Collision detection using the separating axis theorem

Question

Lately I've been working on implementing a good collision detection system for my game. Before implementing it into my game I decided to make a sort of simulation. I've followed many tutorials (mainly this one) but I can get it working. I followed it step by step, but I don't know if my error is in the overlap (checking if the two projections overlap) or in the whole code itself.

The code has got the Vector2D and ConvexPolygon implementation as well. Here's the code:

http://pastebin.com/whV31SDi

It's pasted in pastebin because it's quite long.

It is always a dilemma how much to expand acronyms in question titles—that have to be short too—but a Google match shows that for most computer scientists SAT risks evoking the boolean satisfiability problem. en.wikipedia.org/wiki/Boolean_satisfiability_problem — Pascal Cuoq
– Pascal Cuoq, Commented Sep 1, 2012 at 11:40

john · Accepted Answer · 2012-09-01 11:07:21Z

1

Well as you say the code is rather long, and I didn't go through it all. But two errors stuck out

1) Your operator- is backwards

    Vector2D& operator - (const Vector2D& other)
    {
            Vector2D resultant;
            resultant = Vector2D(other.x - this->x,other.y - this-> y);
            return resultant;
    }

should be

    Vector2D& operator - (const Vector2D& other)
    {
            Vector2D resultant;
            resultant = Vector2D(this->x - other.x, this->y - other.y);
            return resultant;
    }

2) There are lots of style issues with your code. The biggest one is returning all your values by references to local variables. That is just asking for trouble because the variable no longer exists after the function returns. Here's a better way to write your operator+ for instance

   Vector2D operator+ (const Vector2D& other) const
   {
            return Vector2D(other.x + this->x, other.y + this->y);
   }

Still not perfect, because symmetrical operators like operator+ should be global functions not class members, but better than what you have.

answered Sep 1, 2012 at 11:07

john

88.3k5 gold badges61 silver badges95 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

john Over a year ago

@EdoardoTygerDominici Well I don't know much about collision detection but isn't it a problem that your second polygon isn't convex? The tutorial you mentioned says this algorithm works for convex shapes only. In fact some people would say the second polygon isn't a polygon at all, because it's edges cross.

Orrexon · Accepted Answer · 2014-06-17 12:06:27Z

0

I think it has to do with the fact you are using a bool-function to check the if there are overlaps.

In your for-loop you exit the function every time the shapes have not overlapped, and every frame the for loop starts over, and exits at the same place. change it to a void function.

I could be mistaken because I didn't really read all that code thorough, but I have had that problem myself.

answered Jun 17, 2014 at 12:06

Orrexon

34 bronze badges

Comments

lordseanington · Accepted Answer · 2018-07-09 19:25:21Z

0

This answer may be a bit late, however after looking at your code, it seems that you are only checking against the axes of only one of the convex polygons. SAT requires that you check against both shapes axes, which is most likely your issue. To fix this just repeat your collision check for the other shapes axes that weren't checked.

answered Jul 9, 2018 at 19:25

lordseanington

295 bronze badges

Collectives™ on Stack Overflow

Collision detection using the separating axis theorem

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related