DirectPhysics

DirectPhysics

Introduction:
Welcome to the DirectPhysics project. We will be designing an add-on to DirectX using C++ that will allow users to easily implement realtime physics into their games. There will be 2 versions of this. One will be an ActiveX .NET object for the .NET languages. And the other will be an ActiveX COM object to be used with VB6. Anyone is welcome to suggestions, comments, and to contribute anyway you can. Feel free to PM me if you are wanting on the team.

The Team:

• Jacob Roman - Project Director, Math and Physics, Theory, Coder, Tester
• moeur - Co-Project Director, Math and Physics, Theory, Coder
• jcis - Math and Physics, Theory, Coder
• Harsh Gupta - Math and Physics, Theory, Coder
• eyeRmonkey - Math and Physics, Theory
• damasterjo - Math and Physics, Theory
• SAGremlin - Math and Physics, Theory, Tester

Interested in being on the team?
If you want to be a team member in this project, you must have a good math background and some experience with Physics (mainly game related physics), Calculus, and Trig. You don't need much experience with DirectX. Maybe just a little so you have a good grasp on game programming and how the DX syntax is. And if you want to help code, a good understanding of C++ and maybe a little assembly should help.

I am open for an assignment. I am still unclear as to what this will offer and how it will be done. I am usually pretty good in programming math things and stuff like that. So learning the physics should not be much of a problem. Also do you want this to get to a level where, it is included in the directx package. Like one day....
New class added Direct Physics, programmed by Jacob Roman.?

The first thing I would like to suggest is to allow the user to select a numerical integration method from a list of constants to be used in a function. Here are some integration methods they can choose:

• Forward (Explicit) Euler
• Backward (Implicit) Euler
• Symplectic Euler
• Semi-Implicit Euler
• 2nd Order Runge Kutta
• 4th Order Runge Kutta
• 6th Order Runge Kutta
• (Velocity-less) Verlet
• Time Corrected Verlet
• Velocity Verlet
• LeapFrog Verlet
• Newton Stormer Verlet
• LeapFrog
• ...etc.

Then when the function is called in a loop, it'll update the time steps and allow the physics to happen.

Quote:

---

Originally Posted by damasterjo

I am open for an assignment. I am still unclear as to what this will offer and how it will be done. I am usually pretty good in programming math things and stuff like that. So learning the physics should not be much of a problem. Also do you want this to get to a level where, it is included in the directx package. Like one day....
New class added Direct Physics, programmed by Jacob Roman.?

---

Like I said earlier in the first post, it will allow users to easily be able to add realtime physics to their DirectX applications. Hopefully that's clear enough. Physics is an advanced topic, and many people find it too difficult to be able to implent physics on their own. This ActiveX component we are making in C++ will make it easier on them.

It will be a separate reference, like how DirectShow is separate. Unless we work at MS and have access to their DX source code, which we don't. :p

Quote:

---

The first thing I would like to suggest is to allow the user to select a numerical integration method from a list of constants to be used in a function.

---

I'd like to first see more of the big picture.

For instance, how are we going to implement this whole thing?

Give me an example of a task the user wants to perform and then the steps you envision him doing using our new library.

Is this going to mostly involve 2d rigid body collisions?

This is a very interesting under taking...

It's more of a front end for DirectX - 3D gaming development..

Alot of 3d development gets stuck at the pyshics area, or have dead slow code because they use alot of floating point arithmatic.. ( this is something to Aviod completely.) you only NEED to use FPA for scientific calculations..

in 3d gamming nearest estimation works fast, and isn't noticable in the output..

So i would say that the first objective is to find fast alternatives to most physics (FPA) calculations...

Belive it or not simple calculation like 14% (Tax) is faster with bit manipulation than straight multiplication..

Gremmy....

Jake,

I'm going to be very blunt. You are jumping into this WAY too quick. I'm sure you are excited about it, but we need to learn from the mistakes we made in LightFusion. From post #3 I know you are going along way to fast. We need a much much cleared objective laid out and then a plan that details that EXTEREMLY well before we make another move.

Of course that is just my opinion, but I think others will agree. If you don't want to design the plan then deligate it, but I think you should make the plan considering this is your project.

Quote:

---

Originally Posted by eyeRmonkey

Jake,

I'm going to be very blunt. You are jumping into this WAY too quick. I'm sure you are excited about it, but we need to learn from the mistakes we made in LightFusion. From post #3 I know you are going along way to fast. We need a much much cleared objective laid out and then a plan that details that EXTEREMLY well before we make another move.

Of course that is just my opinion, but I think others will agree. If you don't want to design the plan then deligate it, but I think you should make the plan considering this is your project.

---

Ok, let's plan ahead. Sorry bout the rush, but considering the fact that Physics is a very advanced topic, this library is gonna be so big, that I wouldn't know where to start. But we can break physics down into small groups so we don't bring too much overhead to ourselves.

DirectPhysics will be broken down into multiple classes. Each class will represent a special group, such as Force, Mass, Integrators, Time, Collision, Rigid Bodys, Ragdoll, Interpolation, etc., and with each, will also contain special functions and attributes. Like force, for example, will contain functions that calculate forces such as Friction, Applied, Gravity, Spring, Magnetic, Drag, Normal, etc., which will all be added into Net Force, which means the sum of all forces applied to an object. There will be 3 types of these functions within the classes. 1D, 2D, and 3D.

I'd draw you a diagram, but ever since I moved out of my parents house, I don't have a computer just yet, and had to use one at a library, which doesn't allow me to do much except webbrowsing.

Okay thats a good start. I assume you mean a diagram for the outline not for the physics? When you do draw up an outline you also need to inculde what is going to happen when. Not just what sections of physics there are. You also need to set aprox. deadlines for each step of the project. Without all that we are bound for failure again.

Does anyone else have any suggestions?

Set a deadline for each step of the project? To be honest, I dunno how long it's gonna take, but we can work on each class one day at a time, nailing in as much as possible until completed.

