Wireless device and method for positioning wireless device

Abstract

The present invention provides a wireless device and a method for locating the wireless device. Implementation of embodiments of the present invention can improve the accuracy of position determination. In one embodiment, the wireless device receives data from multiple satellites, and the method includes: identifying a neural network model that has been trained to adapt to the region in which the wireless device operates; using the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels; identifying satellite raw measurements with a quality level above a threshold; and calculating the position of the wireless device using the identified satellite raw measurements.

Description

Technical Field

[0001] Embodiments of the present invention relate to positioning methods using raw satellite measurements; more particularly, to wireless devices using neural networks to improve position determination. Background Art

[0002] Using satellite data to determine a wireless device's location in urban areas can be complex. One of the main sources of interference with satellite signals is multipath propagation, where satellite signals are reflected, refracted, and / or absorbed by buildings, foliage, and other objects along their path. This phenomenon causes the satellite signal to take multiple paths. Furthermore, transmission conditions can vary from city to city.

[0003] Most traditional designs use a Kalman filter in satellite positioning receivers. The filter performs smoothing, estimation, and prediction to minimize the least-square error. More specifically, the filter's goal is to minimize the sum of squared residuals (i.e., the difference between the measured values ​​and the fitted values ​​provided by the model). Assuming a linear time-invariant (LTI) system with Gaussian noise, the results obtained from the Kalman filter can be considered optimized. However, when performing positioning for urban scenarios, a major challenge is to continuously adjust the Kalman filter parameters to adapt to the operating environment.

[0004] Therefore, there is a need to improve traditional satellite-based positioning methods. Summary of the Invention

[0005] The present invention provides a wireless device and a method for positioning the wireless device. Implementation of the embodiments of the present invention can improve the accuracy of position determination.

[0006] In one embodiment, a wireless device receives data from multiple satellites, and the present invention provides a method for locating the wireless device, including: identifying a neural network model that has been trained to adapt to the area in which the wireless device operates; using the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels; identifying satellite raw measurements with a quality level above a threshold; and calculating the position of the wireless device using the identified satellite raw measurements.

[0007] In one embodiment, the present invention provides a wireless device that may include: a satellite receiver for receiving data from multiple satellites; a memory for storing one or more neural network models; and a processing circuit, wherein the processing circuit is operative to perform the following steps: identifying a neural network model that has been trained to adapt to the area in which the wireless device operates; using the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels; identifying satellite raw measurements with a quality level above a threshold; and calculating the position of the wireless device using the identified satellite raw measurements.

[0008] As described above, embodiments of the present invention use the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels, identify satellite raw measurements with a quality level above a threshold, and calculate the position of the wireless device using the identified satellite raw measurements, thereby improving the accuracy of the position determination of the wireless device. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 FIG. 1 is a schematic diagram of an environment according to an embodiment of the present invention.

[0010] Figure 2 An overview of a position determination apparatus and method according to one embodiment of the present invention is shown.

[0011] Figure 3 3 is a flow chart of a method 300 for training a region-adaptive neural network model according to an embodiment of the present invention.

[0012] Figure 4 FIG. 4 is a flow chart of a method 400 for model updating according to an embodiment of the present invention.

[0013] Figure 5 is a flowchart of a method 500 for performing position determination according to an embodiment of the present invention. DETAILED DESCRIPTION

[0014] Certain terms are used throughout the specification and claims to refer to specific components. Those skilled in the art will understand that hardware manufacturers may use different terms to refer to the same component. This specification and claims do not distinguish components based on name, but rather on functional differences. Throughout the specification and claims, the terms "including" and "comprising" are open-ended and should be interpreted as meaning "including, but not limited to." "Substantially" or "approximately" means that, within an acceptable range of error, a person skilled in the art can solve the stated technical problem and achieve the stated technical effect. Furthermore, the term "coupled" or "coupled" encompasses any direct and indirect electrical connection. Therefore, if a first device is described as being coupled to a second device, this means that the first device can be directly electrically connected to the second device or indirectly electrically connected to the second device via other devices or connections. The following describes preferred embodiments of the present invention and is intended to illustrate the spirit of the invention rather than to limit its scope, which shall be determined by the appended claims.

