switch to klipkonns algorithm to pos calculation

This commit is contained in:
Marcel Märtens 2022-08-01 20:20:59 +02:00
parent 34c5682485
commit f17bc34883

View File

@ -295,13 +295,44 @@ fn cant_run_during_fall() -> Result<(), Box<dyn Error>> {
Ok(())
}
fn mass_c() -> (f64, f64) {
let mass = 1.0;
let air_friction_co = 0.9_f64;
let air_friction_area = 0.75_f64;
let air_density = 1.225_f64;
let c = air_friction_co * air_friction_area * 0.5 * air_density * mass;
(mass, c)
#[derive(Clone, Copy)]
struct FricParams {
friction_co: f64,
projected_area: f64,
density: f64,
mass: f64,
}
impl Default for FricParams {
fn default() -> Self {
Self {
friction_co: 0.9_f64,
projected_area: 0.75_f64,
density: 1.225_f64,
mass: 1.0_f64,
}
}
}
impl FricParams {
const MAX_ACC: f64 = 9.2;
// https://en.wikipedia.org/wiki/Drag_(physics)
// Todo: old impl was (mass * acc.abs() / (friction_co * projected_area * 0.5 *
// density * mass )).sqrt();
fn v_term(&self, acc: f64) -> f64 {
((2.0 * self.mass * acc.abs()) / (self.density * self.projected_area * self.friction_co))
.sqrt()
}
}
trait MathHelp {
fn coth(&self) -> Self;
fn acoth(&self) -> Self;
}
impl MathHelp for f64 {
// coth(x) = 1/tanh(x)`
fn coth(&self) -> Self { 1.0 / self.tanh() }
// `acoth(x) = atanh(1/x)`
fn acoth(&self) -> Self { (1.0 / self).atanh() }
}
fn test_run_helper(
@ -311,16 +342,20 @@ fn test_run_helper(
test_series: Vec<(/* acc */ f64, /* duration */ f64)>,
) -> (f64, f64) {
let dt = 1.0 / tps as f64;
let (mass, c) = mass_c();
let params = FricParams::default();
let v_term = params.v_term(FricParams::MAX_ACC);
println!("");
println!("dt: {:0>4.4} - mass: {:0>4.4} - c: {:0>4.4}", dt, mass, c);
println!(
"dt: {:0>4.4} - mass: {:0>4.4} - v_term: {:0>4.4}",
dt, params.mass, v_term
);
let (mut vel, mut pos) = (start_vel, start_pos);
let mut i = 0;
for (move_dir, duration) in test_series {
let count = (duration / dt).round() as usize;
println!("move_dir: {:0>4.4}", move_dir);
for _ in 0..count {
(vel, pos) = acc_with_frict_tick(i, move_dir, vel, pos, dt);
(vel, pos) = acc_with_frict_tick(i, move_dir, vel, pos, dt, params);
i += 1;
}
}
@ -352,12 +387,15 @@ macro_rules! veriy_diffs {
/// https://www.energie-lexikon.info/reibung.html
/// https://sciencing.com/calculate-force-friction-6454395.html
/// https://www.leifiphysik.de/mechanik/reibung-und-fortbewegung
fn acc_with_frict_tick(i: usize, move_dir: f64, vel: f64, pos: f64, dt: f64) -> (f64, f64) {
let (mass, c) = mass_c();
const MAX_ACC: f64 = 9.2;
let acc = MAX_ACC * move_dir; // btw: cant accelerate faster than gravity on foot
let old_vel = vel;
fn acc_with_frict_tick(
i: usize,
move_dir: f64,
vel: f64,
pos: f64,
dt: f64,
params: FricParams,
) -> (f64, f64) {
let acc = FricParams::MAX_ACC * move_dir; // btw: cant accelerate faster than gravity on foot
// controller
// I know what you think, wtf, yep: https://math.stackexchange.com/questions/1929436/line-integral-of-force-of-air-resistanc
@ -366,41 +404,74 @@ fn acc_with_frict_tick(i: usize, move_dir: f64, vel: f64, pos: f64, dt: f64) ->
// terminal velocity equals the maximum velocity that can be reached by acc
// alone
let vel_t = (mass * acc.abs() / c).sqrt();
let _vel_tm = (mass * MAX_ACC / c).sqrt();
let vel_t = acc.signum() * params.v_term(acc);
// if our old_vel is bigger than vel_t we can use `coth(x) = 1/tanh(x)` and
// `acoth(x) = atanh(1/x)`
let revert_fak = if vel_t != 0.0 {
old_vel / vel_t
// thanks to kilpkonn for figuring this out
// https://en.wikipedia.org/wiki/Drag_(physics)
