kshana 0.23.0

Open, reproducible PNT-resilience simulator with quantum-sensor performance models
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
// SPDX-License-Identifier: AGPL-3.0-only
//! Impulsive / finite-burn maneuvers, a Lambert two-body transfer solver, and a
//! porkchop (departure × arrival C3 / arrival-V∞) sweep — the maneuver-modeling and
//! trajectory-design beachhead.
//!
//! Scope (honest). This is the *performance-simulation* layer above a full mission-design
//! tool (GMAT / Orekit), not a replacement for one. It ships:
//!   * impulsive ΔV nodes that apply a velocity discontinuity and carry a 6×6 covariance
//!     forward (deterministic burn ⇒ identity state-transition across the instant; the
//!     stochastic execution-error covariance adds into the velocity block, rotated from the
//!     burn frame),
//!   * a finite-burn integration (constant thrust over a burn arc, mass as a state) whose
//!     achieved ΔV is checked against the closed-form **Tsiolkovsky** rocket equation,
//!   * an **Izzo-2015** single-revolution Lambert solver (`r1`, `r2`, time-of-flight ⇒
//!     `v1`, `v2` for a two-body transfer), and
//!   * a **porkchop** sweep that maps a launch-epoch × arrival-epoch grid to departure C3
//!     and arrival V∞, emitted as a 2-D JSON array for browser contour rendering.
//!
//! Validation is fully self-contained and closed-form, **stronger** than reading a value off
//! a GMAT tutorial: every Lambert output is round-tripped through an exact universal-variable
//! Kepler propagator (two-body truth to machine precision — it must land back on `r2`), the
//! finite burn is checked against Tsiolkovsky to < 0.01 %, and the porkchop minimum is checked
//! against the analytic Hohmann-transfer C3 floor for two coplanar circular orbits.
//!
//! Honest residuals (not claimed): multi-revolution (M ≥ 1) Lambert branches, a planetary
//! ephemeris (the porkchop uses a synthetic coplanar-circular heliocentric model so the
//! optimum is analytic — a real DE-ephemeris Earth–Mars C3 cross-check against GMAT needs the
//! same ephemeris and has not been run), and live wiring of the porkchop JSON into the web
//! playground contour widget. Kshana points users to GMAT/Orekit for full multi-burn /
//! low-thrust optimization.

use serde::Serialize;
use std::f64::consts::PI;

type Vec3 = [f64; 3];

/// Standard gravity used to convert specific impulse (s) to effective exhaust velocity (m/s).
pub const G0: f64 = 9.806_65;
/// Heliocentric gravitational parameter (m³/s²), IAU 2015 nominal `GM_Sun`.
pub const MU_SUN: f64 = 1.327_124_400_18e20;
/// Astronomical unit (m), IAU 2012 definition.
pub const AU_M: f64 = 1.495_978_707e11;

// ---------------------------------------------------------------------------
// Small vector helpers (kept local, mirroring the other modules' style).
// ---------------------------------------------------------------------------
fn add(a: Vec3, b: Vec3) -> Vec3 {
    [a[0] + b[0], a[1] + b[1], a[2] + b[2]]
}
fn sub(a: Vec3, b: Vec3) -> Vec3 {
    [a[0] - b[0], a[1] - b[1], a[2] - b[2]]
}
fn scale(a: Vec3, s: f64) -> Vec3 {
    [a[0] * s, a[1] * s, a[2] * s]
}
fn dot(a: Vec3, b: Vec3) -> f64 {
    a[0] * b[0] + a[1] * b[1] + a[2] * b[2]
}
fn cross(a: Vec3, b: Vec3) -> Vec3 {
    [
        a[1] * b[2] - a[2] * b[1],
        a[2] * b[0] - a[0] * b[2],
        a[0] * b[1] - a[1] * b[0],
    ]
}
fn norm(a: Vec3) -> f64 {
    dot(a, a).sqrt()
}
fn unit(a: Vec3) -> Vec3 {
    let n = norm(a);
    [a[0] / n, a[1] / n, a[2] / n]
}

// ===========================================================================
// 1. Impulsive maneuver with covariance propagation.
// ===========================================================================

/// The frame an impulsive ΔV (and its execution-error covariance) is expressed in.
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum ManeuverFrame {
    /// Inertial: ΔV components are already in the propagation (ECI) frame.
    Eci,
    /// Local-vertical/local-horizontal (radial, along-track, cross-track) at the burn state.
    Lvlh,
}

/// An instantaneous velocity change with an optional execution-error covariance.
#[derive(Clone, Copy, Debug)]
pub struct ImpulsiveManeuver {
    /// ΔV components (m/s) expressed in [`ImpulsiveManeuver::frame`].
    pub dv: Vec3,
    /// The frame `dv` and `exec_cov` are given in.
    pub frame: ManeuverFrame,
    /// 3×3 ΔV execution-error covariance (m²/s²) in the same frame (zeros ⇒ perfect burn).
    pub exec_cov: [[f64; 3]; 3],
}

