I wrote a collision detection system for a game I am working on, and I am experiencing a weird glitch where, occasionally, projectiles will go through the player or walls scattered throughout the level. Because the projectiles can be fired at any angle, I decomposed the bounding box of each projectile into multiple, smaller bounding boxes that I then rotate around the center of the texture according to the rotation of the projectile in space. For some reason, occasionally a Spear projectile will go through the player or a wall even through others do not.
I use the following methods to determine the rotation of the texture and to translate the bounding boxes:
public double RotateToFaceTarget()
{
double rotation = Math.Atan2((double)_direction.Y, (double)_direction.X);
return rotation;
}
public List<BoundingBox> TranslateBoundingBox(List<BoundingBox> box, double rotation)
{
List<BoundingBox> newBounds = new List<BoundingBox>();
foreach (BoundingBox b in box)
{
Vector2 boundsOrigin = new Vector2(b.Pos.X + b.Size.X / 2, b.Pos.Y + b.Size.Y / 2);
Vector2 texOrigin = new Vector2(_pos.X + _texture.Width / 2, _pos.Y + _texture.Height / 2);
Vector2 newPosBasedOnOrigin = Vector2.Transform(boundsOrigin - texOrigin, Matrix.CreateRotationZ((float)rotation)) + boundsOrigin;
newBounds.Add(new BoundingBox(newPosBasedOnOrigin, b.Size));
}
return newBounds;
}
_direction is calculated by subtracting the position of the projectile from the target location and normalizing. I use this method to determine if the projectile is colliding with another entity:
public bool ProjectileCollision(Entity e, Projectile entity2)
{
if (entity2.CanCollide)
{
foreach(GameObject.BoundingBox b in entity2.BoundingBox)
{
foreach(GameObject.BoundingBox b2 in e.BoundingBox)
{
if (b2.Intersect(b) && (entity2.IgnoredEntities.Contains(e.Type) == false))
{
entity2.IsActive = false;
e.Health -= entity2.Damage;
return true;
}
return false;
}
}
return false;
}
return false;
}
And this is my Bounding Box Intersection method:
public bool Intersect(BoundingBox intersected)
{
if ((_pos.Y < intersected.Pos.Y + intersected.Size.Y) && (_pos.Y + _size.Y > intersected.Pos.Y) && (_pos.X + _size.X > intersected.Pos.X) && (_pos.X < intersected.Pos.X + intersected.Size.X))
{ return true; }
return false;
}
EDIT: On further testing, it seems that the projectile will always detect a hit if the player hits based on the top left corner ( which makes sense now that I look at my intersect code). Is there another way to re-write my Intersect method to use something more accurate than the top left corner?
EDIT2: I drew the hitboxes for certain objects, and this is one instance of when I catch the spear going through the player:
http://imgur.com/a/fAxZw
the player is the larger pink square. The hitboxes are not being translated correctly, but it shouldn't just stop working, for some and not others, right?
It could happen, because of high velocity and small object, projectile could fly through object. Beside of checking if objects are intersecting, you have to check if object will intersect, to check if object is in line of fire. You could achieve this by raycasting.
On those cases like your i had function that check if object is near other. Simple checking if object is inside of some radius of other object. If yes then i was checking if object is flying toward other object, and checking distance between them. When distance is really close then collision happened.
Related
I am using the Unity Engine with C#.
I have a 1x1 cube which moves forward on a grid of 49, 1x1 cubes (screenshot below) - when I press the start button on the controller.
The movement code for the cube is below.
void MovePlayerCube()
{
transform.Translate(direction * moveSpeed * Time.deltaTime);
}
When this cube passes over a cube with an arrow on it, the cube will change direction to where the arrow is pointing (staying on the same Y axis).
I need to detect the exact point at which the cube is directly over the cube with the arrow on it, and run the 'change direction' code at that point.
