Implementing character movement using real physics

Question

This is the first time I'm seriously developing a game so apologies if I'm missing the obvious.

In my 2D platformer I adopted Box2D since it seemed like a natural solution to model gravitational pull (player dropping off platforms) and jump arcs and in the process get collision detection for free. It was also a breeze to integrate with libgdx which I'm using as the main game engine.

One thing that I'm struggling with though is, ironically, the most simple thing: expressing player movement along the x axis using Box2D. Where previously it was as simple as translating position as a function of time, now I can only indirectly affect player position by applying a horizontal force to the rigid body representing the player, thus increasing its velocity.

I have multiple problems with this approach:

1. If I add friction to the platform fixtures, which seems natural, the user has to keep applying a force to the player for it to keep moving. However, my input system is based on the user pointing to a target platform or block upon which the character moves there, then stops.

2. If I remove the friction from any static objects the player steps on, then it's sufficient to apply force force once for the player to keep moving, but how do I tell it to stop at point x? I couldn't come up with a solution that does not feel convoluted or unnatural. I was thinking for instance to store the target position and then apply a balancing force in the opposite direction once the player reached it. Sounds complicated and prone to failure in corner cases though. I also thought about placing invisible geometry one unit ahead of the target position so that the player would collide with it and stop once reaching the position.

But seriously, both ideas sound complicated and kinda dumb. Am I missing the obvious?

I'm thinking already that maybe I should pursue a mixed approach of physics and direct controls, but I already tried that and Box2D does not like it if I set object positions manually, since this short circuits the computations behind the scenes?

Answer 1

I was going to type up an answer, but then I found this article, which seems to be exactly what you need: http://www.iforce2d.net/b2dtut/constant-speed

I'll copy a few of the relevant bits. This assumes C++, but it's certainly applicable to any language.

[...] in other situations you might want it to start and stop instantaneously. It is very tempting to use the SetLinearVelocity function to explicitly set the velocity for a body to accomplish this, and indeed it does get the job done, but this approach has its drawbacks. While it often looks fine on the screen, setting the velocity directly means the body is not correctly participating in the physics simulation.

[...]

Using SetLinearVelocity

For many applications this may be good enough. Inside the Step() function, we will take whatever action is required each time step:

    //inside Step()
    b2Vec2 vel = body->GetLinearVelocity();
    switch ( moveState )
    {
       case MOVE_LEFT:  vel.x = -5; break;
       case MOVE_STOP:  vel.x =  0; break;
       case MOVE_RIGHT: vel.x =  5; break;
    }
    body->SetLinearVelocity( vel );

[...]

Using forces

Forces are more suited to the gradual acceleration, [...] but if we make the time span very short and the force very large, we can get the same effect as an impulse. First we need to do a little math...

The relationship between force and acceleration is f = ma where m is the mass of the body we're moving, a is acceleration which is measured in "units per second per second", and f is the force we want to calculate. The acceleration could also be called "velocity per second", since velocity and "units per second" are the same thing. So we could write this as f = mv/t where t is the length of time the force will be applied.

We can get m by using the body's GetMass() function. v will be the change in velocity we desire which is the difference between the maximum speed and the current speed. To get an instantaneous speed change effect, we would be applying the force for one time step or 1/60th of a second if using the default testbed framerate. Now we know everything except f, so we do something like this:

    b2Vec2 vel = body->GetLinearVelocity();
    float desiredVel = 0;
    switch ( moveState )
    {
      case MOVE_LEFT:  desiredVel = -5; break;
      case MOVE_STOP:  desiredVel =  0; break;
      case MOVE_RIGHT: desiredVel =  5; break;
    }
    float velChange = desiredVel - vel.x;
    float force = body->GetMass() * velChange / (1/60.0); // f = mv/t
    body->ApplyForce( b2Vec2(force,0), body->GetWorldCenter() );

This should give you the same behaviour as the SetLinearVelocity did, while still remaining a realistic physics scenario.

Using impulses

Astute readers will notice that the code immediately above is basically simulating an impulse. However since impulses already take into account the length of the simulation timestep, we can just take the time part out and get the same effect with ApplyLinearImpulse:

    b2Vec2 vel = body->GetLinearVelocity();
    float desiredVel = 0;
    switch ( moveState )
    {
      case MOVE_LEFT:  desiredVel = -5; break;
      case MOVE_STOP:  desiredVel =  0; break;
      case MOVE_RIGHT: desiredVel =  5; break;
    }
    float velChange = desiredVel - vel.x;
    float impulse = body->GetMass() * velChange; // disregard time factor
    body->ApplyLinearImpulse( b2Vec2(impulse,0), body->GetWorldCenter() ); // because ApplyLinearImpulse knows the exact time factor