/// Columns of the LVLH→ECI rotation `[r̂ | t̂ | ĥ]` at state `(r, v)`:
/// radial `r̂ = r/|r|`, cross-track `ĥ = (r×v)/|r×v|`, along-track `t̂ = ĥ×r̂` (toward `+v`).
pub fn lvlh_to_eci(r: Vec3, v: Vec3) -> [[f64; 3]; 3] {
    let ur = unit(r);
    let uh = unit(cross(r, v));
    let ut = cross(uh, ur);
    // Column-major: m[i] is the i-th basis vector expressed in ECI.
    [ur, ut, uh]
}

/// Rotate a body-frame vector `x` into ECI given the LVLH→ECI columns `m`.
fn rotate(m: &[[f64; 3]; 3], x: Vec3) -> Vec3 {
    [
        m[0][0] * x[0] + m[1][0] * x[1] + m[2][0] * x[2],
        m[0][1] * x[0] + m[1][1] * x[1] + m[2][1] * x[2],
        m[0][2] * x[0] + m[1][2] * x[1] + m[2][2] * x[2],
    ]
}

/// Apply an impulsive maneuver to a 6-state `[r(3), v(3)]` and its 6×6 covariance.
///
/// Position is continuous; velocity jumps by the (frame-resolved) ΔV. Because the burn is a
/// deterministic, state-independent shift, the state-transition across the instant is the
/// identity, so the *a priori* covariance is unchanged — only the execution-error covariance
/// (rotated into ECI) adds into the velocity–velocity block. Returns `(state_after, cov_after)`.
pub fn apply_impulse(
    state: [f64; 6],
    cov: [[f64; 6]; 6],
    man: &ImpulsiveManeuver,
) -> ([f64; 6], [[f64; 6]; 6]) {
    let r = [state[0], state[1], state[2]];
    let v = [state[3], state[4], state[5]];
    // Resolve ΔV and execution covariance into ECI.
    let (dv_eci, q_eci) = match man.frame {
        ManeuverFrame::Eci => (man.dv, man.exec_cov),
        ManeuverFrame::Lvlh => {
            let m = lvlh_to_eci(r, v);
            // Q_eci = M · exec_cov · Mᵀ.
            let mut q = [[0.0f64; 3]; 3];
            for (i, qi) in q.iter_mut().enumerate() {
                for (j, qij) in qi.iter_mut().enumerate() {
                    let mut s = 0.0;
                    for (a, ma) in man.exec_cov.iter().enumerate() {
                        for (b, &mab) in ma.iter().enumerate() {
                            // M[i][a] · exec[a][b] · M[j][b], with M[row=i] = (col a)[i].
                            s += m[a][i] * mab * m[b][j];
                        }
                    }
                    *qij = s;
                }
            }
            (rotate(&m, man.dv), q)
        }
    };
    let mut out = state;
    out[3] += dv_eci[0];
    out[4] += dv_eci[1];
    out[5] += dv_eci[2];
    let mut cov_after = cov;
    for i in 0..3 {
        for j in 0..3 {
            cov_after[3 + i][3 + j] += q_eci[i][j];
        }
    }
    (out, cov_after)
}

// ===========================================================================
// 2. Finite-burn maneuver vs the Tsiolkovsky rocket equation.
// ===========================================================================

/// A constant-thrust burn over a fixed arc, integrated with mass as a state variable.
#[derive(Clone, Copy, Debug)]
pub struct FiniteBurn {
    /// Engine thrust (N).
    pub thrust_n: f64,
    /// Specific impulse (s); effective exhaust velocity is `isp_s · G0`.
    pub isp_s: f64,
    /// Initial wet mass (kg).
    pub m0_kg: f64,
    /// Burn duration (s).
    pub burn_s: f64,
    /// Thrust direction (need not be normalized; only its direction is used).
    pub dir: Vec3,
}

/// Outcome of integrating a [`FiniteBurn`].
#[derive(Clone, Copy, Debug, Serialize)]
pub struct FiniteBurnResult {
    /// Achieved |ΔV| (m/s) from the field-free numerical integration.
    pub dv_ms: f64,
    /// Final (dry-of-burn) mass (kg).
    pub mf_kg: f64,
    /// Closed-form Tsiolkovsky ΔV = `Isp·g₀·ln(m0/mf)` (m/s).
    pub tsiolkovsky_ms: f64,
    /// Relative error |dv − tsiolkovsky| / tsiolkovsky.
    pub rel_err: f64,
}

/// Tsiolkovsky rocket equation: ΔV = `Isp · g₀ · ln(m0 / mf)` (m/s).
pub fn tsiolkovsky(isp_s: f64, m0_kg: f64, mf_kg: f64) -> f64 {
    isp_s * G0 * (m0_kg / mf_kg).ln()
}