[0015] The following description is intended to be the preferred embodiment of the present invention. These descriptions are intended to illustrate the general principles of the present invention and should not be used to limit the present invention. The scope of protection of the present invention should be determined based on the claims of the present invention.

[0016] Embodiments of the present invention improve position determination by using raw satellite measurements classified by a region-adapted neural network. The region-adapted neural network is trained to adapt to different geographic regions, such as dense urban areas and urban canyons. A machine learning-based approach uses this neural network to address the severe multipath error problem in urban environments. Combining the region-adapted neural network with a Kalman filter-based algorithm improves position determination accuracy.

[0017] The region-adaptive neural network is a learning-based model that learns to classify satellite raw measurements for each satellite at each timestamp in the region of interest. The wireless device obtains satellite raw measurements from those satellites that are available and / or trackable to the wireless device. The satellite raw measurements are classified into two or more categories with different quality levels. The region-adaptive neural network model can be trained in the corresponding region and over a sustained period of time (e.g., 12 hours or longer). The trained model (including the weights of each neuron in the model) can be preloaded or downloaded to the wireless device to perform real-time position determination.

[0018] In this document, the term "satellite raw measurements" includes observation data from a device receiver, broadcast orbit information from a satellite, and supporting data (such as meteorological parameters collected from co-located instruments). For example, satellite raw measurements may include the satellite system's identifier, signal-to-noise ratio (SNR), or another signal-to-noise indicator calculated by the device, Doppler information, elevation angle, azimuth angle, clock information, etc. The satellite may be part of a constellation of a Global Navigation Satellite System (GNSS), such as the U.S. Global Positioning System (GPS), the Russian Global Navigation Satellite System (GLONASS), the Chinese BeiDou Navigation Satellite System (BDS), the European Union's Galileo, etc. In this disclosure, the terms "location" and "position" may be used interchangeably.

[0019] A wireless device can calculate a pseudo-range based on raw satellite measurements from multiple satellites. The pseudo-range between the receiver and the satellite is calculated by multiplying the speed of light by the difference between the reception time (expressed in the receiver's time frame) and the transmission time (expressed in the satellite's time frame) of the satellite signal. This pseudo-range corresponds to the distance from the receiver to the satellite, which includes the clock offset value of the receiver and the satellite and other offsets and errors (for example, ionospheric delay error, tropospheric delay error, and multipath propagation error). Typically, a device can determine its position based on the pseudo-ranges of four or more satellites. Due to the large number of satellites deployed in space at the same time, a device may be able to receive satellite signals from more than four satellites (for example, ten or more satellites) at the same time.

[0020] Figure 1 Figure 1 illustrates an environment in which a wireless device 100 with a local adaptive neural network can operate, according to one embodiment of the present invention. In this embodiment, the wireless device 100 receives satellite signals from a plurality of satellites, five of which are shown as satellites 115A-115E (collectively, satellites 115). The received satellite signals include errors due to atmospheric delays, such as ionospheric delay and tropospheric delay. Furthermore, the received satellite signals also include multipath propagation errors (e.g., the signal received from satellite 115E).

[0021] Figure 1The wireless device 100 is also shown to include an antenna 110 coupled to a radio frequency (RF) circuit 120, which also includes a satellite receiver configured to receive satellite signals. The wireless device 100 also includes a processing circuit 130, which may further include one or more programmable processors (e.g., a central processing unit (CPU), a digital signal processor (DSP), a graphics processing unit (GPU), an accelerated processing unit (APU), a neural processing unit (NPU), etc.). The processing circuit 130 may also include a dedicated circuit, such as an application-specific integrated circuit (ASIC). The processing circuit 130 is coupled to a memory 140. The memory 140 may include memory devices such as dynamic random access memory (DRAM), static RAM (SRAM) and / or other volatile and non-volatile memories. It will be understood that Figure 1 The embodiments shown are simplified for illustrative purposes and the present invention may include additional hardware components.

