search

Scirra / Construct-bugs

Public bug report submissions for Construct 3 and Construct Animate. Please read the guidelines then click the 'Issues' tab to get started.
https://www.construct.net
107 stars 83 forks source link

Turret behavior predictive aim bug #7965

Closed TackerTacker closed 6 months ago

TackerTacker commented 6 months ago

Problem description

The predictive aim in the Turret behavior seems to have a bug.

https://github.com/Scirra/Construct-bugs/assets/18266300/57871f98-e732-4513-b885-a8ba8e4215bd

In the video (or running the project) it looks like there are 3 phases

My guess for the 3rd phase is that there is no actual solution at the speed of the bullets and player, so the equation just results in weird angles. I wish it would fail more gracefully, maybe reverting back to just pointing straight at the player without predictive aiming, but the 2nd phase is actually what really feels like a bug to me. Feels like it's off by 180°

( As a side note, I wish the aim angle would be available without the need to actually turn the object itself. Similar to how the Sine behavior has a "Value only" output option [ But I know that's not a bug ] )

Attach a .c3p

turret_auto_aim_bug_01.zip

Steps to reproduce

  1. Just open the project and run it

Observed result

The Turret Behavior is not able to predict the shot reliably, even if the player is moving at a constant speed and the turret rotation speed is high enough to be instantaneous.

Expected result

The Turret Behavior should predict shots more reliable and maybe fail more gracefully in case a shot is impossible.

More details

Affected browsers/platforms:

First affected release:

System details

View details PASTE HERE
AshleyScirra commented 6 months ago

Sorry, I don't know how to fix this. I didn't come up with the predictive aim maths myself, and I don't really know why it would produce weird results like that. It already reverts to basic aim if there is no solution and this project shows that happening, but it seems in extreme cases just before it gets to "no solution" it ends up with "wrong solution". If you can figure out the correct maths or checks to avoid this case perhaps it could be implemented, but otherwise I would say either use basic aim or design the project to avoid this happening.

TackerTacker commented 6 months ago

Hmmm that's too bad, I appreciate you looking into it anyways.

TackerTacker commented 6 months ago

I asked R0j0hound for help and he found a solution. He created a working example using events. I think he figured out that there are 2 possible solutions to the problem and currently the Turret Behavior basically picks the wrong one, or maybe doesn't account for 2 solutions? Idk, I don't understand the math, R0j0hound would have to explain it.

Anyways, here is the example project turret_behavior_aim_bug_and_event_based_solution2.zip

TackerTacker commented 6 months ago

R0j0hound made another version in which he added a 2nd solution he thinks is closer to the original formula of the Turret behavior. turret_behavior_aim_bug_and_event_based_solution3.zip

He says his 2nd version is very similar to the 3rd solution described in this article "Strategy # 3 - Assuming Zero Acceleration" https://www.gamedeveloper.com/programming/predictive-aim-mathematics-for-ai-targeting#close-modal