/// Integrate a finite burn in a field-free setting (isolating the rocket-equation check from
/// gravity losses) with `steps` RK4 sub-steps. The coupled ODE on `[v(3), m]` is
/// `v̇ = (T/m)·d̂`, `ṁ = −T/(Isp·g₀)`; the achieved |ΔV| is compared to Tsiolkovsky.
///
/// Returns an error if the burn would exhaust the mass (`mf ≤ 0`).
pub fn integrate_finite_burn(burn: &FiniteBurn, steps: usize) -> Result<FiniteBurnResult, String> {
    let mdot = burn.thrust_n / (burn.isp_s * G0);
    let mf = burn.m0_kg - mdot * burn.burn_s;
    if mf <= 0.0 {
        return Err(format!(
            "burn exhausts mass: m0={} kg, ṁ={} kg/s, burn={} s ⇒ mf={} kg",
            burn.m0_kg, mdot, burn.burn_s, mf
        ));
    }
    let d = unit(burn.dir);
    let h = burn.burn_s / steps as f64;
    // y = [vx, vy, vz, m]; field-free, so v̇ = (T/m)·d̂, ṁ = −mdot.
    let deriv = |_t: f64, y: &[f64]| -> Vec<f64> {
        let m = y[3];
        let a = burn.thrust_n / m;
        vec![a * d[0], a * d[1], a * d[2], -mdot]
    };
    let mut y = vec![0.0, 0.0, 0.0, burn.m0_kg];
    let mut t = 0.0;
    for _ in 0..steps {
        y = crate::integrator::rk4_step(&deriv, t, &y, h);
        t += h;
    }
    let dv_ms = norm([y[0], y[1], y[2]]);
    let tsiol = tsiolkovsky(burn.isp_s, burn.m0_kg, mf);
    let rel_err = (dv_ms - tsiol).abs() / tsiol;
    Ok(FiniteBurnResult {
        dv_ms,
        mf_kg: mf,
        tsiolkovsky_ms: tsiol,
        rel_err,
    })
}

// ===========================================================================
// 3. Exact universal-variable Kepler propagator (two-body truth for validation).
// ===========================================================================

/// Stumpff functions `c2(ψ)`, `c3(ψ)` (series near zero, closed form otherwise).
fn stumpff(psi: f64) -> (f64, f64) {
    if psi > 1e-6 {
        let s = psi.sqrt();
        ((1.0 - s.cos()) / psi, (s - s.sin()) / (psi * s))
    } else if psi < -1e-6 {
        let s = (-psi).sqrt();
        ((s.cosh() - 1.0) / (-psi), (s.sinh() - s) / ((-psi) * s))
    } else {
        // Series: c2 = 1/2 − ψ/24 + ψ²/720…, c3 = 1/6 − ψ/120 + ψ²/5040…
        (
            0.5 - psi / 24.0 + psi * psi / 720.0,
            1.0 / 6.0 - psi / 120.0 + psi * psi / 5040.0,
        )
    }
}

/// Propagate a two-body state `(r0, v0)` forward by `dt` under gravity parameter `mu`
/// using the universal-variable (Stumpff f/g) formulation. Exact two-body to convergence —
/// the reference truth the Lambert outputs are round-tripped against.
pub fn kepler_universal(r0: Vec3, v0: Vec3, dt: f64, mu: f64) -> (Vec3, Vec3) {
    let sqrt_mu = mu.sqrt();
    let r0n = norm(r0);
    let v0n = norm(v0);
    let rv0 = dot(r0, v0); // r0·v0 (NOT divided by |r0| — the universal-variable term needs r0·v0/√μ)
    let alpha = 2.0 / r0n - v0n * v0n / mu; // 1/a
                                            // Initial χ guess.
    let mut chi = if alpha > 1e-9 {
        sqrt_mu * dt * alpha
    } else if alpha < -1e-9 {
        let a = 1.0 / alpha;
        dt.signum()
            * (-a).sqrt()
            * ((-2.0 * mu * alpha * dt)
                / (dot(r0, v0) + dt.signum() * (-mu * a).sqrt() * (1.0 - r0n * alpha)))
                .ln()
    } else {
        sqrt_mu * dt / r0n
    };
    for _ in 0..100 {
        let psi = chi * chi * alpha;
        let (c2, c3) = stumpff(psi);
        let r = chi * chi * c2 + (rv0 / sqrt_mu) * chi * (1.0 - psi * c3) + r0n * (1.0 - psi * c2);
        let f = sqrt_mu * dt
            - chi * chi * chi * c3
            - (rv0 / sqrt_mu) * chi * chi * c2
            - r0n * chi * (1.0 - psi * c3);
        let dchi = f / r;
        chi += dchi;
        if dchi.abs() < 1e-10 {
            break;
        }
    }
    let psi = chi * chi * alpha;
    let (c2, c3) = stumpff(psi);
    let f = 1.0 - chi * chi / r0n * c2;
    let g = dt - chi * chi * chi / sqrt_mu * c3;
    let r_vec = add(scale(r0, f), scale(v0, g));
    let rn = norm(r_vec);
    let fdot = sqrt_mu / (rn * r0n) * chi * (psi * c3 - 1.0);
    let gdot = 1.0 - chi * chi / rn * c2;
    let v_vec = add(scale(r0, fdot), scale(v0, gdot));
    (r_vec, v_vec)
}

// ===========================================================================
// 4. Izzo-2015 single-revolution Lambert solver.
// ===========================================================================

