Abstract: An RF fingerprinting methodology is generalized to include non-RF related factors. For each fingerprinted tile, there is an associated distance function between two fingerprints (the training fingerprint and the test fingerprint) from within that tile which may be a linear or non-linear combination of the deltas between multiple factors of the two fingerprints. The distance function for each tile is derived from a training dataset corresponding to that specific tile, and optimized to minimize the total difference between real distances and predicted distances. Upon receipt of an inference request, a result is derived from a combination of the fingerprints from the training dataset having the least distance per application of the distance function. Likely error for the tile is also determined to ascertain whether to rely on other location methods.

Abstract: Embodiments calculate an estimated latency between computing devices. A latency service aggregates latency records defining latency measurements and corresponding latency factors from a plurality of computing devices. From the aggregated latency records, the latency service defines relationships between the latency measurements and the corresponding latency factors. Responsive to a request for an estimated latency from a mobile computing device, the latency service applies the defined relationships to estimate the latency based on the latency factors associated with the received request. In some embodiments, the estimated latency includes three portions: a first latency value representing the latency from the mobile computing device to a cell site, a second latency value representing the latency from the cell site to an access point, and a third latency value representing the latency from the access point to a destination computing device.

Abstract: Managing use of a location sensor on a computing device for energy efficiency. The location sensor is briefly initialized to measure the signal quality. The measured signal quality is compared to pre-defined signal criteria values. The signal criteria values correspond to acceptable energy consumption, for example. If the signal criteria values are satisfied, location information for the computing device is obtained. Otherwise, the location sensor is disabled without obtaining the location information. In some embodiments, a lower-energy location sensor is used to obtain location information to determine whether to enable a higher-energy location sensor based on expected energy consumption.

Abstract: Location inference using selected beacons. Data is received representing a set of beacons observed by a computing device. The beacons are located within a first geographic area. A subset (e.g., a clique) of the beacons is selected based on a coverage area of each of the beacons, where each of the beacons in the selected subset has a coverage area that overlaps with the coverage area of each of the other beacons in the selected subset. Using known or estimated positions of the beacons, a second geographic area is defined based on the selected subset of beacons and the beacon reference data and the coverage areas associated therewith. The second geographic area, smaller than the first geographic area, represents an approximate location of the computing device. In some embodiments, the computing device is calculated to be within the second geographic area with 95% probability.

Abstract: Embodiments adjust device error radiuses associated with inferred device positions produced by positioning systems. Inferred beacon positions and associated beacon radiuses are accessed for beacons in a beacon fingerprint from an observing computing device. The beacon radiuses are associated with a pre-defined confidence level (e.g., an in-circle percentage). A Kalman filter is applied to at least one of the beacons using the inferred beacon positions and the beacon radiuses associated therewith to infer a device position for the computing device and to compute a device error radius for the inferred device position. The computed device error radius is adjusted as a function of the quantity of beacons input to the Kalman filter to achieve the pre-defined confidence level.

Abstract: An RF fingerprinting methodology is generalized to include non-RF related factors. For each fingerprinted tile, there is an associated distance function between two fingerprints (the training fingerprint and the test fingerprint) from within that tile which may be a linear or non-linear combination of the deltas between multiple factors of the two fingerprints. The distance function for each tile is derived from a training dataset corresponding to that specific tile, and optimized to minimize the total difference between real distances and predicted distances. Upon receipt of an inference request, a result is derived from a combination of the fingerprints from the training dataset having the least distance per application of the distance function. Likely error for the tile is also determined to ascertain whether to rely on other location methods.

Abstract: Embodiments analyze crowd-sourced data to identify a moved or moving beacon. The crowd-sourced data involving a particular beacon is filtered based on a cluster start time associated with the beacon. A clustering analysis groups the filtered crowd-sourced data for the beacon into a plurality of clusters based on spatial distance. Timestamps associated with the crowd-sourced data in the clusters are compared to select one of the clusters. The crowd-sourced data associated with the selected cluster is used to determine position information for the moved beacon. The cluster start time for the beacon is adjusted based on the earliest timestamp associated with the positioned observations corresponding to the selected cluster. Adjusting the cluster start time removes from a subsequent analysis the positioned observations associated with one or more prior positions of the beacon.

Abstract: Embodiments calculate an estimated latency between computing devices. A latency service aggregates latency records defining latency measurements and corresponding latency factors from a plurality of computing devices. From the aggregated latency records, the latency service defines relationships between the latency measurements and the corresponding latency factors. Responsive to a request for an estimated latency from a mobile computing device, the latency service applies the defined relationships to estimate the latency based on the latency factors associated with the received request. In some embodiments, the estimated latency includes three portions: a first latency value representing the latency from the mobile computing device to a cell site, a second latency value representing the latency from the cell site to an access point, and a third latency value representing the latency from the access point to a destination computing device.

Abstract: Training datasets and test datasets consisting of observations (i.e., RSS measurements) partitioned per a mapping tile system are used to evaluate possible RSS weighting functions for each such tile. The observations from the training dataset are used to determine an optimal weighting function based on the training dataset that minimizes the error for the test data, wherein the error may be a function of the deltas between GPS positions of observations in the test dataset and predicted positions from the RSS weighted functions applied to test data. The accuracy of the optimal weighted function for each tile is characterized to determine whether to use the weighted function or an alternative (such as a non-weighted function) for subsequent inquiries.