[0022] In one embodiment, memory 140 may store a positioning algorithm 160 and one or more neural network models 170. In one embodiment, neural network model 170 is a region-adaptive neural network model that includes weights (i.e., the weights of each neuron in the model) trained using satellite data collected in an area of ​​interest (e.g., the location of wireless device 100). In one embodiment, neural network model 170 may be implemented as a multi-layer perceptron network. Positioning algorithm 160 may be based on a known algorithm, such as a Kalman filter algorithm. Additionally or alternatively, positioning algorithm 160 may include at least one of a least mean square algorithm, a weighted least mean square algorithm, and a closed-form algorithm. Neural network model 170 and positioning algorithm 160 may be implemented by processing circuit 130.

[0023] Figure 2 According to an embodiment of the present invention, an overview of a position determination device and method is shown. The position determination method is performed by a device 200, which is Figure 11 is an example of a wireless device 100 in FIG. Device 200 includes a location detector 250, a model detector 260, a region-adaptive neural network model 270, and a positioning engine 280. The region-adaptive neural network model 270 is stored in a memory and has been trained for the region in which the device 200 operates. The location detector 250, the model detector 260, and the positioning engine 280 may be implemented by hardware circuitry, software, or a combination of hardware and software.

[0024] In step 201, the device 200 obtains satellite raw measurements for each tracked satellite at a given time. In step 202, the device 200 executes the region-adaptive neural network model 270 to classify the satellite raw measurements into two or more quality levels. The raw measurements of each satellite are classified into a corresponding quality level. The device 200 may discard, disregard, or ignore measurements with a quality level below a threshold; for example, measurements with the lowest quality level. In step 203, the positioning engine 280 processes the measurements with a quality level exceeding the threshold to output the device's location, such as latitude and longitude. The positioning engine 280 may execute Figure 1 200. For example, when positioning engine 280 executes a Kalman filter algorithm to calculate the position of device 200, device 200 may set a weight in the Kalman filter for each satellite's raw satellite measurement value based on the quality level in step 202 or 203. Therefore, in step 203, positioning engine 208 may first obtain an initial positioning result using measurements with a quality level exceeding a threshold, and then execute the Kalman filter algorithm on the initial positioning result based on the weights in the Kalman filter assigned to the measurements with a quality level exceeding the threshold to calculate the position of device 200.

[0025] The location detector 250 monitors the output of the positioning engine 280 and detects changes in the device's location. When a new device location is detected, in step 204, the model detector 260 determines whether the stored region-adapted neural network model 270 needs to be updated; for example, when the region-adapted neural network model 270 is no longer applicable to the new location and the model adapted to the new location is not stored in the device 200, it indicates that it needs to be updated. If an update is required, in step 205, the model detector 260 requests the location server 290 to download a new neural network model adapted to the new location of the device 200 via a wireless network (e.g., the Internet). After the model detector 260 receives the downloaded neural network model corresponding to the new location, it uses the downloaded model to update the stored region-adapted neural network model 270. The downloaded model includes weights (i.e., the weights of each neuron in the model) trained using training data collected in the new location or in the area including the new location.

[0026] In some embodiments, device 200 may store multiple region-adaptive neural network models in memory, for example, storing N neural networks for N cities. Each of the N neural networks may have different types of multipath propagation characteristics and / or different types of atmospheric characteristics. Device 200 first estimates its location, such as which city it is located in, and then selects a stored neural network model (e.g., neural network model 270) for that city. After selecting neural network model 270, device 200 may begin position determination by performing steps 201-203 or steps 201-206 described above.

