In terms of speed, we use the speed value returned by the GPS
sensor when it is available; this is more accurate than calculating
the speed from consecutive GPS location points [4]. Otherwise, if
the direct speed is not available, it can be computed from
consecutive location changes. For a sequence of GPS reports we
compute the average speed. This feature has been used in many
existing works [2, 4, 18].
Let {p1, p2, p3, p4…pn} be a finite set of GPS sensor reports
submitted within a time window.
Average speed = (∑i=1 to n piv) / n (2)
where piv is the current ground speed obtained from the GPS
sensor report.