I would suggest each person to work on each class but then we would have to change our code around to make all the classes compatible. Which would suck.

thank you for selecting me into this team. my intentions was to learn all the Maths and Physics stuff which are necessary, interesting and moreover, need to be remembered, ;)

i have few things to say:

1- like i mentioned before, i need to learn the advanced C++ techniques (vectors, maps etc), and atleast basics of Calculus, Trigs, Numerical methods, and Physics so i can get started. so i need atlest 4-5 days (i presume i can start working properly prior to next Saturday).

2- the reason i used "Basics of", because the last time i studied all this, was way 4-5 years back, so i still need to brush up all these by reading the theory and trying few complex problems.

3- like your itinerary, you can assign me the basic modules before Friday and i will try to finish it in my free time.

4- one suggestion, i quite agree with ERM. it is better to be late than be dead/ horrified/ horrostruck etc. etc. :bigyello:

Harsh Gupta

To help brush everyone up on physics, I have many places you can learn.

Physics Classroom
GameDev.net Articles
DMWiki
And ACS Project: Maya
MyPhysicsLab, with Java source code

If I may make a suggestion.

First of all, sort out some kind of core package to deal with vectors. Include vector sums, dot products, cross products, transpose, adjoint, jacobians :) , grad, and anthing else you can think of. Matrices would also be useful.
You really won't regret it, because the last thing you want is lots of different people doing vector algebra in different ways. Get this sorted at the beginning, and the rest will just plug into it.


zaza

Following on from my eariler post, I can recommend this site.
The link to the Numerical Recipes in C++ part, although not containing the full book, does contain the first chapter about implementing vector and matrix classes in C++.
The other links actually host the full books online for Fortran and C. There is lots of useful stuff in here, and it is very much worth reading (and the code pillaging).
I would suggest that if you want to get stuck into some coding from the start, then getting into these basic maths classes is definitely the way to go. You will absoultely need the power they will provide later, and you can be planning the rest of the DirectPhysics engine in parallel.


zaza

Quote:

---

Originally Posted by Jacob Roman

DirectPhysics will be broken down into multiple classes. Each class will represent a special group, such as Force, Mass, Integrators, Time, Collision, Rigid Bodys, Ragdoll, Interpolation, etc., and with each, will also contain special functions and attributes. Like force, for example, will contain functions that calculate forces such as Friction, Applied, Gravity, Spring, Magnetic, Drag, Normal, etc., which will all be added into Net Force, which means the sum of all forces applied to an object. There will be 3 types of these functions within the classes. 1D, 2D, and 3D.

---

First of all thanks for letting me Join....

Now my thoughts on this..

We need to draw up a full list of all the fundamental Physics, including all the sub factors of each fundamental,and start building from each unique factor.

Alot of these factors are only 1D. IE Gravity affects motion in only one direction.. Acceleration/Deceleration and the rate there of, can be coded as a single, 1D class and simply applied on the X,Y & Z axis, as needed..

Only our bottom level classes (those the the user see) need be a full 3D type class... IE. Motion occurs in all three axis simultaniously, and needs to be applied as such...

Our Top Level classes (The Core of the Pysics) need to be compact, but Fast..

The only fundamental I can see problems with is Friction.. Friction values are fluid.. changing according to every concievable factor.. Temp, Mass, Surface, Speed, Direction, just to name a few predominant ones...

Gremmy...

Quote:

---

...Alot of these factors are only 1D...

---

you mean 2D?? i dont know if 1D objects really exists in this world?? :confused:

Quote:

---

Originally Posted by SAGremlin

We need to draw up a full list of all the fundamental Physics, including all the sub factors of each fundamental,and start building from each unique factor.

Alot of these factors are only 1D. IE Gravity affects motion in only one direction.. Acceleration/Deceleration and the rate there of, can be coded as a single, 1D class and simply applied on the X,Y & Z axis, as needed..

Only our bottom level classes (those the the user see) need be a full 3D type class... IE. Motion occurs in all three axis simultaniously, and needs to be applied as such...

Our Top Level classes (The Core of the Pysics) need to be compact, but Fast..

The only fundamental I can see problems with is Friction.. Friction values are fluid.. changing according to every concievable factor.. Temp, Mass, Surface, Speed, Direction, just to name a few predominant ones...

---

I realise I'm not part of the team, so I hope you won't think I'm interfering, but I really would advise against going from the top down. Don't worry so much about what you want to put in it yet, worry about how to get it implemented. Once the maths behind it is solid, then the equations will flow therefrom. In the above statements, for example, I wouldn't go into details about gravity acting in 1 dimension, or applying acceleration / deceleration in 1D on an axis as required. When you have multiple systems, such as modelling the solar system for example, then you'll have multiple gravitational forces acting in different directions. What you'll be interested in is the resultant gravitational force, so you'll be doing vector sums. As i say, IMHO you'd be far better off coding general vector algebra in, then applying this to gravitational forces, magnetic forces, frictional forces etc etc.
Stuff like vector operations (algebra, calculus etc), matrix operations etc will be essential regardless of the physics equations to which you apply it, and you will then have far greater scope to include the physics that you want later. Regardless of the physics you'll need, I guarantee that you will need these maths operations, so better to get it in at the start.


zaza

I agree with zaza about where we should start. However we still need to decide what the end product is going to look like so that we can work toward that goal.

Here is how I picture the end product is going to be used. I'f I'm way off base then say so.
Something like this

VB Code:

Dim myPhysics As clsDirectPhysics
Set myPhysics = New clsDirectPhysics
 
'add bodies
Dim myBody As clsDPBody
Set myBody = New clsDPBody
 
With myBody
    .Mass = 10 'kg
'set initial coordinates and velocity
    With .Rectangular
        .X = 100000 'm
        .Y = 200000 'm
        .Z = 0 'm
        .Vx = 1000 'm/s
        .Vy = 500 'm/s
        .Vz = 0
    End With
 
'Alternatively we could use spherical or cylindrical coordiantes
    With .Spherical
        .Radius = 223607 'm
        .Theta = 0.463 'radians
        .phi = 0
    End With
 
End With
myPhysics.Bodies.Add myBody
 
With myBody
    .Mass = 1000000 'kg
'set initial coordinates and velocity
    With .Rectangular
        .X = 0
        .Y = 0
        .Z = 0
        .Vx = 0 'm/s
        .Vy = 0 'm/s
        .Vz = 0
    End With
End With
myPhysics.Bodies.Add myBody
 