/// Gauss hypergeometric `₂F₁(3, 1; 5/2; x)` by its power series (Izzo's `hyp2f1b`),
/// used for the time-of-flight series near `x = 1`.
fn hyp2f1b(x: f64) -> f64 {
    if x >= 1.0 {
        return f64::INFINITY;
    }
    let mut res = 1.0_f64;
    let mut term = 1.0_f64;
    let mut i = 0.0_f64;
    loop {
        term *= (3.0 + i) * (1.0 + i) / (2.5 + i) * x / (i + 1.0);
        let newres = res + term;
        if newres == res {
            return newres;
        }
        res = newres;
        i += 1.0;
    }
}

fn compute_y(x: f64, ll: f64) -> f64 {
    (1.0 - ll * ll * (1.0 - x * x)).sqrt()
}

fn compute_psi(x: f64, y: f64, ll: f64) -> f64 {
    if (-1.0..1.0).contains(&x) {
        // Elliptic.
        (x * y + ll * (1.0 - x * x)).acos()
    } else if x > 1.0 {
        // Hyperbolic.
        ((y - x * ll) * (x * x - 1.0).sqrt()).asinh()
    } else {
        0.0
    }
}

/// `tof(x) − T0` for the (non-dimensional) Lambert time equation, single revolution.
fn tof_equation_y(x: f64, y: f64, t0: f64, ll: f64) -> f64 {
    let tof = if (0.6_f64).sqrt() < x && x < (1.4_f64).sqrt() {
        let eta = y - ll * x;
        let s1 = 0.5 * (1.0 - ll - x * eta);
        let q = 4.0 / 3.0 * hyp2f1b(s1);
        0.5 * (eta * eta * eta * q + 4.0 * ll * eta)
    } else {
        let psi = compute_psi(x, y, ll);
        (psi / (1.0 - x * x).abs().sqrt() - x + ll * y) / (1.0 - x * x)
    };
    tof - t0
}

fn tof_eq_p(x: f64, y: f64, tof: f64, ll: f64) -> f64 {
    (3.0 * tof * x - 2.0 + 2.0 * ll * ll * ll * x / y) / (1.0 - x * x)
}
fn tof_eq_p2(x: f64, y: f64, tof: f64, dt: f64, ll: f64) -> f64 {
    (3.0 * tof + 5.0 * x * dt + 2.0 * (1.0 - ll * ll) * ll * ll * ll / (y * y * y)) / (1.0 - x * x)
}
fn tof_eq_p3(x: f64, y: f64, _tof: f64, dt: f64, ddt: f64, ll: f64) -> f64 {
    (7.0 * x * ddt + 8.0 * dt - 6.0 * (1.0 - ll * ll) * ll.powi(5) * x / y.powi(5)) / (1.0 - x * x)
}

/// Householder (cubic-order) iteration for the single-rev root `x` of `tof(x) = T`.
fn householder(x0: f64, big_t: f64, ll: f64) -> f64 {
    let mut x = x0;
    for _ in 0..35 {
        let y = compute_y(x, ll);
        let fval = tof_equation_y(x, y, big_t, ll);
        let tof = fval + big_t;
        let fp = tof_eq_p(x, y, tof, ll);
        let fpp = tof_eq_p2(x, y, tof, fp, ll);
        let fppp = tof_eq_p3(x, y, tof, fp, fpp, ll);
        let denom = fp * (fp * fp - fval * fpp) + fppp * fval * fval / 6.0;
        let dx = fval * (fp * fp - fval * fpp / 2.0) / denom;
        x -= dx;
        if dx.abs() < 1e-12 {
            break;
        }
    }
    x
}

/// Solve Lambert's problem (single revolution, `M = 0`) by Izzo's 2015 method.
///
/// Given two position vectors `r1`, `r2`, the time of flight `tof` (s), gravity parameter `mu`,
/// and a `prograde` flag (transfer with `+z` angular momentum), returns the departure and
/// arrival velocities `(v1, v2)` of the connecting two-body arc. Errors on degenerate geometry
/// (collinear `r1`, `r2` ⇒ undefined transfer plane) or non-positive time of flight.
pub fn lambert(
    r1: Vec3,
    r2: Vec3,
    tof: f64,
    mu: f64,
    prograde: bool,
) -> Result<(Vec3, Vec3), String> {
    if tof <= 0.0 {
        return Err("time of flight must be positive".into());
    }
    let r1n = norm(r1);
    let r2n = norm(r2);
    let c_vec = sub(r2, r1);
    let c = norm(c_vec);
    let s = 0.5 * (r1n + r2n + c);
    let ir1 = unit(r1);
    let ir2 = unit(r2);
    let cross12 = cross(ir1, ir2);
    let cross_n = norm(cross12);
    if cross_n < 1e-12 {
        return Err("collinear endpoints: transfer plane is undefined (≈0°/180°)".into());
    }
    let ih = scale(cross12, 1.0 / cross_n);
    let mut ll = (1.0 - (c / s).min(1.0)).sqrt();
    let (mut it1, mut it2);
    if ih[2] < 0.0 {
        ll = -ll;
        it1 = cross(ir1, ih);
        it2 = cross(ir2, ih);
    } else {
        it1 = cross(ih, ir1);
        it2 = cross(ih, ir2);
    }
    if !prograde {
        ll = -ll;
        it1 = scale(it1, -1.0);
        it2 = scale(it2, -1.0);
    }
    it1 = unit(it1);
    it2 = unit(it2);

    let big_t = (2.0 * mu / (s * s * s)).sqrt() * tof;