[0027] In one embodiment, neural network model 270 can be trained and used for positioning in indoor environments. Device 200 can combine satellite raw measurements with data from other wireless systems to perform position determination. For example, these other systems may include cellular networks, Wi-Fi networks, Bluetooth networks, and the like. Data from these other invisible systems may include cell IDs, signal strength from base stations or other wireless system access points, channel state information, and the like. In one embodiment, device 200 can receive terrain and building information from a data collection system. The device can use the terrain or building information to estimate multipath effects on satellite signals and estimate the quality level of satellite raw measurements from a given satellite at a given time. For example, if building information indicates that a given satellite is blocked by a building at a given time, device 200 will ignore data obtained from the given satellite at that time based on this building information. In one embodiment, building information can be incorporated into neural network model 270 during the training phase (i.e., the building information and satellite raw measurements are used as inputs to neural network model 270). Thus, device 200 can classify satellite raw measurements based at least in part on the building information. Additionally or alternatively, as part of the operation of the positioning engine 280, building information can also be applied to the classified satellite raw measurements (i.e., the building information is provided to the classified satellite raw measurements at the output of the neural network model 270 to calibrate the output of the neural network model 270). For example, the positioning engine 280 can set the weights of the Kalman filter based at least in part on the building information, such that satellite raw measurements that are subject to multipath effects are given lower weights in the Kalman filter. Thus, the device 200 can calculate its position using the satellite raw measurements classified as having a high quality rating by the neural network model 270 and the building information added to the Kalman filter.

[0028] In one embodiment, device 200 may receive a pre-loaded neural network model prior to shipment. In one embodiment, the pre-loaded neural network model may not receive updates during the lifecycle of the device or may receive updates over the lifecycle of the device. Model updates may be periodic and may not be related to the location of the device, or may occur when the device changes its location.

[0029] Figure 33 is a flow chart of a method 300 for training a regional adaptive neural network model according to an embodiment of the present invention. The method 300 can be performed by a computing and / or communication system including a satellite receiver. In step 310, the system identifies a target area for training the neural network. The system is located in the target area to collect training data. In addition, at each data collection stage (i.e., a moment in time), the pseudorange between the system and each tracked satellite is known. The target area can be a geographic area described by a center point (latitude and longitude) and a diameter. The target area can be characterized by its location type, such as a dense urban area, an urban canyon, an area of ​​dense foliage, etc. In step 320, for each satellite tracked by the system, the system collects and records satellite raw measurements with a timestamp over a continuous period of time (e.g., at least 12 hours).

[0030] In step 330, the system calculates a pseudorange for each tracked satellite using the timestamp and the clock time in the satellite raw measurement, and calculates a pseudorange error (pseudorange error) that includes at least one multipath propagation error from the satellite raw measurement for a given satellite at a given time. The pseudorange error is the difference between a known pseudorange and the corresponding pseudorange calculated for a given satellite with a given timestamp. This difference is compared to one or more predetermined thresholds to determine the quality level of the satellite raw measurement for the given satellite. For example, if the difference is 10 meters and the threshold is 30 meters, the satellite raw measurement for the given satellite is classified as good quality. If the difference exceeds 30 meters, the satellite raw measurement for the given satellite is classified as poor quality and may be ignored. In scenarios where more than two quality levels exist, multiple thresholds may be predefined to classify the satellite raw measurements into multiple quality levels.

[0031] In step 340, the system labels each set of satellite raw measurements (for one satellite and one timestamp) with a quality rating based on the pseudorange error to generate a training set. In step 350, the system trains a region-adapted neural network model for each target region using the satellite raw data (i.e., satellite raw measurements) as input and the quality rating as output. The trained neural network model can be downloaded to a wireless device (e.g., Figure 1 Device 100 and Figure 2 Device 200 in is used for location determination.

[0032] In one embodiment, the first system may perform data collection steps 310 and 320 and the second system may perform training steps 330 to 350. For example, the first system may collect and record raw satellite measurements and provide these measurements to the second system for relevant training.

