I'm coding the physics for a little platformer I'm making but I'm not sure how to design a friction calculation that does what I expect. When running along a platform I want the player's character to pick up speed but be prevented from going too fast due to friction counteracting their movement. Then, if they stop running I want them to continue to run for a little bit proportional to their speed (if they're going slow they should stop almost instantly).

void Update(V2!float forces, Elapsed time){
    position += (velocity * time.deltaS);
    forces -= 0.5 * velocity^3; //friction???
    velocity += (forces * time.deltaS);

The forces variable contains anything accelerating the object such as gravity, or acceleration from the player hitting a directional key, etc. What keeps happening is that whatever I try to plug in as friction doesn't bring the object its moving to a nice clean stop. It will look like it's stopped but then move a pixel or two after that because its velocity, while very small, doesn't hit zero for several seconds.

I can fix this with a bunch of conditional statements, but if there's a simple solution I'd much rather do that!


Did you try using a different power for the velocity or a combination of powers? Power 3 means that for small speed the friction is almost zero, eg 0.1^3=0.001. Drag from air flow is normally ~v^2 for higher and ~v for low speed so you could use a linear combination av^2+bv. Sliding friction is even independent of speed, so you could add an extra constant which will dominate the friction for low v.

  • \$\begingroup\$ Same issue even with different powers and a combination. What would a constant that dominates at low speeds consist of/look like? \$\endgroup\$
    – NovaCrist
    Mar 16 '15 at 8:34
  • \$\begingroup\$ Something like 42*(v<0?+1:-1), pick your favorite number instead of 42. OK, it has a conditional in it, but this is what nature would do. But thinking of it, in reality nature has another conditional testing for v=0 (static friction) so this gets complicated. Still I wonder why it is not working with the linear term even after a second or so. Can you give some numbers for a typical case, like initial speed, time.delta and time and "physical" distance to real standstill? \$\endgroup\$
    – Holger
    Mar 16 '15 at 9:52
  • \$\begingroup\$ time.delta is around 0.0002-0.0005 seconds as it's just a tick of the game, they should = 1 if added up over one second. With a, b, and c set to 0 there's no drag. When they're at 1 My speed caps out to velocity = 13 (pixels/s^2) (when providing +200 force) but I still have the same problem. When c = 42 it doesn't keep sliding, but it stops way too abruptly... So I think fiddling with the values I can get what I want through trial an error? forces -= ( a * velocity^2 ) + ( b * velocity ) + c; \$\endgroup\$
    – NovaCrist
    Mar 16 '15 at 11:19
  • \$\begingroup\$ Say you have force-=b*v and from user input you get a force f. The velocity should grow to v=f/b, because then the total force is be zero. Say you switch off all external forces while going at speed v. After time 4/b the velocity has fallen to 2% of v, ie it stops. And till then the thing moves by v/b pixels. For example take b=5 and force 100. Then the maximum speed should be 20 pix per sec (position is in pixels, right?). It should stop after 0.8s = 4 pixels. If this is what you observe you can fine tune the parameters. If its way off, then we should look for the problem elsewhere. \$\endgroup\$
    – Holger
    Mar 16 '15 at 21:53
  • \$\begingroup\$ Plugging in those numbers... yep, I get those results! But the third and fourth pixel it moves by are spread out so it looks like it's come to a stop after the first two, and then a slightly delayed movement occurs, and then finally a very delayed movement occurs. I think I'm going to use this rechneronline.de/function-graphs to figure out a curve that's steep in the right areas to get the effect I want. Thanks for the help! \$\endgroup\$
    – NovaCrist
    Mar 17 '15 at 9:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.