I'm currently using Vector3.Distance to check if the X and Z coordinates of the 2 cubes are close enough together (if they are less than 0.03f in distance), I can't check if they are equal due to floating point imprecision.
However this is really ineffective as half the time this code doesn't register for probably the same reason, and if I increase the 0.03f to a point where it never misses it becomes really noticeable that the cube isn't aligned with the grid anymore.
There has to be a proper solution to this and hopefully I've clarified the situation enough?
Any advice is appreciated.
You are moving your cube via
transform.Translate(direction * moveSpeed * Time.deltaTime);
which will never be exact an might overshoot your positions.
=> I would rather implement a coroutine for moving the cube exactly one field at a time, ensuring that after each iteration it fully aligns with the grid and run your checks once in that moment.
It doesn't even have to match exactly then, you only need to check if you are somewhere hitting a cube below you.
So something like e.g.
private Vector3Int direction = Vector3Int.left;
private IEnumerator MoveRoutine()
{
// depends on your needs if this runs just forever or certain steps
// or has some exit condition
while(true)
{
// calculate the next position
// optional round it to int => 1x1 grid ensured on arrival
// again depends a bit on your needs
var nextPosition = Vector3Int.RoundToInt(transform.position) + direction;
// move until reaching the target position
// Vector3 == Vector3 uses a precision of 1e-5
while(transform.position != nextPosition)
{
transform.position = Vector3.MoveTowards(transform.position, nextPosition, moveSpeed * Time.deltaTime);
yield return null;
}
// set target position in one frame just to be sure
transform.position = nextPosition;
// run your check here ONCE and adjust direction
}
}
start this routine only ONCE via
StartCoroutine(MoveRoutine());
or if you have certain exit conditions at least only run one routine at a time.
A Corouine is basically just a temporary Update routine with a little bit different writing => of course you could implement the same in Update as well if you prefer that
private Vector3Int direction = Vector3Int.left;
private Vector3 nextPosition;
private void Start()
{
nextPosition = transform.position;
}
private void Update()
{
if(transform.position != nextPosition)
{
transform.position = Vector3.MoveTowards(transform.position, nextPosition, moveSpeed * Time.deltaTime);
}
else
{
transform.position = nextPosition;
// run your check here ONCE and adjust direction
// then set next position
nextPosition = Vector3Int.RoundToInt(transform.position) + direction;
}
}
Then regarding the check you can have a simple raycast since you only run it in a specific moment:
if(Physics.Raycast(transform.position, Vector3.down, out var hit))
{
direction = Vector3Int.RountToInt(hit.transform.forward);
}
assuming of course your targets have colliders attached, your moved cube position (pivot) is above those colliders (assumed it from your image) and your targets forward actually points int the desired new diretcion
I would do it this way. First I would split the ability of certain objects to be "moving with certain speed" and "moving in a certain direction", this can be done with C# interfaces. Why? Because then your "arrow" cube could affect not only your current moving cube, but anything that implements the interfaces (maybe in the future you'll have some enemy cube, and it will also be affected by the arrow modifier).
IMovingSpeed.cs
public interface IMovementSpeed
{
float MovementSpeed{ get; set; }
}
IMovementDirection3D.cs
public interface IMovementDirection3D
{
Vector3 MovementDirection { get; set; }
}
Then you implement the logic of your cube that moves automatically in a certain direction. Put this component on your player cube.
public class MovingStraight: MonoBehaviour, IMovementSpeed, IMovementDirection3D
{
private float _movementSpeed;
Vector3 MovementSpeed
{
get { return _movementSpeed; }
set { _movementSpeed = value; }
}
private Vector3 _movementDirection;
Vector3 MovementDirection
{
get { return _movementDirection; }
set { _movementDirection= value; }
}
void Update()
{
// use MovementSpeed and MovementDirection to advance the object position
}
}
Now to implement how the arrow cube modifies other objects, I would attach a collision (trigger) component for both moving cube and the arrow cube.
