Waypoint Averaging


23 April 2009 @ 1:37 PM  / Trail Tech /


 
A common misconception of many outdoor GPS device users is they need to average every waypoint they mark.  However, modern advanced GPS chipsets do significant amounts of calculations and filtering before reporting a position, diminishing the benefits of short-term averaging.   For many waypoints (campsites, houses, trailheads, etc.), averaging is unnecessary because even in the most challenging GPS conditions, the waypoint will likely be within 30ft of the ground truth.  In some circumstances, such as saving a geocache location or marking a waypoint in dense tree cover, it is beneficial to gain additional confidence the marked location is as close to ground truth as possible.

We recently completed detailed research on various waypoint averaging techniques and their respective benefits.  Our research culminated in an improved multi-sample averaging technique which outperforms the traditional short-term averaging.  The first implementation offering the multi-sample technique is offered as a new application on our Oregon platform.

 

Traditional Short-term averaging involves averaging a location over a few minutes.   This technique removes large temporary errors in GPS position and is most beneficial when conditions are challenging (such as dense tree cover) but can not mitigate errors from the current satellite constellation.  For best results in difficult conditions, our research found short-term averaging should be done for at least 5 minutes, preferably 7 to 10 minutes.  Oregon’s application indicates how long it recommends averaging based on current conditions.

Multi-sample averaging involves returning multiple times to collect short-term average samples, mitigating errors from the current satellite constellation and allowing a waypoint to converge to ground truth as additional samples are added.  For multi-sample averaging, our research uncovered two important factors in achieving the greatest error reduction: the number of samples collected and the elapsed time between samples.  For optimal results, at least 4-8 samples should be collected spaced at least 90 minutes apart (allowing the satellite constellation sufficient time to change).