    // Single-revolution initial guess (Izzo `_initial_guess`, M = 0 branch).
    let t0 = ll.acos() + ll * (1.0 - ll * ll).sqrt(); // T(x = 0)
    let t1 = 2.0 / 3.0 * (1.0 - ll * ll * ll); // T(x = 1), parabola
    let x0 = if big_t >= t0 {
        (t0 / big_t).powf(2.0 / 3.0) - 1.0
    } else if big_t < t1 {
        2.5 * t1 / big_t * (t1 - big_t) / (1.0 - ll.powi(5)) + 1.0
    } else {
        (t0 / big_t).powf((t1 / t0).log2()) - 1.0
    };
    let x = householder(x0, big_t, ll);
    let y = compute_y(x, ll);

    let gamma = (mu * s / 2.0).sqrt();
    let rho = (r1n - r2n) / c;
    let sigma = (1.0 - rho * rho).sqrt();
    let vr1 = gamma * ((ll * y - x) - rho * (ll * y + x)) / r1n;
    let vr2 = -gamma * ((ll * y - x) + rho * (ll * y + x)) / r2n;
    let vt1 = gamma * sigma * (y + ll * x) / r1n;
    let vt2 = gamma * sigma * (y + ll * x) / r2n;
    let v1 = add(scale(ir1, vr1), scale(it1, vt1));
    let v2 = add(scale(ir2, vr2), scale(it2, vt2));
    Ok((v1, v2))
}

// ===========================================================================
// 5. Porkchop sweep over a synthetic coplanar-circular heliocentric system.
// ===========================================================================

/// A body on a coplanar circular heliocentric orbit (synthetic, so the transfer optimum is
/// analytic). `phase0_rad` is its true longitude at epoch `t = 0`.
#[derive(Clone, Copy, Debug)]
pub struct CircularCoplanarBody {
    /// Orbit radius (m).
    pub radius_m: f64,
    /// True longitude at `t = 0` (rad).
    pub phase0_rad: f64,
    /// Central gravity parameter (m³/s²).
    pub mu_central: f64,
}

impl CircularCoplanarBody {
    /// Mean angular rate `n = √(μ/r³)` (rad/s).
    pub fn angular_rate(&self) -> f64 {
        (self.mu_central / self.radius_m.powi(3)).sqrt()
    }
    /// Heliocentric position (m) at time `t` (s).
    pub fn position(&self, t: f64) -> Vec3 {
        let th = self.phase0_rad + self.angular_rate() * t;
        [self.radius_m * th.cos(), self.radius_m * th.sin(), 0.0]
    }
    /// Heliocentric velocity (m/s) at time `t` (s).
    pub fn velocity(&self, t: f64) -> Vec3 {
        let n = self.angular_rate();
        let th = self.phase0_rad + n * t;
        let speed = self.radius_m * n;
        [-speed * th.sin(), speed * th.cos(), 0.0]
    }
}

/// A porkchop grid: departure C3 (km²/s²) and arrival V∞ (km/s) over launch × arrival epochs.
/// Degenerate cells (non-positive TOF or near-collinear ≈0°/180° geometry) are stored as NaN.
#[derive(Clone, Debug, Serialize)]
pub struct PorkchopGrid {
    /// Departure epochs (s).
    pub dep_epochs_s: Vec<f64>,
    /// Arrival epochs (s).
    pub arr_epochs_s: Vec<f64>,
    /// `c3_km2s2[i][j]` = departure C3 for `dep_epochs_s[i]` → `arr_epochs_s[j]`.
    pub c3_km2s2: Vec<Vec<f64>>,
    /// `vinf_arr_kms[i][j]` = arrival hyperbolic-excess speed for the same cell.
    pub vinf_arr_kms: Vec<Vec<f64>>,
}

impl PorkchopGrid {
    /// Serialize the grid (epoch axes + the two 2-D arrays) for browser contour rendering.
    pub fn to_json(&self) -> String {
        // `PorkchopGrid` is only `Vec<f64>` / `Vec<Vec<f64>>` fields — no maps and no
        // fallible custom `Serialize` impls — so JSON serialisation cannot fail. (A
        // non-finite C3/v∞ entry serialises to `null`; it does not error.)
        serde_json::to_string_pretty(self)
            .expect("PorkchopGrid (only Vec<f64>/Vec<Vec<f64>> fields) always serialises")
    }
    /// The minimum finite C3 over the grid and its `(dep_index, arr_index)`.
    pub fn min_c3(&self) -> Option<(f64, usize, usize)> {
        let mut best: Option<(f64, usize, usize)> = None;
        for (i, row) in self.c3_km2s2.iter().enumerate() {
            for (j, &c3) in row.iter().enumerate() {
                let better = match best {
                    None => true,
                    Some((b, _, _)) => c3 < b,
                };
                if c3.is_finite() && better {
                    best = Some((c3, i, j));
                }
            }
        }
        best
    }
}