In the component of the arrow cube, you can implement an action when something enters this trigger zone, in our case we check if this object "has direction that we can change", and if so, we change the direction, and make sure that the object will be aligned, by forcing the arrow cube's position and the other object's position to be the same on the grid.
public class DirectionModifier: MonoBehaviour
{
private Vector3 _newDirection;
private void OnTriggerEnter(Collider collider)
{
IMovementDirection3D objectWithDirection = collider as IMovementDirection3D ;
if (objectWithDirection !=null)
{
objectWithDirection.MovementDirection = _newDirection;
// to make sure the object will continue moving exactly
// from where the arrow cube is
collider.transform.position.x = transform.position.x;
collider.transform.position.y = transform.position.y;
}
}
}
If you made your trigger zones too large, however, then the moving cube will "jump" abruptly when it enters the arrow cube's trigger zone. You can fix it by either starting a coroutine as other answers suggested, or you could make the trigger zones pretty small, so that the jump is not noticeable (just make sure not to make them too small, or they may not intersect each other)
You could then similarly create many other modifying blocks, that would change speed or something
I think that it is enough for you to check if the X and Z coordinates are equal, since the movement occurs only along them
Example
if(_player.transfom.position.x == _gameSquare.transfom.position.x && _player.transfom.position.z == _gameSquare.transfom.position.z)
{
Debag.Log("Rotate!")
}
I have run into a slight issue with the collision resolution in my game engine. If two objects collide and that collision causes the velocity to go to zero, the edges of the objects will overlap each other and they'll be stuck.
Is there a way to implement a catchall for this kind of a situation? i.e. move the objects just enough in the right direction so they are not stuck.
Here is how I am checking collisions and moving objects. When update is called on an entity, it moves the (x,y).
public static void Update()
{
for (var iterator = 0; iterator < PhysicsEntities.Count; iterator++)
{
for (var index = iterator + 1; index < PhysicsEntities.Count; index++)
{
if (!Collision.ResolveCollision(PhysicsEntities[iterator],
PhysicsEntities[index], Detection)) continue;
PhysicsEntities[iterator].Update();
PhysicsEntities[iterator].Collided = true;
PhysicsEntities[index].Update();
PhysicsEntities[index].Collided = true;
}
}
foreach (var entity in PhysicsEntities)
{
entity.Update(velocity: true);
entity.Collided = false;
}
}
}
Here is the update function for the entities:
public void Update(bool velocity = false)
{
if(!Movable) return;
if (!Collided)
{
var moveX = Velocity.X / Universe.UpdateInterval;
var moveY = Velocity.Y / Universe.UpdateInterval;
Position.Move(moveX, moveY);
BoundingBox.Move(moveX, moveY);
}
if(velocity) UniversalForces();
}
private void UniversalForces()
{
Velocity.Scale(1 - Universe.Friction);
Velocity.Add(Universe.GravityMag, Universe.GravityDir);
}
Finally, here is a image of one simulation where the objects get stuck. As you can see, it is just the edges that are getting stuck:
The quick solution is to move both objects back to the previous tic's position and any other object that causes a collision with to move back as well. It works, but it looks messy and causes some behavior that looks really bad - things like pushing directly at a wall leaves a gap, but angling towards a wall leaves a smaller gap. Very messy.
The better solution is to move both objects back just far enough along their negative velocity vector so that they are no longer touching. Usually some dot product math can give you what you need for this, though iterating backwards can work (slow).
Long story short, don't ever allow objects to overlap. Take care of it before it happens and you avoid stuck jitters, can't move stuff, etc.
When one object collides with another, get it to backtrack up its own movement vector so that the distance between the centroids of both objects is equal to the two radius. This works if its a circle - if its a complex polygon you really need to do edge collision detection instead of bounding sphere detection. If the bounding spheres collide, then you move to complex edge detection. In the end the trick is the same; check for collision then back up the movement vector until you find the exact (or nearly exact) point of collision.
