Thanks for the pointers, guys. Here are the settings that I have in the game:
1. The two dynamite actions on either side of the door are linked to each other. They work properly when the dynamite goes off whether planted on the inside or outside (while testing the game, I have personally set it off from outside the door since my bot engineers won't go there)
2. Route #0 at the Allied spawn points to the dynamite action (#4) and Route #1 at the Documents points to the extraction point (#32).
3. The alt roam Group ID numbers are all 0 and they also are set to be "active forever."
4. The route_team is set to Allied (#2).
I just now fired up the map and rechecked those settings to be sure!
Mal, your explanation of the route system was very clear to me and explains fully how the system works. Let's throw some math at it to see how picky the alt roam placement really is. Using Mal's example:
Let D1 be the short route from the bot to the objective (red to blue)
Let D2 be the longer route via the alt roam (red to green to blue)
Let f be the fraction of D2 that extends from red to green
There are two cases to consider:
CASE 1: the alt roam (green) is too close to the start (red).
The bot will travel to the alt roam then turn around and go back past the starting point and take the short way to blue. To keep this from happening, the following inequity must be satisfied
D1 + 2* f*D2 > (1-f) * D2
solving for f,
f > (D2 - D1) / 3*D2
CASE 2: the alt roam (green) is too close the the finish (blue). The bot will take the shortest route to the alt roam (which in this case is via the short route through the finish and on to the alt roam) then turn around and go back to the finish (blue). To keep this from happening, the following inequity must be satisfied
D1 + (1-f)*D2 > f*D2
solving for f,
f < (D1 + D2) / 2*D2
So we have the following conditions placed on f (it must lie somewhere between the upper and lower limits)
(D2 - D1) / 3*D2 < f < (D1 + D2) / 2*D2
Let's put some numbers in to get an idea of how critical things are: start out with a large difference in route lengths, say D1 = 100 and D2 = 200. A quick trip through the calculator yields
0.1667 < f < 0.75
So the alt roam must be at least 33.3 units away from the start (red) and not closer that 50 units from the finish (blue). This leaves 116.7 units available -- plenty of latitude available for placing the alt roam.
What if the routes are almost the same length? Try D1 = 100 and D2 = 101. We get
0.0033 < f < 0.995
The alt roam in this case must be at least 0.33 units away from the start and not closer than 0.505 units from the finish. This is much less critical than where the route lengths are considerably different.
The most stringent situation is where the long route is much, much longer than the direct route. In this case, the limiting valuse are
0.33 < f < 0.5
So you can use a rule-of-thumb that the alt roam must be placed at least a third of the way down the long route and not more than halfway. Put it at 0.415 from the start and you should never go wrong. Now all we need is a way to measure the path lengths accurately!
Please check my reasoning and my math, because my conclusion is that the alt roam placement shouldn't be picky as long as you stay away from each end of the route.
The experiment that I did last night of placing the alt roam at short intervals along the route should have worked if the above reasoning is correct.
NOTE: what I'm seeing is that the engineer bots run directly to the dynamite point by the shortest way. They do not travel down the long route to the alt roam and then turn back. They do not go through the finish to the alt roam and then turn back (they can't in this case because there's a door in the way). The longer route is simply ignored.
I must have something else wrong (and I'm going to feel real stupid when it's pointed out to me)!
THIRTY MINUTES LATER:
I thought that I'd try to measure the length of my two route paths to see what they really were. My "measurement" technique was to use a stopwatch to time how long it took to run each route, including descending a ladder on each one and climbing some boxes on one. When I got to the "long, safer" route, I needed to disconnect a couple of branch-offs. After doing that, my engineer bot refused to leave the spawn point. Uh Oh! There must be a break in the path. Sure enough, I found it at the long ladder descent -- I only had an upward node connection! I'd seen bots climbing it but hadn't noticed that they never went down.
P.S. The "long" safe route is one second FASTER than the supposed short route.
I said that I was going to feel stupid . . .