[0033] Figure 4 4 is a flow chart of a method 400 for model updating according to an embodiment of the present invention. In step 410, the device estimates the current position. The estimate can be based on Kalman filter positioning without the benefits of a neural network (i.e., the estimate does not use a neural network model). Alternatively, the estimate can be based on information from the user (e.g., in which city the user is located) or information from another wireless system (e.g., a cellular system). In step 420, the device determines whether a neural network model corresponding to the estimated position is already stored in the device memory. If the corresponding neural network model is already stored in the device, in step 430, the device uses the stored model to process the satellite raw measurements, for example, by classifying the satellite raw measurements into two or more quality level categories. In step 440, the device determines its current position using the satellite raw measurements classified as having a high quality level (e.g., exceeding a quality level threshold).

[0034] If the corresponding neural network model is not stored in the device, then at step 450 the device downloads the model (including the weights of the model (i.e., the weights of each neuron in the model)) from the cloud server. At step 460, the device performs a model update, for example, replacing the stored model with the downloaded model. At step 470, the device processes the satellite raw measurements using the downloaded model, for example, by classifying the satellite raw measurements into two or more quality level categories. At step 480, the device determines its current position using the satellite raw measurements classified as having a high quality level (e.g., exceeding a quality level threshold).

[0035] Figure 5 FIG2 is a flow chart illustrating a method 500 for determining position according to an embodiment of the present invention. In step 510, the device identifies a neural network model that has been trained to adapt to the region in which the wireless device is operating. In step 520, the device uses the neural network model to classify raw satellite measurements from each satellite at a given time into corresponding quality levels. In step 530, the device identifies raw satellite measurements with a quality level above a threshold. In step 540, the device calculates the position of the wireless device using the identified raw satellite measurements.

[0036] In one embodiment, a neural network is trained to classify satellite raw measurements based on pseudorange errors calculated during a training phase, the pseudorange errors including at least multipath propagation errors. The neural network model may be a multilayer perceptron network. The satellite raw measurements may include at least one of the following satellite raw measurements: satellite system identity, signal-to-noise ratio (SNR), Doppler information, elevation angle, azimuth angle, and clock information. The neural network model may include input nodes in an input layer, each of which receives one of the satellite raw measurements.

[0037] In one embodiment, the device may ignore raw satellite measurements with a quality level below a threshold. The device may estimate its location to determine whether a neural network model stored in the device memory corresponds to the estimated location. When the neural network model stored in the device memory does not correspond to the estimated location, the device may download a neural network model corresponding to the estimated location from a server.

[0038] In one embodiment, when classifying satellite raw measurements, the device may use building information to estimate multipath effects on information from a given satellite at a given time, and classify the satellite raw measurements based at least in part on the building information. In another embodiment, when calculating the location of a wireless device, the device may use location information to estimate multipath effects on information from a given satellite at a given time, and calculate the location of the wireless device using the identified satellite raw measurements and the building information. The identified satellite raw measurements are values ​​classified as having a high quality rating. Additionally, when calculating the location of the wireless device, the device may use information from one or more other wireless systems, including one or more of a cellular system, a Wi-Fi system, and a Bluetooth system.

[0039] In one experiment, a five-layer multilayer perceptron with 46,000 neurons was used as a neural network model to classify the quality of satellite raw measurements. The quality classification was used to weight the satellite raw measurements (i.e., satellite raw measurements with high quality levels were assigned high weights (e.g., high weights in the Kalman filter), while satellite raw measurements with low quality levels were assigned low weights (e.g., low weights in the Kalman filter) or ignored). The satellite raw measurements were refined according to the weights and used as input to the Kalman filter. The complexity of the neural network was kept small enough to accommodate mobile devices with limited storage and computation. By using an accelerator to speed up the execution of the neural network, positioning performance for a typical urban case with 15 available satellites was achieved with 4.5 milliseconds and 2.7 megabytes of memory access, meeting the real-time requirements of the positioning system. Furthermore, experimental evaluations showed significant improvements in positioning accuracy compared to a standard Kalman filter in urban areas. Thus, by combining a region-adaptive neural network with a positioning algorithm such as a Kalman filter-based approach, the disclosed position determination method is shown to mitigate the effects caused by corrupted measurements such as multipath errors and atmospheric errors in challenging urban areas.

