Problems with and ideas for improving voice announce of GPX Waypoints

[bhugh]

I have been experimenting with creating GPX bicycle routes with various waypoints that are things like historical markers, buildings, and places along the route. So they are places you will want to know about or stop at as you are cycling along the route.

The OsmAnd capability to audibly announce GPX waypoints as you are traveling the route is ideal for this, and OsmAnd is the only navigation program I know of that will do this.

I posted this on the Reddit forum and it looks like at least a few people thought the issue worth looking at.

#1. Very often the GPX Waypoint is announced repeatedly, like 6-8 times, one right after the other. Sometimes up to perhaps 20 times? These seem to be points that are somewhat distant from the current location, so I wonder if it is just crossing/recrossing the radius that triggers the announcement repeatedly.

Today on a ride I counted 17 announcements for the same GPX Waypoint in a row. This time I was pretty much right on top of the GPX Waypoint location.

It seems to do this the most when you are just on the edge of the range where it will announce GPX Waypoints (say, between 0.5 and 1.0 miles away from the GPX Waypoint) OR exactly underneath the GPX Waypoint (say, less than 20 meters from the GPX Waypoint center).

#2. Some GPX Waypoints that are, in fact, immediately adjacent to the route are never announced at all.

#3. It will announce waypoints in the form of "You are passing Waypoint X" long, long before you are actually passing it. For example, nearly a mile away. Then when you are ACTUALLY passing the waypoint, it says nothing at all.

#4. Sometimes it will again announced you are "passing" the waypoint, but again you are a good 1/4 mile away and nowhere near actually passing it. And when you actually do pass it, nothing.

Possible solutions:

#1. Ensure that some announcement is made for every GPX Waypoint you come close to.

#2. Ensure each announcement is read only once (or some limit like once per 120 seconds at most)

#3. Allow user adjustment or other context-sensitive adjustment of trigger radius for the "getting near" and "passing" announcements. For bicycle purposes those radii are way, way too large.

#3. Would it be too hard to actually calculate the closest approach distance between a route and a GPX Waypoint and then use the point of closest approach as the basis for the announcement triggers, rather than simple distance from the GPX Waypoint? There are only a limited number of waypoints in any GPX route and the calculation would only have to be done once when the route is loaded. Then the announcement for "passing" the GPX Waypoint could be given in relation to the point where you are indeed, actually passing that paypoint.

#4. If that is too difficult, at least change the message to indicate "GPX Waypoint X is 500 meters distant at heading 328 degrees)" or similar, rather than vaguely saying "PASSING" the GPX waypoint when in fact it is half a mile straight ahead or at some other random position, where you are not necessarily passing it at all.

#5. Also do give an indication when actually "passing" close to the GPX waypoint (which would probably be like 75 meters or closer) if you do indeed pass it.

#6. On the other hand, don't keep giving the 250 meter and 700 meter announcements repeatedly when you are actually less than 50 meters from the GPX Waypoint

#7. Since the GPX Waypoint audible announcements can be pretty verbose, allow the option for the user to enable or disable the 1) distant approach, 1) getting closer, and 3) passing very close announcements, and/or set the distance radius for each, and/or make them more context-sensitive to speed or trip type (driving vs cycling vs walking etc).

I'm using OsmAnd 3.4.5, Android 5.1.1 (though my wife's phone had the same behavior and is Android 8.x).

I tried running some simulated routes and didn't get any wrong GPX Waypoint announcements there (aside from the GPX Waypoints being announced a very long distance away from where the actual Waypoint is--but no double, triple, etc announcements). I don't know if that is because there is something different about running it live, or because my simulated routes were insufficient to identify the problem.

I've attached a sample GPX file of the type I'm using. It is a kind of historical tour, so it has a fair number of waypoints (that is the point of it).

I took a gander at the source code but couldn't spot the exact problems right away. I could possibly spend a little time on this over the next several months if that would help.

SFE0301PGpxWithWaypoints201090803.zip