 
'use a gravitational force
Dim Force1 As clsGravity
Set Force1 = New clsGravity
 
myPhysics.Forces.Add Force1
Set Force1 = Nothing
 
'add a frictional force
Dim Force2 As clsFriction
Set Force2 = New clsFriction
Force2.Coeffecient = 0.5
 
myPhysics.Forces.Add Force2
Set Force2 = Nothing
 
 
 
'time step
Dim DeltaT As Single
DeltaT = 0.001 's
 
 
With myPhysics
 
    'generate new coordinates
    .Advance DeltaT
 
 
    'see new coordinates
    With .Bodies(0).Rectangular
        Debug.Print .X
        Debug.Print .Y
    End With
 
End With

I would say, mour, that that looks pretty much exactly how it should work. Perhaps a more general refinement would be to initialise each body at the origin of its own set of local coordinates, and then use coordinate transforms to get from each set to another set or the global, but as an overarching view of things that would look pretty good.

zaza

Quote:

---

Originally Posted by Harsh Gupta

you mean 2D?? i dont know if 1D objects really exists in this world?? :confused:

---

Yes.. I agree.. Objects are mostly 2D and 3D, But forces are 1D.. Gravity is a downward force.. Acceleretion is a foward force.. Deceleration is a backward force..

The only force i can think of that may be 2D is rotation.. (or Direction change..)

Quote:

---

Originally Posted by moeur

Here is how I picture the end product is going to be used. I'f I'm way off base then say so.
Something like this

---

I like how your looking at this..

This would be where i think working with 1D Force classes could apply..

we have X, X-Velocity, X-Friction, Mass, X-Acceleration :we can calculate the new X position real time...

we also have Z, Z-Velocity, Z-Friction, Mass, Z-Acceleration : and can get the New Z position..

then with Y axis we have Y, Y-Velocity, Y-Friction, Mass, Y-Acceleration, Gravity : and can get the New Y position..

In this case only the Y axis is affected by gravity.. (Looking at motion on the surface..)

so all three axis can be calculated by a single 1D function.. Gravity can be set to 0 for X and Z (and for Space flight can be used for X,Y & Z pulls)

A seperate function can calculate effective gravity according to mass and distance..

The top levels i'm talking about are our Core Basic functions like E=MCC

The middle levels should be the functions to do the calculations for the lower levels.

Gremmy..

Quote:

---

Yes.. I agree.. Objects are mostly 2D and 3D, But forces are 1D.. Gravity is a downward force.. Acceleretion is a foward force.. Deceleration is a backward force..

The only force i can think of that may be 2D is rotation.. (or Direction change..)

---

Forces are not 1-D, they are all 3-D. Sometimes a problem can be reduced to 2-D or 1-D. For example, consider the force of gravity. As you mentioned near the surface of the earth where the surface can be approximated by a flat surface gravity points in one direction, however once the surface begins to look curved then you need to add another dimension.

The problem of two bodies can be reduced to 2-D, but add a third and now you need all three dimensions.

Acceleration, decceleration and rotation are not forces.

I think what he was reffering to is the fact that vertical and horizontal motion are independent. You can therefore right one function for simple projectile motion and call it once for horizontal and once for vertical. The function is therefor a 1-D function. Correct me if I am wrong SAGremlin.

Quote:

---

Originally Posted by eyeRmonkey

I think what he was reffering to is the fact that vertical and horizontal motion are independent. You can therefore right one function for simple projectile motion and call it once for horizontal and once for vertical. The function is therefor a 1-D function. Correct me if I am wrong SAGremlin.

---

Thank you .. Some one who understands what i'm saying..

Quote:

---

Originally Posted by moeur

Acceleration, decceleration and rotation are not forces.

---

Actualy they are relative to a applied force.. Acceleration is the Application of a force to increase the relative velocity.. Deceleration is the Application of a force to decrease the relative velocity.. Ratation is a bit more advanced as it has several forces applied is several directions, that causes a circular motion rather than linear motion..

So for the purpoise of this progect we can look at acceleration as a force. And rotation as multiple forces..

Also this isn't a Plugin project for Planatary movemnet calculation, it's mostly for 3D game development. Therfore we can assume that all surfaces are flat, and that motion is linear, and not circular..

All the 3D space games that i've played, had different Graphics engines for surface and space flight, because of the different dynamics.

can anyone suggest me a good C++ book which can help me in improving my C++ skills, plus helps in coding for this project.

for this project, i mean if there is a book which exclusively helps in coding in specific area. since most books include definition and explanation of C++ function (basic) but doesnot cover special areas like "Making DLLs in C++", etc.

sorry, if i confused you, i could not make out how to explain it to you ppl.

thank you.

PS: right now i am going through C++ Black book ( :Sigh: Accelerated C++ is not available nowadays), but it doenot cover making DLLs, or making Win32 etc.

Quote:

---

You can therefore right one function for simple projectile motion and call it once for horizontal and once for vertical

---

you can always break down each vector into 2 horizontal and one vertical components if that is what you mean, but no one would call it a 1-d problem.

I think what SA was saying is that he is only concerned with situations where gravity is always pointing along one of those components (vertical) and doesn't think we should worry about the rest (i.e. ballistic missles, rocket ships and planets).

SA,

I know you understand force and acceleration, but let's not start confusing people by calling things by the wrong name. Acceleration is the result you get when you apply a force to a mass.

A constant rotation is an angular velocity and does not require force. An angular acceleration, on the otherhand, is the product of a force applied to a mass.

Quote:

---

Originally Posted by moeur

you can always break down each vector into 2 horizontal and one vertical components if that is what you mean, but no one would call it a 1-d problem.

---

But you could use the same 1D function to calculate each vector component. Thus you need 1/3rd of the code, making it easier to debug and tweek..

Quote:

---

Originally Posted by moeur

I think what SA was saying is that he is only concerned with situations where gravity is always pointing along one of those components (vertical) and doesn't think we should worry about the rest (i.e. ballistic missles, rocket ships and planets).

