Update unstick.bt

[superdupont-fr]

Extented with a second part, from 3 to 5 seconds without suicide.
The first part, was modified a few, and apply for 7 to 15 second and end with suicide in case still stuck.

[slipher]

I'm wary of triggering it after so little time (3 seconds) as this will frequently happen when the bot is not really stuck.

Can you give some qualitative explanation of how this improves the unsticking performance? Why is it good to move in a direction for only 100 ms and then change direction?

[superdupont-fr]

I'm wary of triggering it after so little time (3 seconds) as this will frequently happen when the bot is not really stuck.

In fact, I tried with time at 1 seconds to ensure quicker unstuck. Worked fine, but not easy to debug for the moment, so, I keep the 3 seconds time.

Can you give some qualitative explanation of how this improves the unsticking performance? Why is it good to move in a direction for only 100 ms and then change direction?

In many case, moving for short time like 100ms or 200ms in diferent directions before trying to roam can unstick in many case.
Long sequence of walking in same direction is useless. I watched bots in Thunder stucked in large corridor ending to suicide because of long time in same direction.
Also, the mechanism that check if the bot is still stuck checking each seconds, that I think is too long. Many bots have time to come back at the almost same position in this time. My first BT with several small steps was already a good improvement.
My second update introduce a check if there is enemy nearby, and lets bots react if enemy appear in the sequence for unstick.
My third update is almost the same, but with include behavior to shorter files and improve readability. (I'm doing the third update now).

[superdupont-fr]

The 4th update will be without debug info.

[necessarily-equal]

Suggested change

	condition distanceTo( E_ENEMY ) <= 300
	{
	sequence
	{
	action resetStuckTime
	condition ( alertedToEnemy && distanceTo( E_ENEMY ) <= 300 )
	{
	sequence
	{
	action resetStuckTime

I think you are calling resetStuckTime accidentally if you don't check there is an enemy. Otherwise, if I'm reading correctly, the distance to E_ENEMY is 0 if there is no enemy in sight.

[superdupont-fr]

Always need alertedToEnemy with distanceTo( E_ENEMY ) ?

[necessarily-equal]

Yeah basically, otherwise you'll read a distance of 0

[superdupont-fr]

But, it's worked in my tests. (of course, without debug commands, I can't be sure that it's works really).

[superdupont-fr]

I more understood some BT parts. I rewrote some parts. Will test more before update.

[superdupont-fr]

I also removed the part specific for Humans, when stucked with the CT.

[illwieckz]

This repository has no longer use as it is not a submodule anymore:

• Undoing unvanquished_src.dpkdir submodule Unvanquished/Unvanquished#2625

If there is still something to merge from it, this PR should be redone on https://github.com/Unvanquished/Unvanquished instead.

I let this PR open so the one redoing the PR on the other repo can close it when it's redone if it has to be redone, so we know what PR is redone or not yet.