[0040] Already referenced Figure 1 and 2 The exemplary embodiments describe Figure 3-5 However, it should be understood that the Figure 1and 2 Embodiments of the present invention other than the embodiments of Figure 3-5 The operation of the flowchart, Figure 1 and 2 Embodiments of the invention may perform operations different from those discussed with reference to the flowcharts. Figure 3-5 The flowcharts show a particular order of operations performed by certain embodiments of the present invention, but it should be understood that this order is exemplary (e.g., alternative embodiments may perform operations in a different order, combine certain operations, overlap certain operations, etc.).

[0041] Various functional components, blocks, or modules have been described herein. As will be understood by one of ordinary skill in the art, the functional blocks or modules may be implemented by circuits (either dedicated circuits or general-purpose circuits operating under the control of one or more processors and coded instructions), which typically include transistors configured to control the operation of the circuits according to the functions and operations described herein.

[0042] Although the present invention is disclosed above with reference to preferred embodiments, it is not intended to limit the scope of the invention. Anyone with ordinary knowledge in the art may make slight changes and modifications without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the scope of the claims.

Claims

1. A method for locating a wireless device, wherein the wireless device receives data from a plurality of satellites, characterized in that: The method includes: identifying a neural network model that has been trained to be adapted to the region in which the wireless device operates; Using the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels; Identifying satellite raw measurements with a quality level above a threshold; calculating a position of the wireless device using raw measurements of the identified satellites; The method further includes: estimating a location of the wireless device to determine whether a neural network model corresponding to the estimated location is already stored in a memory of the wireless device; and When the neural network model stored in the memory of the wireless device does not correspond to the estimated location, the neural network model corresponding to the estimated location is downloaded from the cloud server.

2. The method for locating a wireless device according to claim 1, wherein: The neural network is trained to classify the satellite raw measurements for each satellite according to the pseudorange errors calculated during the training phase, the pseudorange errors including at least multipath propagation errors.

3. The method for locating a wireless device according to claim 1, wherein: The neural network model is a multi-layer perceptron network.

4. The method for locating a wireless device according to claim 1, wherein: The satellite raw measurement value of each satellite includes at least one of the following satellite raw measurement values: satellite system identity, signal-to-noise ratio, Doppler information, elevation angle, azimuth angle, and clock information.

5. The method for locating a wireless device according to claim 4, wherein: The neural network model includes input nodes located in an input layer, and each of the input nodes receives a satellite raw measurement value.

6. The method for locating a wireless device according to claim 1, wherein: Also includes: Ignore satellite raw measurements with a quality level below a threshold.

7. The method for locating a wireless device according to claim 1, wherein: Also includes: Incorporating building information to estimate the multipath effect on the signal from a given satellite at a given time; The satellite raw measurements for each satellite are classified based at least in part on the building information.

8. The method for locating a wireless device according to claim 1, wherein: Also includes: Incorporating building information to estimate the multipath effect on the signal from a given satellite at a given time; The location of the wireless device is calculated based on the identified satellite raw measurements and the building information.

9. The method for locating a wireless device according to claim 1, wherein: Also includes: incorporating information from one or more other wireless systems in calculating the location of the wireless device; The one or more other systems include one or more of a cellular system, a Wi-Fi system, and a Bluetooth system.

10. The method for positioning a wireless device according to claim 1, wherein: Calculating the position of the wireless device using the identified satellite raw measurements includes executing a positioning algorithm, wherein the positioning algorithm includes at least one of a Kalman filter algorithm, a least mean square algorithm, a weighted least mean square algorithm, and a closed form algorithm.