---

Correct.. :thumb:

Quote:

---

Originally Posted by moeur

I know you understand force and acceleration, but let's not start confusing people by calling things by the wrong name. Acceleration is the result you get when you apply a force to a mass.

---

Okay.. ;) :thumb:

Quote:

---

Originally Posted by moeur

A constant rotation is an angular velocity and does not require force. An angular acceleration, on the otherhand, is the product of a force applied to a mass.

---

This is very true.. But you need to apply forces to the object to get the rotation, we need to consider this because we need to be able to calculate the angular velocity and angular rotation..

I must appologise too.. I last used physics almost 15 years ago, and have forgotten alot of the correct terms, but I remember the theory and maths behind it all..

I'va already stated digging out alot of my old theory books to keep around as handy reference.. :D

Quote:

---

Originally Posted by SAGremlin

Also this isn't a Plugin project for Planatary movemnet calculation, it's mostly for 3D game development. Therfore we can assume that all surfaces are flat, and that motion is linear, and not circular..

---

To chip in some more cents worth, I really think that this would be a bad way to go, for a number of reasons.

1) There are plenty of people who would like to see such stuff included. FPS games are not the only 3D games out there, nor are games the only uses for 3D modelling.

2) Things such as gravity are simply specifics of a generalised force. Once a "force" calc is implemented, then it just needs a magnitude and a direction. I appreciate your point made earlier about keeping the calculations fast, but doing this by restricting so severely what the user can do is, in my opinion, definitely not the way to go. Of course there have to be limits, but they can be further out than this.

3) Sticking to linear motion and ignoring angular motion would be disaster! Nobody would want to use the engine.

4) This may, in fact, complicate things. Here is an example:

Suppose I wanted to model a juggler, moving balls around in a circle. In cartesian (XYZ) coordinates, I would need to apply exactly the correct force in moving the jugglers left hand, just the right amount of compensatory force to his right hand, offset against gravity and hope that the resultant calculation would move the balls around in a circle, rather as though he were holding two table-tennis bats and just bouncing the balls. To actually work out the starting parameters to script something like this would be a nightmare and I suspect that most of the attempts to get it to go would end up with balls flying off everywhere. This will be a major problem in all these circumstances where complex motion is involved. I recall JR's problems just getting linear friction to work on his SF2 game...
Alternatively, one could just put the balls in their own frame of reference using cylindrical polars with no z-axis movement. No force calc would be necessary, because the balls would move at steady angular velocity, and a separation of 1mm between balls and juggler hands would ensure that the hands didn't interfere whilst looking to the user as though they were causing the juggling. This would be significantly easier for the user to implement, as well as actually being quicker to calculate.

I suspect that, for such a situation, one would simply have to accept that DirectPhysics could not handle the challenge or that at best it would be rather complicated, when in fact a different programming approach would make it very easy.

Quote:

---

But you could use the same 1D function to calculate each vector component. Thus you need 1/3rd of the code, making it easier to debug and tweek..

---

Of course, I have chosen a scenario which fits my argument, but if this is to actually be a useful piece of programming then it would be better not to restrict what the user can do or make things overly complicated just for the sake of making the coding of the engine easier.
But I maintain that allowing different coordinate systems to be used, implementing a single (or multiple) "Force" object in that coordinate system (whatever it should happen to be) and then applying a characteristic(s) such as "Gravitational" or "Magnetic" to the force object(s) would be an infinitely more useful and in no way more difficult road to take. This is what moeur suggested above.


zaza

Quote:

---

Originally Posted by moeur

I agree with zaza about where we should start. However we still need to decide what the end product is going to look like so that we can work toward that goal.

Here is how I picture the end product is going to be used. I'f I'm way off base then say so.
Something like this

VB Code:

Dim myPhysics As clsDirectPhysics
Set myPhysics = New clsDirectPhysics
 
'add bodies
Dim myBody As clsDPBody
Set myBody = New clsDPBody
 
With myBody
    .Mass = 10 'kg
'set initial coordinates and velocity
    With .Rectangular
        .X = 100000 'm
        .Y = 200000 'm
        .Z = 0 'm
        .Vx = 1000 'm/s
        .Vy = 500 'm/s
        .Vz = 0
    End With
 
'Alternatively we could use spherical or cylindrical coordiantes
    With .Spherical
        .Radius = 223607 'm
        .Theta = 0.463 'radians
        .phi = 0
    End With
 
End With
myPhysics.Bodies.Add myBody
 
With myBody
    .Mass = 1000000 'kg
'set initial coordinates and velocity
    With .Rectangular
        .X = 0
        .Y = 0
        .Z = 0
        .Vx = 0 'm/s
        .Vy = 0 'm/s
        .Vz = 0
    End With
End With
myPhysics.Bodies.Add myBody
 
 
'use a gravitational force
Dim Force1 As clsGravity
Set Force1 = New clsGravity
 
myPhysics.Forces.Add Force1
Set Force1 = Nothing
 
'add a frictional force
Dim Force2 As clsFriction
Set Force2 = New clsFriction
Force2.Coeffecient = 0.5
 
myPhysics.Forces.Add Force2
Set Force2 = Nothing
 
 
 
'time step
Dim DeltaT As Single
DeltaT = 0.001 's
 
 
With myPhysics
 
    'generate new coordinates
    .Advance DeltaT
 
 
    'see new coordinates
    With .Bodies(0).Rectangular
        Debug.Print .X
        Debug.Print .Y
    End With
 
End With

---

It would look something like that, only with a DirectX style of syntax, which is where me and damasterjo will come in and sort out.

I'll be back to respond to the other posts. Got to get back to work. ;)

Nobody is working because our supervisor has not clarified what classes we need and what we are working on.

Quote:

---

Originally Posted by zaza

If I may make a suggestion.

First of all, sort out some kind of core package to deal with vectors. Include vector sums, dot products, cross products, transpose, adjoint, jacobians :) , grad, and anthing else you can think of. Matrices would also be useful.
You really won't regret it, because the last thing you want is lots of different people doing vector algebra in different ways. Get this sorted at the beginning, and the rest will just plug into it.


zaza

---

There will be a Math class in DirectPhysics that will contain math done to vectors and matrices.