Abstract: Determining semantics for locations based on user data such as user activities and/or user communications. The user data is captured and analyzed to identify keywords including locations. Location data describing the locations is obtained and associated with the user and keywords. The associations represent user context for the locations. The associations are used to deliver services and/or products to the user at various times such as when user approaches or enters a particular location.

Abstract: Determining geospatial patterns from device data collected from a plurality of computing devices. The devices represent, for example, a plurality of sources providing the device data. The device data describes the computing devices and/or environments thereof. Some embodiments present the determined patterns to users for editing, update maps with the edited patterns, and distribute the maps to the users. The maps are stored to create a searchable map library.

Abstract: Managing use of a location sensor on a computing device for energy efficiency. The location sensor is briefly initialized to measure the signal quality. The measured signal quality is compared to pre-defined signal criteria values. The signal criteria values correspond to acceptable energy consumption, for example. If the signal criteria values are satisfied, location information for the computing device is obtained. Otherwise, the location sensor is disabled without obtaining the location information. In some embodiments, a lower-energy location sensor is used to obtain location information to determine whether to enable a higher-energy location sensor based on expected energy consumption.

Abstract: Embodiments infer position information using captured content items. A computing device such as a mobile computing or a cloud service analyzes one or more content items to generate information describing the content items. For content items that include photographs, for example, the photographs are processed to identify faces, objects, places, text, etc. The generated information is compared to reference content items (e.g., image comparison), user data (e.g., calendar data, contact information, etc.), and/or public data (e.g., address information, event information, etc.) to identify at least one match. Location information associated with the matched reference content items, user data, and/or public data is obtained. A position is determined for the content items based on the location information obtained from the matched reference content item, user data, and/or public data.

Abstract: Embodiments enhance the functionality of a vehicle, a user device, or both by the selection and sharing of data. Upon detection of each other, the vehicle device and the user device obtain and share data. The data may be associated with the user, the user computing device, and/or the vehicle and may be stored in cloud-based services. Functionality of the vehicle and/or user device is customized to the user based on the shared data. For example, the user device may provide assisted global positioning system (GPS) data to the vehicle to reduce a time-to-fix (TTF) when determining a location of the vehicle. In other examples, settings of the vehicle are personalized to the user, and location-relevant content is downloaded to the user device.

Abstract: Embodiments respond to a position inference request from a computing device to determine a location of a computing device. The position inference request received from the computing device identifies a set of beacons observed by the computing device. A geographic area is estimated in which the computing device is located using the set of beacons. At least one location method is selected to identify a location of the computing device within the geographic area. In some cases two or more location methods may he employed and their results combined using, for example, a weighting function. The location of the computing device is determined within the geographic area using the set of beacons and the selected location method(s). The location that is determined is communicated to the computing device.

Abstract: Generate tags for content from metadata pre-filtered based on context. A plurality of data items is accessed. Each of the data items has metadata. A context for a user is determined (e.g., at a moment of content capture). One or more of the data items are selected based on the determined context. Upon receipt of content, the received content is compared with the selected data items to identify matches. Metadata is selected from the metadata associated with the matching data items. The selected metadata is associated with the captured content.

Abstract: Training datasets and test datasets consisting of observations (i.e., RSS measurements) partitioned per a mapping tile system are used to evaluate possible RSS weighting functions for each such tile. The observations from the training dataset are used to determine an optimal weighting function based on the training dataset that minimizes the error for the test data, wherein the error may be a function of the deltas between GPS positions of observations in the test dataset and predicted positions from the RSS weighted functions applied to test data. The accuracy of the optimal weighted function for each tile is characterized to determine whether to use the weighted function or an alternative (such as a non-weighted function) for subsequent inquiries.

Abstract: An RF fingerprinting methodology is generalized to include non-RF related factors. For each fingerprinted tile, there is an associated distance function between two fingerprints (the training fingerprint and the test fingerprint) from within that tile which may be a linear or non-linear combination of the deltas between multiple factors of the two fingerprints. The distance function for each tile is derived from a training dataset corresponding to that specific tile, and optimized to minimize the total difference between real distances and predicted distances. Upon receipt of an inference request, a result is derived from a combination of the fingerprints from the training dataset having the least distance per application of the distance function. Likely error for the tile is also determined to ascertain whether to rely on other location methods.

Abstract: Embodiments provide a position service experimentation system to enable comparison of modeling and inference methods as well as characterization of input datasets for correspondence to output analytics. Crowd-sourced positioned observations are divided into a training dataset and a test dataset. A beacons model is generated based on the training dataset, while device position estimations are calculated for the test dataset based on the beacons model. The device position estimations are compared to the known position of the computing devices generating the positioned observations to produce accuracy values. The accuracy values are assigned to particular geographic areas based on the position of the observing computing device and aggregated to enable a systematic analysis of the accuracy values based on geographic area and/or positioned observations characteristics.

Abstract: Embodiments order observed beacons based on relative signal strength to create a correspondence between beacon sets and positions. A computing device such as a mobile device provides a positioned observation including a plurality of observed beacons and a position of the mobile device during observation. The observed beacons are ordered based on quality indicators such as signal strength relative to each other. A set of the beacons are selected based on the ordering (e.g., the beacons with the strongest signal strength are selected in order). The position of the observing mobile device is associated with the beacon set to enable location inference for other devices providing observations including the same beacon set.