11. The method for positioning a wireless device according to claim 10, wherein: When the positioning algorithm is a Kalman filter algorithm, the method further includes: Setting a weight in a Kalman filter for the satellite raw measurements of each satellite according to the classified quality level; and Calculating the position of the wireless device using the identified satellite raw measurements includes: Calculate the initial positioning result by using the raw measurement value of the identified satellite; A Kalman filter algorithm is executed on the initial positioning result according to the weight of the identified satellite raw measurement value in the Kalman filter to calculate the position of the wireless device.

12. A wireless device, characterized in that: include: a satellite receiver for receiving data from a plurality of satellites; A memory for storing one or more neural network models; and Processing circuitry, wherein the processing circuitry is operable to perform the following steps: identifying a neural network model that has been trained to be adapted to the region in which the wireless device operates; Using the neural network model to classify satellite raw measurements from each satellite at a given time into corresponding quality levels; Identifying satellite raw measurements with a quality level above a threshold; calculating a position of the wireless device using raw measurements of the identified satellites; The processing circuit also performs: estimating a location of the wireless device to determine whether a neural network model corresponding to the estimated location is already stored in a memory of the wireless device; and When the neural network model stored in the memory of the wireless device does not correspond to the estimated location, the neural network model corresponding to the estimated location is downloaded from the cloud server.

13. The wireless device according to claim 12, wherein The neural network is trained to classify the satellite raw measurements for each satellite according to the pseudorange errors calculated during the training phase, the pseudorange errors including at least multipath propagation errors.

14. The wireless device of claim 12, wherein: The neural network model is a multi-layer perceptron network.

15. The wireless device of claim 12, wherein: The satellite raw measurement value of each satellite includes at least one of the following satellite raw measurement values: satellite system identity, signal-to-noise ratio, Doppler information, elevation angle, azimuth angle, and clock information.

16. The wireless device of claim 15, wherein: The neural network model includes input nodes located in an input layer, and each of the input nodes receives a satellite raw measurement value.

17. The wireless device of claim 12, wherein: When identifying satellite raw measurements with a quality level above the threshold, the processing circuit further performs: ignoring satellite raw measurements with a quality level below the threshold.

18. The wireless device of claim 12, wherein: When using the neural network model to classify the satellite raw measurements from each satellite at a given time into corresponding quality classes, the processing circuitry further performs: Incorporating building information to estimate the multipath effect on the signal from a given satellite at a given time; The satellite raw measurements for each satellite are classified based at least in part on the building information.

19. The wireless device of claim 12, wherein: When calculating the position of the wireless device using the identified satellite raw measurements, the processing circuitry further performs: Incorporating building information to estimate the multipath effect on the signal from a given satellite at a given time; The location of the wireless device is calculated based on the identified satellite raw measurements and the building information.

20. The wireless device of claim 12, wherein When calculating the position of the wireless device using the identified satellite raw measurements, the processing circuitry further performs: incorporating information from one or more other wireless systems in calculating the location of the wireless device; The one or more other systems include one or more of a cellular system, a Wi-Fi system, and a Bluetooth system.

21. The wireless device of claim 12, wherein: When calculating the position of the wireless device using the identified satellite raw measurements, the processor calculates the position of the wireless device in conjunction with a positioning algorithm, wherein the positioning algorithm includes at least one of a Kalman filter algorithm, a least mean square algorithm, a weighted least mean square algorithm, and a closed form algorithm.

22. The wireless device of claim 21, wherein When the positioning algorithm is a Kalman filter algorithm, the processor further executes: Setting weights in the Kalman filter for the satellite raw measurements of each satellite according to the classified quality level; as well as When calculating the position of the wireless device using the identified satellite raw measurement values, the processor calculates the position of the wireless device in conjunction with a positioning algorithm, including: Calculate the initial positioning result by using the raw measurement value of the identified satellite; A Kalman filter algorithm is executed on the initial positioning result according to the weight of the identified satellite raw measurement value in the Kalman filter to calculate the position of the wireless device.

Images