Quote:

---

Originally Posted by moeur

Nobody is working because our supervisor has not clarified what classes we need and what we are working on.

---

Damn gotta go back to work. When it's lunch time, I shall show you all the classes we need to create. ;)

Moeur, bear in mind you are also a project director, so you can help design the classes too. ;)

Anyways, there's only some I can think of to create at the moment, but while we work on the project, you better believe others will be added. Here are the 7 that I thought of:

DPForce - This will calculate all of the forces applied to an object in both 2D/3D, add them up, and return the result in the Net Force.

DPSprite - This will contain some important info for a 2D/3D sprite, such as the Position, Velocity, Acceleration, Force, Mass, etc.

DPIntegration - This will contain a number of time based integration technics the user can choose from, such as Euler, Backwards Euler, Runge Kutta 4, Verlet, LeapFrog, etc.

DPRigidBody - This will contain all the functions needed to produce rigid body physics.

DPCollision - This will contain a number of different methods to test for collision for 2D/3D polygons.

DPRagdoll - Not sure if a class is needed for this, but this class will contain functions needed to produce ragdoll physics.

DPMath - This will contain some useful math functions needed for Vector math, Matrix math, Quaternions, etc., and contain some very handy constants.

And all of these classes will be a part of the main class, DirectPhysics. If anybody has any comments on any of these classes, or ideas for others, let me know. ;)

OK, let's see if I understand this.
the DirectPhysics class will contain
a collection of Sprites of type DPSprite
Each sprite will contain:

mass - scalar
Position - 3-d point
Acceleration - Vector
Velocity - Vector
Forces - collection of Vectors for local forces
NetForce - Vector

There is also a global collection of force vectors that apply to all sprites.

So to move the animation forward by one timestep we use the DPIntegration method which updates all the data in all the sprites?

I agree with what moeur said before about acceleration not being a force. If net force is equal to zero then the object is still. If the net force is not equal to zero then there is some for of acceleration or decceleration (properly called negative acceleration).

Quote:

---

Originally Posted by moeur

OK, let's see if I understand this.
the DirectPhysics class will contain
a collection of Sprites of type DPSprite
Each sprite will contain:

mass - scalar
Position - 3-d point
Acceleration - Vector
Velocity - Vector
Forces - collection of Vectors for local forces
NetForce - Vector

There is also a global collection of force vectors that apply to all sprites.

So to move the animation forward by one timestep we use the DPIntegration method which updates all the data in all the sprites?

---

That's sounds fine. I want to add that we should be very carefull about where and when each force would be applied, I think the best is giving the user a full control over that. In most games, the user player is still in the center of the screen, no real movement applied (I'm talking about his movement according to scene, not his own inner movement, like moving his hands). In this cases, the movement effect is done by applying reverse player speed to the global scene, so, maybe the DPIntegration should update all sprites except the player sprite, all the other players will be moved by DPIntegration and by its own speeds.
I'm talking just about an example about an implementation, for instances, a soccer game:
The BALL sprite, would be always in the center of the screen, no real movement, all ball movement would be applied in reverse to the whole scene, so, but when a player has the ball, he is also still there, no movement for him, but all the other players should be moved with the scene.
So, I think that we could add 1 more property to the DPsprite class that would say if a DPsprite must be affected by the global/scene/camera movement or not. Or that should be done by the user himself?

Quote:

---

Originally Posted by moeur

OK, let's see if I understand this.
the DirectPhysics class will contain
a collection of Sprites of type DPSprite
Each sprite will contain:

mass - scalar
Position - 3-d point
Acceleration - Vector
Velocity - Vector
Forces - collection of Vectors for local forces
NetForce - Vector

There is also a global collection of force vectors that apply to all sprites.

So to move the animation forward by one timestep we use the DPIntegration method which updates all the data in all the sprites?

---

DPIntegration is a class, not a method, but there will be a method in the DPIntegration class that will update all the physics data. Guess we can call it .Update, which will be similar to Direct3D's .Present method.

Quote:

---

Originally Posted by eyeRmonkey

I agree with what moeur said before about acceleration not being a force. If net force is equal to zero then the object is still. If the net force is not equal to zero then there is some for of acceleration or decceleration (properly called negative acceleration).

---

You're right, acceleration is NOT a force. It is a vector that represents a change in velocity.

Quote:

---

Originally Posted by jcis

That's sounds fine. I want to add that we should be very carefull about where and when each force would be applied, I think the best is giving the user a full control over that. In most games, the user player is still in the center of the screen, no real movement applied (I'm talking about his movement according to scene, not his own inner movement, like moving his hands). In this cases, the movement effect is done by applying reverse player speed to the global scene, so, maybe the DPIntegration should update all sprites except the player sprite, all the other players will be moved by DPIntegration and by its own speeds.
I'm talking just about an example about an implementation, for instances, a soccer game:
The BALL sprite, would be always in the center of the screen, no real movement, all ball movement would be applied in reverse to the whole scene, so, but when a player has the ball, he is also still there, no movement for him, but all the other players should be moved with the scene.
So, I think that we could add 1 more property to the DPsprite class that would say if a DPsprite must be affected by the global/scene/camera movement or not. Or that should be done by the user himself?

---

I think it might be best to have both methods, but for many sprites, you would need to create a buffer, so that 1 update call can update all the sprites data in memory. Similar to Direct3D's method that's done to polygons.

Quote:

---

Originally Posted by Jacob Roman

You're right, acceleration is NOT a force. It is a vector that represents a change in velocity.

---

Okay Now that we done Ragging Gremmy for his terible memory.. :o :p ;)

lets get this on a roll... I'm liking what you got going there... Give us more...

Gremmy..

Quote:

---

Ragging Gremmy for his terible memory..

---

thanks to me :bigyello:

@JR or moeur, could you please elaborate more on DPMaths, Force etc, the functions they include, on how we want these functions to work?

Harsh Gupta

DPMath is gonna contain some the same functions as the Direct3DX class, see here, from DirectX, only it will contain more physics based math functions, and will have just the vector, matrix, and quaternion math from the other library I have in the link. The reason why we must recreate these functions is because I don't want DirectPhysics to be dependent on DirectX in order to run the math functions. I also want to rename them to DP instead of D3DX.

