may include at least one processor 1401, such as a data processor (DP) and at least one memory (MEM) 1402 coupled to the at least one processor 1401. The apparatus 1400 may further include one or more transmitters TX, one or more receivers RX 1403, or one or more transceivers coupled to the one or more processors 1401, which are related to wireless local communication network technologies (such as WLAN, UWB, ) and wireless telecommunication technologies such as 2 / 3 / 4 / 5 / 6G (generations) or any combination thereof. In addition, the apparatus 1400 may have one or more wired communication means that connect the apparatus to a computer cloud network or system, such as the network 110. The MEM 1402 stores a program (PROG) 1404. The PROG 1404 may include instructions that, when executed on the associated processor 1401, enable the apparatus 1400 to operate according to the embodiments of the present disclosure, such as executing one of the methods 300 and 400. The combination of the at least one processor 1401 and the at least one MEM 1402 may form a processing circuit system or module 1405 suitable for implementing various embodiments of the present disclosure.

[0088] Various embodiments of the present disclosure may be implemented by a computer program executable by one or more processors 1401, software, firmware, hardware, or any combination thereof.

[0089] The MEM 140 may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as semiconductor-based storage devices, magnetic storage devices and systems, optical storage devices and systems, fixed memories, and removable memories, as non-limiting examples.

[0090] The processor 1401 can be of any type suitable for the local technical environment and can include, by way of non-limiting example, one or more general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), and processors based on multi-core processor architectures.

[0091] In general, the various exemplary embodiments can be implemented using hardware or special-purpose circuits, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented using firmware or software executable by a controller, microprocessor, or other computing device, although the present invention is not limited thereto. Although the various aspects of the exemplary embodiments of the present invention can be illustrated and described as block diagrams, flowcharts, or using some other graphical representation, it is well understood that the block diagrams, apparatus, systems, techniques, or methods described herein can be implemented, by way of non-limiting example, in hardware, software, firmware, special-purpose circuits or logic, general-purpose hardware or controllers, or other computing devices, or some combination thereof.

[0092] Accordingly, it should be understood that at least some aspects of the exemplary embodiments of the present invention can be implemented in various components, such as integrated circuit chips and modules. Accordingly, it should be understood that the exemplary embodiments of the present invention can be implemented in a device embodied as an integrated circuit, where the integrated circuit can include circuitry (and possibly firmware) for embodying at least one or more of a data processor, a digital signal processor, a baseband circuitry, and a radio frequency circuitry, and the circuitry is configurable to operate in accordance with the exemplary embodiments of the present invention.

[0093] It should be understood that at least some aspects of the exemplary embodiments of the present invention can be embodied in computer-executable instructions executed by one or more computers or other devices, such as in one or more program modules. In general, program modules include routines, programs, objects, components, data structures, etc., which perform specific tasks or implement specific abstract data types when executed by a processor in a computer or other device. The computer-executable instructions can be stored on a computer-readable medium, such as a non-transitory computer-readable medium, such as a hard disk, an optical disk, a removable storage medium, a solid-state memory, a RAM, etc. As will be understood by those skilled in the art, in various embodiments, the functions of the program modules can be combined or distributed as needed. In addition, the functions can be embodied in whole or in part in firmware or hardware equivalents, such as integrated circuits, field-programmable gate arrays (FPGAs), etc.

[0094] The term "circuitry" as used in this application can refer to one or more or all of the following:

[0095] (a) Only hardware circuit implementations (such as implementations using only analog and / or digital circuitry), and

[0096] (b) A combination of hardware circuitry and software, such as (where applicable):

[0097] (i) A combination of (one or more) analog and / or digital hardware circuitry and software / firmware, and

[0098] (ii) Any portion of (one or more) hardware processors (including (one or more) digital signal processors) with software and (one or more) memories, which work together to enable a device, such as a mobile phone or a server, to perform various functions, and

[0099] (c) (One or more) hardware circuitry and / or (one or more) processors, such as (one or more) microprocessors or a portion of (one or more) microprocessors, which require software (e.g., firmware) to operate, but the software may be absent when not required to operate.

[0100] This definition of "circuitry" applies to all uses of the term in this application, including in any claims. As a further example, as used in this application, the term "circuitry" also encompasses implementations that are only hardware circuitry or processors (or multiple processors) or a portion of hardware circuitry or processors and their (or their) accompanying software and / or firmware. The term "circuitry" also encompasses, for example, and if applicable to a particular claim element, a baseband integrated circuit or a processor integrated circuit of a mobile device, or a similar integrated circuit in a server, a cellular network device, or other computing or network devices.

[0101] The present invention includes any novel feature or combination of features or any generalization thereof that is explicitly disclosed herein. When read in conjunction with the accompanying drawings, various modifications and adaptations of the foregoing exemplary embodiments of the present invention may become apparent to those skilled in the relevant art. However, any and all modifications will still fall within the scope of the non - limiting and exemplary embodiments of the present invention.

Claims

1. A method performed at a positioning server, comprising: Determining weights of at least two access points, wherein the weight of an access point is associated with the detectability of the access point for positioning a target device; For each of the at least two access points, constructing respective weighted kernel functions; For each of the at least two access points, calculating the inner product of the weight of the corresponding access point and the weighted kernel function constructed for the corresponding access point; Constructing an estimation error function by using the inner products for the at least two access points; Determining weight parameters of the corresponding weighted kernel functions to optimize the estimation error function with L1 regularization constraints; And Selecting, based on the determined weight parameters, one or more access points from the at least two access points to be used for position estimation of the target device.

2. The method according to claim 1, wherein, Determining the weights of the at least two access points further comprises: Receiving weights from the corresponding access points among the at least two access points.

3. The method according to claim 1, wherein, The selection further comprises: If the determined weight parameter of the weighted kernel function constructed for an access point is non-zero, determining that the access point will be used for the position estimation.

4. The method according to claim 1, wherein, The optimization of the estimation error function is further constrained by a distance range within which the target device may move.

5. The method according to claim 1, further comprising: Receiving channel state information of a radio link between the target device and a specific access point among the at least two access points; And Estimating the position of the target device by using the received channel state information.

6. The method according to claim 5, further comprising: Based on the channel state information, determining the detectability of the specific access point for positioning the target device; And Based on the determined detectability, determining whether the specific access point will be used for position estimation of the target device.

7. The method according to claim 5, wherein Determining whether the specific access point will be used for position estimation of the target device comprises: If the detectability is lower than a predetermined threshold, determining that the specific access point will not be used for position estimation of the target device.

8. The method according to claim 5, further comprising: Receiving information about an adjustment to a radio link between the target device and the specific access point, And wherein the determination of the detectability of the specific access point is further based on the adjustment.

9. The method according to claim 5, further comprising: Based on the determined detectability, determining the weight of the specific access point.

10. An apparatus at a positioning server, comprising: At least one processor; And At least one memory including computer program code, the memory and the computer program code being configured to, with the processor, cause the apparatus to at least: Determine weights of at least two access points, wherein the weight of an access point is associated with the detectability of the access point for positioning a target device; For each of the at least two access points, construct respective weighted kernel functions; For each of the at least two access points, calculate the inner product of the weight of the corresponding access point and the weighted kernel function constructed for the corresponding access point; Construct an estimation error function by using the inner product of the at least two access points; Determine the weight parameters of the corresponding weighted kernel function to optimize the estimation error function with the constraint condition of L1 regularization; And According to the determined weight parameters, select one or more access points from the at least two access points to be used for the position estimation of the target device.

Images