Skip to content

Rocket Launch Simulation

Jump to bottom
Wiki Sync edited this page Mar 12, 2026 · 2 revisions

Rocket Launch Simulation

RocketPy (6-DOF) + custom orbital mechanics solver + trajectory optimization + powered descent guidance.

Overview

Simulation Tool Launch Site Result
Calisto (HPR) RocketPy 6-DOF Spaceport America, NM Apogee 3,305 m, Mach 0.86
H3-Inspired Sounding RocketPy 6-DOF Tanegashima (Yoshinobu LP-2) Apogee 3,341 m, Mach 0.87
H3-22S Orbital Custom gravity-turn Tanegashima (Yoshinobu LP-2) LEO 293x365 km, e=0.005
H3-22S Optimized scipy.optimize — MECO T+288s/203km (err 3.5%)
Powered Descent G-FOLD (cvxpy SOCP) — 1,523 kg fuel, 0.0 m/s landing
RL Landing PPO (PyTorch) — Training (GPU required)
Orbital Transfer poliastro — GTO Δv=2.43 km/s, TLI Δv=3.11 km/s

H3-22S Orbital Insertion (Optimized)

Vehicle Configuration

Component Thrust Isp Propellant Burn Time
SRB-3 (x2) 3,200 kN (avg) 280 s 132 ton 114 s
1st Stage (LE-9 x2) 2,942 kN 425 s (vac) 209 ton ~288 s
2nd Stage (LE-5B-3) 137 kN 448 s 28 ton ~470 s
Total Liftoff 6,142 kN 423 ton T/W = 1.48

Pitch Profile Optimization (Step A+B)

scipy.optimize.differential_evolution で7パラメータのピッチプログラムを最適化。 S1推進薬を180t→209t に修正（実機222t に対して最適化で推定）。

Optimization Results

Cost function: 0.765 → 0.070 (91% reduction)

Flight Timeline vs Real H3

Event Before Optimized Real H3 (TF2/3) Improvement
SRB-3 Separation T+114s, 54 km T+114s, 47 km T+116s, ~50 km Alt: 54→47km
Fairing Jettison T+187s T+225s T+207-214s Closer
1st Stage MECO T+246s, 245 km T+288s, 203 km T+298s, ~200 km 52s→10s (80%)
SECO (Orbit) T+812s, 567 km T+766s, 363 km T+~850-1017s More realistic

Orbital Parameters

Parameter Before Optimized Note
Perigee 481 km 293 km Closer to 300 km target
Apogee 576 km 365 km
Eccentricity 0.0069 0.0054 More circular
Period 95.1 min 91.0 min

Flight Profile

H3-22S Trajectory

Accuracy Assessment

Aspect Before After Notes
SRB separation A (1.7%) A (2.2%) Stable
MECO timing C+ (17%) A- (3.5%) Major improvement
MECO altitude C (22%) A (1.7%) 93% improvement
Orbit altitude C (60%+) B+ (~10%) Much closer to 300 km
Eccentricity B+ A- 0.0069 → 0.0054

Overall: 70-80% → 90%+ quantitatively accurate.

Powered Descent Guidance (G-FOLD)

Convex optimization (SOCP) for H3 booster recovery — SpaceX-style hoverslam landing.

Powered Descent

Configuration

Parameter Value
Booster mass 25 ton (dry 20t + fuel 5t)
Engine LE-9, 0-1,471 kN
T/W at max thrust 6.0 (hoverslam required)
Initial altitude 1,500 m
Initial descent rate 60 m/s

Results

Metric Value
Optimal landing time 19.0 s
Fuel consumption 1,523 kg (30.5% of budget)
Landing speed 0.000 m/s
Landing error 0.00 m
Tsiolkovsky efficiency 94% (vs ideal 1,439 kg)

Algorithm

  • G-FOLD (Guidance for Fuel-Optimal Large Diverts)
  • Lossless convexification transforms non-convex thrust bounds into SOCP
  • Solved with CLARABEL via cvxpy
  • Time-optimal search over tf ∈ [15, 35] s

RL Landing Agent (PPO)

Custom 2D environment with H3 booster physics. PPO (Proximal Policy Optimization) with curriculum learning.