Also, DPForce will contain functions that calculate Gravity, Air Resistance (aka Air Drag), Friction, Normal force, Magnetic, Applied, Buoyant, Electrical, Vacuum, Tensional, and Spring force. Torque, I'm not sure should go there or be in a separate class, but we will see. There will even be a function called Add, that will add whatever forces you apply to it, which in effect automatically puts it into Net force.

There will be 2 to 4 versions of each force, depending on whatevers possible between 1D to 3D. 4D is for when people use 'w', which is usually always 1.

has any body seen my car keys....

oh uh...

oh all this math talk is confusing me... Not really but anyway...

I downloaded visual c++ 2005 today, I couldnt even see where to type in code! I might be willing to learn, if it will benifit me in the future. I do plan on seeking employment in this field, as I have not grown bored of it. Yet...

So if anyone has anygood links on good visuall c++ 2005 , that would be great.

I figure for this project, I might not have to learn to much, but I know it will be harder than I think

Does anyone have any other questions before we begin coding? :ehh:

Ok, let's begin then. To help everyone code this, I'm gonna bring out a series of formulas for every force, or some of them for now. Some of them don't have any formula.

Table
N - Normal force
μ - Friction coefficient
m - Mass
g - Gravity
C - Drag coefficient
p - Air density
v - Velocity
A - Area
x - Position
x0 - Initial Position
k - Spring stiffness


Formulas

Gravitational: None
Friction: μ * N
Normal: m * g
Air Drag: 0.5 * C * p * A * v²
Spring: -k * (x - x0)

Not to nitpick, but:

Gravitational force: F(g) = GMm/r^2
The gravtitaional force between two objects of masses M and m, separated by r. G is the gravitational constant (very small). It acts along the line joining their centres of gravity.

Weight: F(w) = mg
For an object of mass m, the weight is mg where g is the acceleration due to gravity. On Earth, this is ~9.81ms^-2. On the local scale, this is always straight down.

Normal force: F(N) is of variable magnitude, but always acts perpendicular to the surface. When an object is on the tabletop and in equilibrium, then the Normal force is equal to the weight. However, things can get much more complicated....see diagram.

Friction: F(f) is the frictional force; usually F(f) <= uN where u is the coefficient of friction and N is the Normal force. F(f) acts to oppose the direction of motion. The frictional force does not increase forever, nor is it a constant and equal to uN.

Springs: F(s) = -kx where k is the spring constant and x is the extension of the spring. Only valid in the regime in which the spring returns to its initial length when released. Springs can be overstretched.


zaza

Actually zaza, GMm/r^2 is used to calculate the gravity of a planet or a moon, which isn't really that useful in a physics engine, and is hardly ever used.

Although Normal force and weight have the same formula m*g, they are not the same thing cause a normal force isn't always going straight down like weight is.

And in the Spring force equation -k*x, x is calculated (x - x0) to obtain the length of the spring.

Evening,

I don't wish to be a nuisance, but it's worth getting these things right.

Quote:

---

GMm/r^2 is used to calculate the gravity of a planet or a moon, which isn't really that useful in a physics engine, and is hardly ever used.

---

Actually, this is the Universal Law of gravitation. It applies to everything! For someone such as yourself, or any other object on the Earth, the gravitational force must be equal and opposite to the weight.
Hence:

mg = GMm/r^2

The m's cancel, leaving g (accel due to gravity) = GM/r^2. These are (more or less) constants, G = 6.67x10^-11 m^3/kg/s^2, M (of Earth) = 5.976x10^24 kg, radius (of Earth) = 6378137m. Fill these in, and you get g = 9.81ms^-2, exactly as we all know. So arguably, this formula is of more use than simply a constant g.

Quote:

---

Although Normal force and weight have the same formula m*g, they are not the same thing cause a normal force isn't always going straight down like weight is.

---

Having re-read my post, I realise I didn't necessarily make it clear that the Normal force usually acts in the opposite direction to the weight. However, this should be clear from the diagram. They very much are not the same thing, and the Normal force does not have the same formula as the weight.


My point about the spring equation is simply that springs are not necessarily infinitely extensible. I realise that your x-x0 is intended to represent the extension.



Can I make the further plea - please consider putting universal gravity in. Your users will honestly love you for it, and it really isn't difficult.

:)

zaza

Fine, I'll have that Gravity formula in, but since we already are gonna have many constants in our engine that will represent gravity for certain planets and our moon, not to mention the user can put in his/her custom gravity constant, I don't really see the need, but it'll be there.

By the way, weight and Normal force are the same equation, m * g. Just that Normal force has it as a vector, and optionally multiplied by an angle:

Fn.x = m * g.x
Fn.y = m * g.y
Fn.z = m * g.z

In the diagram above, for an object not on horizontal ground (as most objects will not be), the normal force is not equal and opposite to the weight. It is equal to the component of the weight force in the direction perpendicular to the surface.

It is clear in the diagram that the weight has no component in the x direction, but the Normal force does. Also, the magnitude of the Normal force is not the same as that of the weight. The motion of the object down the slope depends upon whether or not the component of the weight in the direction parallel to the slope is greatr than that of the friction in the other direction. Tilt your head to the left and it will be clearer. This also explains why the frictional force can vary - it can't be bigger than the component of the weight along the direction of the slope because otherwise the object would float uphill. It can be less, however, and so the object slides down the slope.


zaza

Okay just to break up all the serious talk here.. I have seen objects float Uphill.:D..

Seriously.. So parts here we have some hills with very strong magnetic iron ore.. You stop the car, let go of the brakes and the car Slowly rolls UPHILL.. And just to check that it is indeed uphil, we dumped a drum of water on the road(Been non magnetic), it went in the opposite direction.

A compas in that area is useless.. From about a Km (2/3 Mile) away you still pick up residule effects of the magnetic field..

Gremmy..

Sounds to me like we need to get a strong vector engine up and running before anything else.

Also, are we going to be inculding any non-physical physics (okay, that didn't make sense :p). I mean are we going to be doing anything with light or sound or anything? I kind of doubt it, but I thought I might bring it up.