I was able to figure it out with the suggestions people made. Some of the changes I made include having each object collide with any other object only once per update, and moving the object after the collision until it is no longer colliding. Here is the code that I used to do that, feel free to use this on any project and let me know if you have any questions about it.
public static void Update()
{
foreach (var a in PhysicsEntities)
{
foreach (var b in PhysicsEntities)
{
if (a.Equals(b) ||
!Collision.ResolveCollision(a, b, Detection) || b.Collided) continue;
while (Detection == Detection.BoundingBox ?
Collision.BoundingBox(a, b) :
Collision.PixelPerfect(a, b))
{
const float moveBy = .5F;
var moveX = a.Position.X > b.Position.X ? moveBy : -moveBy;
var moveY = a.Position.Y > b.Position.Y ? moveBy : -moveBy;
if (a.Movable)
{
a.Move(moveX, moveY);
a.Velocity.Scale(-1);
}
else if (b.Movable)
{
b.Move(moveX * -1, moveY * -1);
b.Velocity.Scale(-1);
}
}
a.Update();
b.Update();
a.Collided = a.Movable;
}
}
foreach (var entity in PhysicsEntities)
{
entity.Update(velocity: true);
entity.Collided = false;
}
}
I am trying to handle the position of a cube when I move it close to another one. When I drag a cube next to another, I want it to have the same rotation and be perfectly next to it. Here is an example of what I am trying to do :
When I drag the second square close to the one on the top, it must end up like this :
I implemented a function to clip the selected cube to another :
public override void clip(Cube other){
//print("North:" + allSides[north] +" East:" + allSides[east] +" South:" + allSides[south] +" West:" + allSides[west]);
float distX = this.transform.position.x - other.transform.position.x;
float distZ = this.transform.position.z - other.transform.position.z;
bool isClips = false;
if(Math.Abs(distX)<2.2 && Math.Abs (distZ)< 0.5){
transform.rotation = other.transform.rotation;
if(distX<0){
move(east,other.transform.position.x-SQUARE_BASE_LENGTH,other.transform.position.z,other);
}
if(distX>0){
move(west,other.transform.position.x+SQUARE_BASE_LENGTH,other.transform.position.z,other);
}
isClips = true;
}
if(Math.Abs(distZ)<2.2 && Math.Abs (distX)< 0.5){
transform.rotation = other.transform.rotation;
if(distZ<0){
move(north,other.transform.position.x,other.transform.position.z-SQUARE_BASE_LENGTH,other);
}
if(distZ>0){
move(south,other.transform.position.x,other.transform.position.z+SQUARE_BASE_LENGTH,other);
}
isClips = true;
}
}
But I have a result like this :
or
depending on which side is the closest. How can I make so that my 2 cubes attach each other as expected?
Some points for further calculation:
The diagonal distance between the two objects has to be your edge length afterwards.
Easiest would probably be to calculate the directional vector from the other into the direction you want to attach to, resize it to fit your edge length and add it to the position of the other object to get the position of the object you want attach.
You could also do some stuff with angles (sin/cos/tan) and pythagoras where the hypothenus would be your edge length, but I think this requires more calculation.
I have to make a game with Windows Forms for school. My game consists of a user having to get through a maze. I'm trying to prevent my user from going straight through the walls using collision detection, but am getting stuck because of the varying shape of the rectangles being used to represent walls. Here's an image of the game. This question may be similar to this one, however with my movement I believe that it is quite different, as I don't have a grid system or graphical map laid out.
As you can see, the walls are fairly thick. Each wall is represented by a C# Rectangle, as is my Player image (the little yellow ghost). I know how to determine if the player is crossing through these walls using C#'s IntersectsWith(Rectangle r) method, but I'm not exactly sure how to use this information in order to handle the collision and STOP the player from moving through the walls at all.