/// Sweep a launch-epoch × arrival-epoch grid, solving Lambert for each cell and recording the
/// departure C3 (km²/s²) and arrival V∞ (km/s). Cells within `0.5°` of a 0°/180° transfer (or
/// with TOF ≤ 0) are marked NaN — the transfer plane is ill-conditioned there.
pub fn porkchop(
    dep: &CircularCoplanarBody,
    arr: &CircularCoplanarBody,
    dep_epochs_s: &[f64],
    arr_epochs_s: &[f64],
    mu_helio: f64,
) -> PorkchopGrid {
    let collinear_guard = (0.5_f64).to_radians();
    let mut c3 = vec![vec![f64::NAN; arr_epochs_s.len()]; dep_epochs_s.len()];
    let mut vinf = vec![vec![f64::NAN; arr_epochs_s.len()]; dep_epochs_s.len()];
    for (i, &td) in dep_epochs_s.iter().enumerate() {
        for (j, &ta) in arr_epochs_s.iter().enumerate() {
            let tof = ta - td;
            if tof <= 0.0 {
                continue;
            }
            let r1 = dep.position(td);
            let r2 = arr.position(ta);
            // Skip near-collinear geometry.
            let cosang = (dot(r1, r2) / (norm(r1) * norm(r2))).clamp(-1.0, 1.0);
            let ang = cosang.acos();
            if ang < collinear_guard || (PI - ang) < collinear_guard {
                continue;
            }
            if let Ok((v1, v2)) = lambert(r1, r2, tof, mu_helio, true) {
                let vinf_dep = sub(v1, dep.velocity(td));
                let vinf_arr = sub(v2, arr.velocity(ta));
                c3[i][j] = dot(vinf_dep, vinf_dep) / 1.0e6; // m²/s² → km²/s²
                vinf[i][j] = norm(vinf_arr) / 1000.0; // m/s → km/s
            }
        }
    }
    PorkchopGrid {
        dep_epochs_s: dep_epochs_s.to_vec(),
        arr_epochs_s: arr_epochs_s.to_vec(),
        c3_km2s2: c3,
        vinf_arr_kms: vinf,
    }
}

/// Closed-form departure C3 (m²/s²) of the Hohmann transfer between two coplanar circular
/// orbits `r1 → r2` about `mu`: `C3 = (v_transfer,peri − v_circ,1)²`. This is the global
/// minimum-energy departure for circular-to-circular transfer — the porkchop floor.
pub fn hohmann_departure_c3(r1: f64, r2: f64, mu: f64) -> f64 {
    let a_t = 0.5 * (r1 + r2);
    let v_circ1 = (mu / r1).sqrt();
    let v_peri = (mu * (2.0 / r1 - 1.0 / a_t)).sqrt();
    let vinf = v_peri - v_circ1;
    vinf * vinf
}

/// Time of flight (s) of the Hohmann transfer `r1 → r2` about `mu`: half the transfer-ellipse
/// period, `π·√(a_t³/μ)` with `a_t = (r1+r2)/2`.
pub fn hohmann_tof(r1: f64, r2: f64, mu: f64) -> f64 {
    let a_t = 0.5 * (r1 + r2);
    PI * (a_t * a_t * a_t / mu).sqrt()
}

#[cfg(test)]
mod tests {
    use super::*;
    use std::f64::consts::PI;

    fn approx(a: f64, b: f64, tol: f64) {
        assert!((a - b).abs() <= tol, "expected {a} ≈ {b} (tol {tol})");
    }

    // ----- 1. Impulsive maneuver + covariance -----

    #[test]
    fn impulsive_eci_dv_jumps_velocity_and_adds_only_velocity_covariance() {
        let state = [7.0e6, 0.0, 0.0, 0.0, 7.5e3, 0.0];
        let mut cov = [[0.0f64; 6]; 6];
        for (i, row) in cov.iter_mut().enumerate() {
            row[i] = (i as f64 + 1.0) * 10.0; // distinct diagonal
        }
        let man = ImpulsiveManeuver {
            dv: [10.0, -5.0, 2.0],
            frame: ManeuverFrame::Eci,
            exec_cov: [[0.04, 0.0, 0.0], [0.0, 0.09, 0.0], [0.0, 0.0, 0.01]],
        };
        let (out, p) = apply_impulse(state, cov, &man);
        // Position unchanged, velocity jumps by ΔV exactly.
        approx(out[0], 7.0e6, 0.0);
        approx(out[3], 10.0, 1e-12);
        approx(out[4], 7.5e3 - 5.0, 1e-9);
        approx(out[5], 2.0, 1e-12);
        // Position block of P unchanged; velocity block gains exactly exec_cov.
        for i in 0..3 {
            for j in 0..3 {
                approx(p[i][j], cov[i][j], 1e-12);
            }
        }
        approx(p[3][3], cov[3][3] + 0.04, 1e-12);
        approx(p[4][4], cov[4][4] + 0.09, 1e-12);
        approx(p[5][5], cov[5][5] + 0.01, 1e-12);
    }