RL Landing

Component Detail
Observation 8-dim: [alt, x, vx, vy, θ, ω, fuel_frac, t_frac]
Action 2-dim continuous: [throttle, gimbal]
Reward Shaped: landing bonus, crash penalty, fuel efficiency, approach shaping
Architecture Actor-Critic MLP (128-128)
Training Curriculum: low alt → full difficulty, 800+ episodes

Status: Requires GPU training (>10,000 episodes) for convergence. Demo results on CPU.

Orbital Transfer Analysis (poliastro)

Orbital Transfer

Delta-v Budget

Mission Δv (km/s) Transfer Time
LEO 300km → GTO 2.426 2.6 hours
GTO → GEO 1.467 2.6 hours
LEO → GEO (total) 3.893 5.3 hours
LEO → Moon (TLI) 3.106 5.0 days
Moon orbit insertion 0.830 —

H3-24L Fuel Feasibility

  • 2nd stage available Δv: 5,710 m/s
  • Required for GTO: 2,426 m/s → Margin: +3,284 m/s
  • Lunar TLI: 3,106 m/s → Feasible

Key Findings

  • Hohmann transfer is more efficient than bielliptic for GEO (ratio 1.093)
  • TLI velocity is 99.1% of escape velocity — nearly parabolic orbit
  • H3-24L has sufficient margin for direct lunar injection

Sounding Rocket Simulation (RocketPy 6-DOF)

Tanegashima Launch

Tanegashima Sounding Rocket

  • Motor: Cesaroni M2245 (9,978 Ns)
  • Heading: 110 deg (SE, Pacific)
  • Apogee: 3,341 m AGL, Mach 0.87

Calisto (Spaceport America)

Calisto HPR

  • Motor: Cesaroni M1670 (6,026 Ns)
  • Apogee: 3,305 m AGL, Mach 0.86
  • Dual-deploy recovery (drogue + main)

Google Earth KML Files

File Description
papers/repos/RocketPy/jaxa_h3_orbital.kml H3-22S full trajectory to orbit (phase-colored, animated)
papers/repos/RocketPy/jaxa_tanegashima.kml Sounding rocket from Tanegashima (4-phase, animated)
papers/repos/RocketPy/trajectory_pro.kml Calisto from Spaceport America (enhanced)

KML features: phase-colored trajectories, event placemarks (SRB sep, MECO, SECO), gx:Track time animation, HTML flight data popups.

Source Code

Script Description
papers/repos/RocketPy/jaxa_h3_orbital.py H3-22S gravity-turn orbital simulation + KML
papers/repos/RocketPy/optimize_h3_trajectory.py scipy.optimize pitch profile + S1 mass optimization
papers/repos/RocketPy/h3_powered_descent.py G-FOLD convex optimization landing guidance
papers/repos/RocketPy/h3_rl_landing.py PPO reinforcement learning landing agent
papers/repos/RocketPy/h3_orbital_transfer.py poliastro GTO/lunar transfer analysis
papers/repos/RocketPy/jaxa_tanegashima_launch.py Tanegashima sounding rocket (RocketPy) + KML
papers/repos/RocketPy/run_simulation.py Calisto basic simulation (RocketPy)
papers/repos/RocketPy/generate_kml.py Enhanced KML generator for Calisto

Cloned Repositories

Repo Purpose
papers/repos/RocketPy/ 6-DOF trajectory simulation (Python)
papers/repos/poliastro/ Orbital mechanics / astrodynamics (Python)
papers/repos/gfold-py/ G-FOLD powered descent guidance (cvxpy)
papers/repos/lcvx-pdg/ Lossless convexification PDG
papers/repos/RocketLander/ RL rocket landing (PyBox2D + PyTorch)
papers/repos/MAPLEAF/ 6-DOF rocket simulation framework
papers/repos/awesome-space/ Curated list of space-related OSS

References

Clone this wiki locally