If anyone is interested I attached a .DOC that I made for my physics class. It has most of the basic formulas and some stuff that we won't use at all. I think it is pretty helpful. I will update it from time to time and post it here.

I like what i see in the document..

Now we need to get these into fast code..

PS .. Jacob .. You must add that i'm the VB 6 class tester on the front page :D..

Afternoon,

I've tried to hint at it before - this time I'll be more explicit. Why not just implement the general code to express any force in any direction? To get the direction, all you'd need is the vector joining one object to another object. To get the magnitude, you can then bolt on whichever "Force calc" you like. This gives maximum flexibility to the end user whilst still being easy to code.

In order to keep the speed up and the number of calcs to a minimum, you could give an object two properties for a force, e.g. Object.EmitGravity and Object.FeelGravity. When you work out the gravitational force on objects, you only calculate the force on those with Object.FeelGravity = True and calculate the magnitude in the direction of all those objects with Object.EmitGravity = True. This will be of much use when there are large numbers of objects, because you can set the FeelGravity = True property only when a collision occurs on an object. Hence as you are driving your tank towards a stack of crates, you don't need to be constantly calculating forces to check whether they are balanced or not - just set the FeelGravity property when the collision occurs and let DirectPhysics take over.
In this regime, all you then need to do is bolt in the expression for the gravitational force between objects and call it in when needed. Do the same for magnetic, frictional, etc.
To simulate gravity, you would then just need to create a point mass at a large distance below the game area and set its EmitGravity property to True. This would be the only emitting object (so you aren't calculating tiny forces between the player and a nearby tree), and the user could easily set the size of the mass, distance away etc. And hey presto, you've built in a fully functioning gravity simulator that will cover any circumstance. The same applies for all the other forces.

Suppose I have a stack of crates in the above diagram rather than just one. What is the Normal force on the bottom crate? You have not only the weight of that crate to worry about, but all the ones stacked on top of it, hence just arbitrarily bunging in an (mg cos theta) won't be correct. You need to calculate it for each crate in relation to the other crates - and hence you need to be able to swiftly do these sums. Clearly in this circumstance, as you iterate through all the objects in the game area and calculate all the forces on each one, you need to do them in the correct order otherwise you might not work out that the top crate is resting on the bottom crate. Or you have to do all calcs simultaneously - not great for speed. Or, you implement a "contact force" whenever two objects are touching and set the flag on this just as you do for any other force. Then, when working out the net force on any given object, you can trace back through the contact forces to work out the overall net force.

These are things you need to think about now, not in 3 months when the engine is halfway done and suddenly nothing is behaving as expected.

I understand the desire to get coding, but a few hours spent now will definitely save many weeks later.

zaza

Quote:

---

Originally Posted by eyeRmonkey

Sounds to me like we need to get a strong vector engine up and running before anything else.

Also, are we going to be inculding any non-physical physics (okay, that didn't make sense :p). I mean are we going to be doing anything with light or sound or anything? I kind of doubt it, but I thought I might bring it up.

If anyone is interested I attached a .DOC that I made for my physics class. It has most of the basic formulas and some stuff that we won't use at all. I think it is pretty helpful. I will update it from time to time and post it here.

---

Light and sound physics are not used in physics engines cause they are not needed. Just done in classrooms and labs. Nice Doc by the way. ;)

Quote:

---

Originally Posted by SAGremlin

I like what i see in the document..

Now we need to get these into fast code..

PS .. Jacob .. You must add that i'm the VB 6 class tester on the front page :D..

---

Done. ;)

@ zaza:

I already planned on adding angles for any of the forces. ;)

The most important thing that is gonna be in our physics engine is the Integration class, which should be easy to code. It's gonna allow the user to choose an integration method, and will have a function to execute it. Some methods will be inaccurate and unstable, while others have their own pros and cons. It all depends on what you do. Like for example, Runge Kutta 4 would be perfect for spring simulations. While Runge Kutta methods (especially the higher order ones) are the most accurate and stable yet computationally expensive, Euler is the easiest to implement, yet is inaccurate and can easily become unstable. For future reference just incase I seem like I'm jumping the gun, here are some, which happen to be the most popular choices of integration methods:


x - Position
v - Velocity
a - Acceleration
dt - Delta Time Step Size


Euler
Aliases: Forward Euler, Explicit Euler
----------------------------------------
x = x + v * dt
v = v + a * dt

2nd Order Euler
----------------------------------------
x = x + v * dt + 0.5 * a * dt * dt
v = v + a * dt

Verlet
Aliases: Velocity-less Verlet
----------------------------------------
v = x - old_x + a * dt * dt
x = x + v
old_x = x

Time Corrected Velocity-less Verlet
----------------------------------------
v = (x - old_x) * (dt / old_dt) + a * dt * dt
x = x + v
old_x = x
old_dt = dt

Prescaled Euler
Aliases: Newton Stomer Verlet Variant
----------------------------------------
v = v + a * dt * dt
x = x + v

Re-ordered traditional Euler
Aliases: Semi-Implicit Euler, Newton Stormer Verlet, Backwards Euler
----------------------------------------
v = v + a * dt
x = x + v * dt

Velocity Verlet
----------------------------------------
x = x + v * dt + 0.5 * Old_A * dt * dt
v = v + 0.5 * (Old_A + a) * dt
Old_A = a

2nd Order Runge Kutta,
Aliases: Midpoint Method, RK2
----------------------------------------
k1 = dt * v
l1 = dt * a

k2 = dt * (v + k1 / 2)
l2 = dt * a

x = x + k2
v = v + l2

3rd Order Runge Kutta
Aliases: RK3
----------------------------------------
k1 = dt * v
l1 = dt * a

k2 = dt * (v + k1 / 2)
l2 = dt * a

k3 = dt * (v - k1 + 2 * k2)
l3 = dt * a

x = x + k1 * 1 / 6 + k2 * 2 / 3 + k3 * 1 / 6
v = v + l1 * 1 / 6 + l2 * 2 / 3 + l3 * 1 / 6

