Abstract: Some embodiments include a method and apparatus which obtains sensor data associated with a vehicle. The sensor data includes three-dimensional acceleration data and three-dimensional rotation data, with respect to a sensor coordinate system. A gravity vector is obtained, and it is determined when the vehicle starts moving. Acceleration direction is used as a forward direction of the vehicle. It is determined from the rotation data when the vehicle changes direction, the rotation data indicating two different directions. The gravity vector is used to distinguish which is an upward direction by selecting directions the direction which has larger angle with respect to the gravity vector as the upward direction. The forward direction and upward direction are used to determine a rightward direction. The forward direction, upward direction and rightward direction represent the vehicle coordinate system. The orientation of the apparatus with respect to the orientation of the vehicle is determined.

Abstract: Respective location data of a plurality of users are distributed into respective buckets for each of the users. Each of the location data indicates a location of the respective user, and each of the buckets includes a respective time interval. The time intervals cover the time period, and there is a bucket for each of the users for each of the time intervals. For each pair of the users and for each of the time intervals, a respective sub-distance is calculated between the users of the respective pair based on the location data. For each of the pairs of the users, the sub-distances are aggregated over the time period to obtain a distance between the users of the respective pair. The users having a closest distance from each other are clustered.

Abstract: Respective location data of a plurality of users are distributed into respective buckets for each of the users. Each of the location data indicates a location of the respective user, and each of the buckets includes a respective time interval. The time intervals cover the time period, and there is a bucket for each of the users for each of the time intervals. For each pair of the users and for each of the time intervals, a respective sub-distance is calculated between the users of the respective pair based on the location data. For each of the pairs of the users, the sub-distances are aggregated over the time period to obtain a distance between the users of the respective pair. The users having a closest distance from each other are clustered.