//
// upper and lower are upper and lower bound for integral
let revert_fak = vel / vel_t;
let (vel, pos) = if acc.abs() < f64::EPSILON {
// https://www.wolframalpha.com/input?i=m%2F%28Cx+%2B+m%2FV%29+dx+integrate
let c = params.density * params.projected_area * params.friction_co;
let lower = params.mass * params.mass.ln() / c;
let upper = params.mass * (c * vel * dt + params.mass).ln() / c;
let pos = pos + (upper - lower);
let vel = params.mass
/ (params.density * params.projected_area * params.friction_co * dt
+ params.mass / vel);
(vel, pos)
} else if revert_fak <= 0.0 {
// Handle passing through 0 differently as the function changes
// https://www.wolframalpha.com/input?i=V*tan%28x*g%2FV+%2B+atan%28v%2FV%29%29+%3D+0+solve+for+x
let dt_to_zero = vel_t * (vel / vel_t).atan() / acc.abs();
if dt_to_zero < dt {
// Step with only part of dt that is left after reaching 0 vel
let lower = vel_t.powi(2) * (vel / vel_t).atan().cos().ln() / acc;
let upper =
-vel_t.powi(2) * (acc * dt_to_zero / vel_t + (vel / vel_t).atan()).cos().ln() / acc;
let pos = pos + (upper - lower);
let dt = dt - dt_to_zero;
// https://www.wolframalpha.com/input?i=V+*+tanh%28xg%2FV%29+dx+integrate
// lower bound is 0
let pos = pos + vel_t.powi(2) * (acc * dt / vel_t).cosh().ln() / acc;
let vel = vel_t * (dt * acc / vel_t).tanh();
(vel, pos)
} else {
// https://www.wolframalpha.com/input?i=V+*+tan%28xg%2FV+%2B+atan%28v%2FV%29%29+dx+integrate
let lower = -vel_t.powi(2) * (vel / vel_t).atan().cos().ln() / acc;
let upper = -vel_t.powi(2) * (acc * dt / vel_t + (vel / vel_t).atan()).cos().ln() / acc;
let pos = pos + (upper - lower);
let vel = vel_t * (dt * acc / vel_t + (vel / vel_t).atan()).tan();
(vel, pos)
}
} else if revert_fak >= 1.0 {
// https://www.wolframalpha.com/input?i=V+*+coth%28xg%2FV+%2B+acoth%28v%2FV%29%29+dx+integrate
let lower = (vel_t.powi(2) * (vel / vel_t).acoth().cosh().ln()
+ (vel / vel_t).acoth().tanh().ln())
/ acc;
let upper = (vel_t.powi(2) * (acc * dt / vel_t + (vel / vel_t).acoth()).cosh().ln()
+ (acc * dt / vel_t + (vel / vel_t).acoth()).tanh().ln())
/ acc;
let pos = pos + (upper - lower);
let vel = vel_t * (dt * acc / vel_t + (vel / vel_t).acoth()).coth();
(vel, pos)
} else {
1337.0
// https://www.wolframalpha.com/input?i=V+*+tanh%28xg%2FV+%2B+atanh%28v%2FV%29%29+dx+integrate
let lower = vel_t.powi(2) * ((vel / vel_t).atanh()).cosh().ln() / acc;
let upper = vel_t.powi(2) * (acc * dt / vel_t + (vel / vel_t).atanh()).cosh().ln() / acc;
let pos = pos + (upper - lower);
let vel = vel_t * (dt * acc / vel_t + (vel / vel_t).atanh()).tanh();
(vel, pos)
};
let vel = if revert_fak.abs() >= 1.0 {
vel_t / ((1.0 / revert_fak).atanh() + acc.signum() * (c * acc.abs()).sqrt() * dt).tanh()
} else {
vel_t * ((revert_fak).atanh() + acc.signum() * (c * acc.abs()).sqrt() * dt).tanh()
};
//physics
let distance_last = mass / c * (((old_vel * (c / acc / mass).sqrt()).atanh()).cosh()).ln();
let distance = mass / c
* (((old_vel * (c / acc / mass).sqrt()).atanh() + dt * (c * acc / mass).sqrt()).cosh())
.ln();
let diff = distance - distance_last;
let diff = if !diff.is_finite() { 0.0 } else { diff };
let pos = pos + diff;
let ending = ((i + 1) as f64 * dt * 100.0).round() as i64;
let line = format!(
"[{:0>2.2}]: acc: {:0>5.4}, revert_fak: {:0>4.4}, vel: {:0>4.4}, dist: \
{:0>7.4}, pos: {:0>7.4}",
"[{:0>2.2}]: acc: {:0>5.4}, revert_fak: {:0>4.4}, vel: {:0>4.4}, pos: {:0>7.4}",
(i + 1) as f64 * dt,
acc,
revert_fak,
vel,
distance,
pos
);
if ending % 10 != 0 {