Here's what I've tried:
This is my actual movement code. Because the game is built in WinForm, the movement is triggered by keyboard events such as OnKeyPressed and OnKeyUp
public void Move(Direction dir)
{
HandleCollision(); // Handle collision while player is trying to move.
if (dir == Direction.NORTH)
{
this.y -= moveSpeed;
}
if (dir == Direction.SOUTH)
{
this.y += moveSpeed;
}
if (dir == Direction.EAST)
{
this.x += moveSpeed;
}
if (dir == Direction.WEST)
{
this.x -= moveSpeed;
}
}
This is my collision method, HandleCollision():
private void HandleCollision()
{
// First, check to see if the player is hitting any of the boundaries of the game.
if (this.x <= 0)
{
this.x = 0;
}
if (this.x >= 748)
{
this.x = 748;
}
if (this.y <= 0)
{
this.y = 0;
}
if (this.y >= 405)
{
this.y = 405;
}
// Second, check for wall collision.
foreach (Rectangle wall in mazeWalls)
{
if (playerRectangle.IntersectsWith(wall))
{
if (player.X > wall.X) { player.X += wall.Width; }
else if (player.X < wall.X) { player.X -= wall.Width; }
else if (player.Y > wall.Y) { player.Y += wall.Height; }
else if (player.Y < wall.Y) { player.Y -= wall.Height; }
}
}
}
Now this code above kind of works. However, because the player's coordinates are having the wall's width/height added to it, this makes some weird collision teleportation across the map where the player ends up bouncing around. So what would be the most efficient way of going about implementing a collision detection system which can replace all the code within the if (playerRectangle.IntersectsWith(wall)) { block?
In the move, save the current position, perform the move, check for collision, and if true, restore the old position.
For this to work, HandleCollision would return a Boolean value, true for each successful test (successful=collision detected), false at the end if no condition was met. The method would not modify any x or y value at all. To reflect its new function, this method should be renamed to CheckCollision.
From the gif in your other post it seems that your playing field is divided into squares, and walls and sprite are composed of multiple squares? Then your move should proceed in increments of one square until the intended increment is met. For each square the collision has to be checked.
For a more advanced method with only one check per move you need some linear mathematics. The collision check will need the current point P and the increment d of the move. You need to check if between P and P+d the wall is crossed. This is usually indicated by a sign change in a linear condition. For a vertical wall W to the left of P, if d points to the wall, P.x+d.x-(W.x+W.width) will be negative. Determine s with W.x+W.width<=P.x+s*d.x. For horizontal walls, you have to do the same check in the y-coordinate. Take the minimal s over all objects hit and P+s*d as the new position.
SO I currently have an object which moves in the Y axis (up and down). How would I be able to program it so the program knows whether the object is moving up or down?
I understand I'd need an if statement like the following:
if (object is moving up)
{
//set direction to 1
}
else
{
//object must be going down, set direction to 2
}
I just don't understand what syntax I'd need to use. This would be easy if I was holding down the key and it was moving up or down however that's not the case. The object is a bouncing ball and therefore when you set a power the ball jumps, and bounces so it is constantly changing.
Thanks for your help, let me know if this wasn't described well and you need more info.
You need to save the current coordinates so that you can check them after the update method is called. Then you can check the differences between the previous and current location:
Declare it in Game1() constructor:
Point previous = new Point();
previous.X = myObject.InitialX;
previous.Y = myObject.InitialY;
In Update:
int deltaX = object.X - previous.X;
int deltaY = object.Y - previous.Y;
if (deltaX < 0)
{
object is moving upwards
}
else
{
object is moving downwards
}
if (deltaY < 0)
{
object is moving to the left
}
else
{
object is moving to the right
}
//update the previous state
previous.X = myObject.X;
previous.Y = myObject.Y;
When you update the position of the ball you have one position and you calculate the position for the next frame -> You can calculate the distance vector.
With the sign of the y-value of this distance vector you can decide if it's moving upwards, downwards or isn't moving (=0).