    #[test]
    fn lvlh_execution_covariance_is_rotated_trace_preserving_and_symmetric() {
        // A non-trivial inclined state so the LVLH frame is genuinely rotated.
        let state = [7.0e6, 1.0e6, 5.0e5, 1.0e3, 7.2e3, 1.5e3];
        let cov = [[0.0f64; 6]; 6];
        let (a, b, c) = (0.05, 0.20, 0.02);
        let man = ImpulsiveManeuver {
            dv: [3.0, 0.0, 0.0],
            frame: ManeuverFrame::Lvlh,
            exec_cov: [[a, 0.0, 0.0], [0.0, b, 0.0], [0.0, 0.0, c]],
        };
        let (_out, p) = apply_impulse(state, cov, &man);
        // Rotation preserves the trace of the covariance added to the velocity block.
        let tr = p[3][3] + p[4][4] + p[5][5];
        approx(tr, a + b + c, 1e-12);
        // The added velocity-block covariance is symmetric.
        approx(p[3][4], p[4][3], 1e-15);
        approx(p[3][5], p[5][3], 1e-15);
        approx(p[4][5], p[5][4], 1e-15);
        // And the LVLH→ECI columns are orthonormal.
        let r = [state[0], state[1], state[2]];
        let v = [state[3], state[4], state[5]];
        let m = lvlh_to_eci(r, v);
        for col in &m {
            approx(norm(*col), 1.0, 1e-12);
        }
        approx(dot(m[0], m[1]), 0.0, 1e-12);
        approx(dot(m[0], m[2]), 0.0, 1e-12);
        approx(dot(m[1], m[2]), 0.0, 1e-12);
    }

    // ----- 2. Finite burn vs Tsiolkovsky -----

    #[test]
    fn finite_burn_matches_tsiolkovsky_to_better_than_a_hundredth_percent() {
        // Hand check: ṁ = 500/(300·9.80665) = 0.169970 kg/s; mf = 1000 − 16.9970 = 983.003 kg;
        // Δv = 300·9.80665·ln(1000/983.003) = 2941.995·0.0171419 ≈ 50.43 m/s.
        let burn = FiniteBurn {
            thrust_n: 500.0,
            isp_s: 300.0,
            m0_kg: 1000.0,
            burn_s: 100.0,
            dir: [0.0, 1.0, 0.0],
        };
        let res = integrate_finite_burn(&burn, 2000).unwrap();
        let mdot = 500.0 / (300.0 * G0);
        approx(res.mf_kg, 1000.0 - mdot * 100.0, 1e-9);
        approx(
            res.tsiolkovsky_ms,
            tsiolkovsky(300.0, 1000.0, res.mf_kg),
            0.0,
        );
        // Closed-form sanity on the magnitude itself.
        approx(res.tsiolkovsky_ms, 50.43, 0.05);
        // RK4 of v̇ = T/m reproduces the exact rocket equation to well under 0.01 %.
        assert!(
            res.rel_err < 1e-4,
            "finite-burn ΔV rel err {} ≥ 1e-4",
            res.rel_err
        );
    }

    #[test]
    fn finite_burn_errors_when_mass_is_exhausted() {
        let burn = FiniteBurn {
            thrust_n: 5000.0,
            isp_s: 200.0,
            m0_kg: 100.0,
            burn_s: 100.0,
            dir: [1.0, 0.0, 0.0],
        };
        assert!(integrate_finite_burn(&burn, 100).is_err());
    }

    // ----- 3. Kepler propagator self-check -----

    #[test]
    fn kepler_universal_round_trips_forward_then_back() {
        let r0 = [7.0e6, 0.0, 0.0];
        let v0 = [0.0, 8.0e3, 1.0e3];
        let mu = crate::orbit::MU_EARTH;
        let (r1, v1) = kepler_universal(r0, v0, 2000.0, mu);
        let (r2, v2) = kepler_universal(r1, v1, -2000.0, mu);
        approx(norm(sub(r2, r0)), 0.0, 1e-3);
        approx(norm(sub(v2, v0)), 0.0, 1e-6);
    }

    // ----- 4. Lambert: round-trip against exact two-body truth -----

    #[test]
    fn lambert_recovers_the_velocities_of_a_known_two_body_arc() {
        let mu = crate::orbit::MU_EARTH;
        let r1 = [7.0e6, 0.0, 0.0];
        let v1_true = [0.0, 8.0e3, 1.0e3]; // h = r×v has +z ⇒ prograde
        let tof = 2000.0;
        // Exact two-body truth: where does this arc end, and at what velocity?
        let (r2, v2_true) = kepler_universal(r1, v1_true, tof, mu);
        // Lambert must reconstruct both boundary velocities.
        let (v1, v2) = lambert(r1, r2, tof, mu, true).unwrap();
        approx(norm(sub(v1, v1_true)), 0.0, 1e-3);
        approx(norm(sub(v2, v2_true)), 0.0, 1e-3);
    }

    #[test]
    fn lambert_output_propagates_back_onto_r2() {
        // Independent of the orbit fixture: any Lambert solution, when propagated, must hit r2.
        let mu = MU_SUN;
        let r1 = [1.0 * AU_M, 0.2 * AU_M, 0.0];
        let r2 = [-0.3 * AU_M, 1.3 * AU_M, 0.05 * AU_M];
        let tof = 200.0 * 86400.0;
        let (v1, v2) = lambert(r1, r2, tof, mu, true).unwrap();
        let (r_end, v_end) = kepler_universal(r1, v1, tof, mu);
        approx(norm(sub(r_end, r2)), 0.0, 1.0e3); // < 1 km on a ~1.5e11 m arc
        approx(norm(sub(v_end, v2)), 0.0, 1e-3);
    }