4th Order Runge Kutta
Aliases: RK4
----------------------------------------
k1 = dt * v
l1 = dt * a

k2 = dt * (v + k1 / 2)
l2 = dt * a

k3 = dt * (v + k2 / 2)
l3 = dt * a

k4 = dt * (v + k3)
l4 = dt * a

x = x + k1 / 6 + k2 / 3 + k3 / 3 + k4 / 6
v = v + l1 / 6 + l2 / 3 + l3 / 3 + l4 / 6






And here's some code examples, which show the results of each of them:

VB Code:

Option Explicit
 
Private Sub Form_Activate()
    
    Dim k1 As Single, k2 As Single, k3 As Single, k4 As Single
    Dim l1 As Single, l2 As Single, l3 As Single, l4 As Single
 
    Dim x As Single, v As Single, a As Single
 
    Dim dt As Single, dt2 As Single
    Dim f As Single, m As Single
 
    Dim old_x As Single, old_a As Single, old_dt As Single
 
    Dim i As Long
 
    a = 9.8
    dt = 1 / 100
    
    'List1.AddItem ""
    List1.AddItem "Euler Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    
    For i = 0 To 99
    
        x = x + v * dt
        v = v + a * dt
    
        List1.AddItem "X: " & x & "    " & "V: " & v
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "2nd Order Euler Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    
    For i = 0 To 99
    
        x = x + v * dt + 0.5 * a * dt * dt
        v = v + a * dt
    
        List1.AddItem "X: " & x & "    " & "V: " & v
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "Newton Stormer Verlet (Semi Implicit Euler)"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    
    For i = 0 To 99
    
        v = v + a * dt
        x = x + v * dt
    
        List1.AddItem "X: " & x & "    " & "V: " & v
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "Velocity-less Verlet Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    
    For i = 0 To 99
    
        v = x - old_x + a * dt * dt
        old_x = x
        x = x + v
    
        List1.AddItem "X: " & x & "    " & "V: " & v / dt
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "Newton Stormer Verlet Variant Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    dt2 = dt * dt
    
    For i = 0 To 99
    
        v = v + a * dt2
        x = x + v
    
        List1.AddItem "X: " & x & "    " & "V: " & v / dt
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "Velocity Verlet Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    old_a = a
    
    For i = 0 To 99
    
        x = x + v * dt + 0.5 * old_a * dt * dt
        v = v + 0.5 * (old_a + a) * dt
        old_a = a
    
        List1.AddItem "X: " & x & "    " & "V: " & v
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "Time Corrected Velocity-less Verlet Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    old_x = x
    old_dt = dt
    
    For i = 0 To 99
    
        v = x - old_x * (dt / old_dt) + a * dt * dt
        old_x = x
        old_dt = dt
        x = x + v
        
        List1.AddItem "X: " & x & "    " & "V: " & v / dt
    
    Next i
 
    List1.AddItem ""
    List1.AddItem "RK2 Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
 
    For i = 0 To 99
 
        k1 = dt * v
        l1 = dt * a
        
        k2 = dt * (v + k1 / 2)
        l2 = dt * a
        
        x = x + k2
        v = v + l2
        
        List1.AddItem "X: " & x & "    " & "V: " & v
 
    Next i
    
    List1.AddItem ""
    List1.AddItem "RK3 Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
    
    For i = 0 To 99
    
        k1 = dt * v
        l1 = dt * a
                
        k2 = dt * (v + k1 / 2)
        l2 = dt * a
                
        k3 = dt * (v - k1 + 2 * k2)
        l3 = dt * a
                
        x = x + k1 * 1 / 6 + k2 * 2 / 3 + k3 * 1 / 6
        v = v + l1 * 1 / 6 + l2 * 2 / 3 + l3 * 1 / 6
        
        List1.AddItem "X: " & x & "    " & "V: " & v
    
    Next i
    
    List1.AddItem ""
    List1.AddItem "RK4 Integration"
    List1.AddItem "---------------------"
 
    x = 0
    v = 0
 
    For i = 0 To 99
 
        k1 = dt * v
        l1 = dt * a
        
        k2 = dt * (v + k1 / 2)
        l2 = dt * a
        
        k3 = dt * (v + k2 / 2)
        l3 = dt * a
        
        k4 = dt * (v + k3)
        l4 = dt * a
        
        x = x + k1 / 6 + k2 / 3 + k3 / 3 + k4 / 6
        v = v + l1 / 6 + l2 / 3 + l3 / 3 + l4 / 6
        
        List1.AddItem "X: " & x & "    " & "V: " & v
 
    Next i
 
End Sub

All this code is good.. IN VB its slow, Or too slow for multiple calculations..
And it uses a lot of floating point calculations.. Writing this in C++ and using some ASM to do some of the FPC.. Also using Ref. tables and Aproximation helps to speed up things..

Gremmy...

sorry for no replies in past few days, but i am having some problem with my broadband connection, but i have started working on dpMatrix class (in C++), will try to post it for your views assp.

Harsh Gupta

Quote:

---

Originally Posted by SAGremlin

All this code is good.. IN VB its slow, Or too slow for multiple calculations..
And it uses a lot of floating point calculations.. Writing this in C++ and using some ASM to do some of the FPC.. Also using Ref. tables and Aproximation helps to speed up things..

Gremmy...

---

It was just an example on showing the results of each integrator. Of course it's gonna be rewritten for C++ and possibly some ASM. ;)

Quote:

---

Originally Posted by Harsh Gupta

sorry for no replies in past few days, but i am having some problem with my broadband connection, but i have started working on dpMatrix class (in C++), will try to post it for your views assp.

Harsh Gupta

---

Can't wait to see it. Hopefully it's got the same functions as DirectX's.

Speaking of no replies, it almost seemed as though everyone on the team gave up until you two replied.

Quote:

---

Originally Posted by Jacob Roman

Speaking of no replies, it almost seemed as though everyone on the team gave up until you two replied.

---

IMHO, this is the result of poor planning. You need to lay out a detailed plan and assign tasks if you want to see anything happen. Take a look at frogect rotamiton. The project was broken down into parts and each person was assigned their own task(s). You can't expect use to just read your mind and go off and get things done without some idea of what you want done.

Yes, you have given us some idea, but that is about all. If you don't break this project down into a detailed list of tasks and make goals about when you want things done then this project will remain paralyzed.

I'm not trying to be a jerk, I just want this project to succeed.