Sorry it’s been very quiet here. I have written 5 posts recently but never posted them as they were so uninteresting to write, there’s no way they’d have been interesting to read. So instead I’ve been quietly refining the autopilot code for GPS waypoint tracking. Now it’s time to start testing. First step, passively record pre-flight waypoints to define where the flight should go. Here’s what I got today morning: 5 waypoints with 3 readings taken at each:
On the plus side, it’s consistent: like I said, each point on the graph is three or four GPS samples and they are identical.
On the down side, it’s inaccurate: the first four waypoints are plausible in direction / distance / location, but the last one’s direction and distance are very wrong: it should be back at the starting point yet it’s 8 meters away on the graph. What’s not shown on the graph is the parked car between the first and fourth point which might have affected it due to satellite signals bouncing off it? So with the car moved, the test was repeated:
This was a similar, not identical test. Each point had 3 GPS samples and show excellent consistency again. However, comparing the distances / directions between the two test graphs…
- step one is believable in its own right and in comparison with the previous test distance, but they disagree on direction; I suspect both are wrong, and reality lies somewhere in between.
- step two is believable in its own right and a very close match with the previous test
- step three seems to only have a passing resemblance in its own right and with comparison with the previous test – I suspect test one’s direction / distance is rearer reality; it’s worth noting that point 4 is nearest to the car
- step fours distance is reasonable, if perhaps a little long, but the significant disagreement on step three meant it too finished a long way from the starting point – 6.6 meters.
So what does this mean? Fundamentally, if this is the best possible, then GPS waypoint tracking is too inaccurate to run on my drive and without a new test location, the GPS tracking phase is permanently blocked. However, the similarity is close enough that I’ll do a few more tests removing all obstacles and hopefully they will provide results which are good enough to proceed.