I have been coding for some years in C# and have now decided to try out the XNA framework to make games.
It went great until I started to implement collision handling in my very simple game.
I can understand how to do collision detection using BoundingBoxes and BoundingSpheres, but after looking at the refresh rate in my game, it quickly became a concern of mine if the two colliding objects were never detected as colliding.
Let me try to explain with an example:
- If a character is shooting with a gun at another character.
- The bullet is heading straight for the other character.
- The bullet gets rendered just before the character.
- Because of the bullets high velocity it now gets rendered on the other side of the character.
In this scenario the bullet and the character never collides, because they are never rendered in their colliding state.
So how do you make sure to detect a collision in this scenario?
For very fast-moving objects, the regular approach fails in the scenario you described. What you need to check for is whether the bullet has collided with any item in the interval between the two consecutive game ticks. This is called continuous collision detection (this is a rather related SO post).
You can do this by basically casting a ray between the middle of the bullet's binding box in the current position and the middle of the box from the old position, and checking if that collides with any other blocks/spheres. This is a rather fast solution and if your bullets are small enough, it should work fine.
For more precision, you can cast four different so-called collision rays from each corner of the current bullet box, to their corresponding positions from the previous game tick. Note that in the rare event of high-speed items smaller than bullets, this might also fail. In this case you would need a more advanced collision detection system. But this would just be a corner case.
If that extra precision is a must, a free 3D physics library, such as Bullet could represent a solution. It even has bindings for XNA!
I think your problem is in tunneling that matches your description.
I recommend you to use Box2D physics engine for XNA. It's fast, simple and will avoid your any problems with collisions in your game.
Check in this manual on 4th page continuous collision and check how Box2D deal with tunneling effect.
Related
I'm just wondering how to make shots in my game more realistic. I mean there are two main ways in implementing shots. First: throw a small projectile and let it detect collisions. Second: use raycast. But in real world (and apparently I noticed this in PUBG) bullets fly really quick but not immediately. Here is why you should aim near your target if it's far away from you to hit it. Because if you aim right where it is it will move and your bullet will miss the target.
I'm just curious if any of you, guys have a nice solution to this problem. Also I wish to find a way to use raycast not every frame. In such things like detecting if you will actually hit the wall when shooting. If you have any good ideas how to implement spread and recoil for different types of weapons I will be happy to read them.
Blue point is bullet point of frame. and in very frame send a new raycast to check cross any human. like red line in this picture.
and bullet path use unity built-in Physics system to do it.
Im making a carrom kind of game where Im using physics for striker movement, the problem is my striker object bounces from the edge of the board even before there was a real collision between my striker object collider and the board edge colliders.
Please check this video where I have highlighted my project setup, is I'm doing anything wrong here?
https://drive.google.com/file/d/1xifD--sXHvQWjLbud-Lpt-qG3GAXu-hS/view?usp=sharing
I tried to fix it in many ways but wasn't able to reduce this offset and without resolving this issue I wasn't able to proceed further, because if there are 2 objects very near to each other and if the striker hits only one of those objects accurately still the nearest another object is also considering it as a collision even though there wasn't a real collision!
Is this a known bug? if not how can I remove this offset and can have accurate collisions that are required for carrom kind of games (where multiple objects are near to other)?
I got the mistake that I have done!
In the editor, every time while going to play mode striker object collision detection mode was set to "Continuous Speculative" and at run time I used to change it to "Discrete" or "Continuous" or "Continuous Dynamic" assuming I will be able to change it at run time but that's not the case, even though editor allows us to change collision detection mode at run time, technically it will only consider the mode which was set initially(i.e while starting the play mode), so If I have to see the behavior of "Continuous Dynamic" then I have to stop the play mode and set collision detection mode to "Continuous Dynamic" and run it again!
I am building a game in XNA 4.0, where a player moves about a 2 dimensional (vertical perspective) map consisting of blocks. My issue is creating proper collision between the payer and the blocks (basic game physics.) The player moves more than 1px per frame, so .Intersects() just isn't enough, I need physical contact collision that can function in a gravity environment. The current version I currently have is a piece of garbage and only works occasionally.
Basically, all that the collision system needs to do is stop gravity when the player lands on a block, and provide some decent physics when the player hits blocks (movement in that direction ceases). The idea behind my current solution is to move the next Position around until it finds a clear spot, but it doesn't work well. I have an idea why, just have no idea how to do it properly.
I know there must be a better way to do this. What would be the best method of making this kind of collision work properly?
Thanks
This does the job perfectly, you just need to sort through a few syntax errors. http://go.colorize.net/xna/2d_collision_response_xna/
I have implemented a full per pixel collision system that accounts for rotation, and its very accurate. It returns a simple bool on collision.
However I am not sure how to handle the collision from the player movement point of view.
E.g. In the picture above, if the player is holding up on the left stick, he should be stopped, but if he is holding up + right diagonal on the left stick, he should slide northeast alongside the side of the red square almost naturally.
How do I go about this, to make the player’s momentum stop, but still give control for the player to move in direction not blocked by a collision.
I could do this kind of thing with simple untransformed rectangles, but going into per-pixel texture collision has made my brain explode today so I’m hoping you guys can help. Any advice would be massively appreciated.
It is possible. For that you need to implement physics engine with your pixel based collision if you are willing to give natural effect in your game.
For that either you write your own game engine or use engine that is already there. For 2D game I highly recommended Farseer Physics Engine. It is out there for a long time. Now in stable condition and it surely wonderful.
It is developed using XNA only from scratch, and also performance wise it is far good.
Have a look at this. I hope I was able to give you answer. Please let me know if more details needed.
So I have been using Farseer for a platformer with physics, and I started running into some concerns as shown below:
When I want my character to jump, I don't only want to apply an impulse to the character body, but more realistically, the character body should be pushing down on objects below it (1) thus propelling itself upwards (2).
The same applies when I want my character to move. When the character moves, it should apply force and push off the objects below it (1), not carry the object with it and thus moving in the opposite direction (2).
Although I will be using skeletal animations for this game, I would like to keep it to be a bounding-box shape. This will be a networked game, and the physics will be handled via the server. I prefer to avoid server stress, so I will keep characters to a bounding box and keep all skeletal animations to the client side.
So here are my questions:
Am I forced to "fake" these forces?
Is it reasonable to duplicate skeletal animations on the server side,
and just use capsule bodies for the different bones of each character?
Will this be too intense?
The game should be able to support 20-30 people per room over multiple rooms. If it comes to it, I would like to avoid very heavy server stress.
I tried to post this to GameDev, but I don't have the reputation to post images there yet, so I will try to find an answer here first.
Instead of applying a force in the direction you want your player to move, apply it in the opposite direction. This will essentially create a collision, and the box will push your character up. If you play around with the body restitution property, you might be able to reach an acceptable "bounciness" that could work for you.
You would need to check if your player is resting on top of a box first, or else your player will just fly downwards.
Farseer bodies should keep track of any current collisions, this is how the character knows it is supported by something underneath. I would recommend applying a force to whatever the character is supported by, and if you are going for realism, take into account the mass of both the character and the supported object, and how high the jump is going to be. Whatever is calculated as downward force to be applied to the supporting body should be reversed and also applied to the character.
If you want to try and use Farseer to make this happen, rather than calculate it yourself, you might be able to do this using springs, but otherwise pushing down on the supporting object isn't going to get the player to push upward.