    #[test]
    fn lambert_rejects_collinear_endpoints() {
        let mu = crate::orbit::MU_EARTH;
        let r1 = [7.0e6, 0.0, 0.0];
        let r2 = [-9.0e6, 0.0, 0.0]; // exactly 180°, plane undefined
        assert!(lambert(r1, r2, 3000.0, mu, true).is_err());
    }

    // ----- 5. Porkchop: round-trip consistency + Hohmann floor -----

    #[test]
    fn porkchop_cells_are_round_trip_consistent_and_min_is_near_the_hohmann_floor() {
        // Synthetic coplanar-circular heliocentric system: 1.0 AU → 1.524 AU.
        let r1 = AU_M;
        let r2 = 1.524 * AU_M;
        let tof_h = hohmann_tof(r1, r2, MU_SUN);
        let dep = CircularCoplanarBody {
            radius_m: r1,
            phase0_rad: 0.0,
            mu_central: MU_SUN,
        };
        // Place the target so a 0 → tof_h transfer sweeps ~180° (the Hohmann geometry),
        // but we never sample exactly there (the guard / grid offsets keep us off the singularity).
        let n2 = (MU_SUN / r2.powi(3)).sqrt();
        let arr = CircularCoplanarBody {
            radius_m: r2,
            phase0_rad: PI - n2 * tof_h,
            mu_central: MU_SUN,
        };
        let day = 86400.0;
        let dep_epochs: Vec<f64> = (-20..=20).map(|k| k as f64 * 5.0 * day).collect();
        let arr_epochs: Vec<f64> = (0..=40)
            .map(|k| tof_h - 100.0 * day + k as f64 * 5.0 * day)
            .collect();
        let grid = porkchop(&dep, &arr, &dep_epochs, &arr_epochs, MU_SUN);

        // Every finite cell's departure velocity must propagate two-body back onto r2.
        let mut finite_cells = 0;
        for (i, &td) in dep_epochs.iter().enumerate() {
            for (j, &ta) in arr_epochs.iter().enumerate() {
                if grid.c3_km2s2[i][j].is_finite() {
                    finite_cells += 1;
                    let tof = ta - td;
                    let rr1 = dep.position(td);
                    let rr2 = arr.position(ta);
                    let (v1, _) = lambert(rr1, rr2, tof, MU_SUN, true).unwrap();
                    let (r_end, _) = kepler_universal(rr1, v1, tof, MU_SUN);
                    assert!(
                        norm(sub(r_end, rr2)) < 5.0e3,
                        "cell ({i},{j}) round-trip miss {} m",
                        norm(sub(r_end, rr2))
                    );
                }
            }
        }
        assert!(finite_cells > 100, "too few finite cells: {finite_cells}");

        // The closed-form Hohmann departure C3 is the theoretical floor.
        let c3_floor = hohmann_departure_c3(r1, r2, MU_SUN) / 1.0e6; // km²/s²
        let (min_c3, mi, mj) = grid.min_c3().unwrap();
        // Floor property: no cell beats Hohmann (small numerical slack).
        assert!(
            min_c3 >= c3_floor - 1e-6,
            "grid min C3 {min_c3} below Hohmann floor {c3_floor}"
        );
        // The finite grid samples near (but not at) 180°, so the minimum is just above the floor.
        assert!(
            min_c3 < c3_floor * 1.05,
            "grid min C3 {min_c3} not within 5% of Hohmann floor {c3_floor}"
        );
        // The minimizing cell's TOF is near the Hohmann TOF.
        let tof_min = arr_epochs[mj] - dep_epochs[mi];
        approx(tof_min, tof_h, 60.0 * day);
    }

    #[test]
    fn porkchop_json_round_trips_with_matching_dimensions() {
        let dep = CircularCoplanarBody {
            radius_m: AU_M,
            phase0_rad: 0.0,
            mu_central: MU_SUN,
        };
        let arr = CircularCoplanarBody {
            radius_m: 1.524 * AU_M,
            phase0_rad: 1.0,
            mu_central: MU_SUN,
        };
        let day = 86400.0;
        let dep_epochs: Vec<f64> = (0..5).map(|k| k as f64 * 10.0 * day).collect();
        let arr_epochs: Vec<f64> = (0..6).map(|k| (200 + k * 10) as f64 * day).collect();
        let grid = porkchop(&dep, &arr, &dep_epochs, &arr_epochs, MU_SUN);
        let json = grid.to_json();
        let parsed: serde_json::Value = serde_json::from_str(&json).unwrap();
        assert_eq!(parsed["dep_epochs_s"].as_array().unwrap().len(), 5);
        assert_eq!(parsed["arr_epochs_s"].as_array().unwrap().len(), 6);
        assert_eq!(parsed["c3_km2s2"].as_array().unwrap().len(), 5);
        assert_eq!(parsed["c3_km2s2"][0].as_array().unwrap().len(), 6);
        assert_eq!(parsed["vinf_arr_kms"].as_array().unwrap().len(), 5);
    }
}