Skip to content

Rocket ​

  • 3 state
  • 1 control
  • 1 variable
  • Mayer cost
  • free final time
  • 3 state box
  • 1 control box
  • 1 variable box
  • 4 boundary
  • OptimalControl valid
  • OptimalControl, scalarized valid
  • JuMP valid

The Goddard rocket problem is a classical optimal control problem that models the ascent of a vertically launched rocket. The objective is to maximise the final altitude by optimally controlling the thrust, while accounting for atmospheric drag, gravity, and fuel consumption. This problem is notable for its inclusion of singular arcs, where the control is neither at its maximum nor minimum, complicating the solution process. Historically, it was first proposed by Robert H. Goddard in 1919 (A Method of Reaching Extreme Altitudes), one of the founding works in modern rocketry.

Mathematical formulation ​

The problem can be stated as

where

  • Drag:
  • Gravity:
  • Fuel constant:
  • Maximum thrust:

Qualitative behaviour ​

The optimal trajectory often exhibits a bang–singular–bang structure, with thrust at maximum, then along a singular arc, and finally at minimum. The singular arc arises from the interplay of drag and gravity, producing a non-trivial optimal control law. This behaviour illustrates the complexity of the problem and the challenges in numerically resolving singular arcs.

Characteristics ​

  • Nonlinear coupled dynamics with drag and gravity.

  • Singular arcs in the optimal control.

  • Classical benchmark for testing optimal control algorithms, especially in aerospace applications.

References ​

  • Goddard, R. H. (1919). A Method of Reaching Extreme Altitudes. Smithsonian Institution, Washington D.C. The original formulation of the vertical rocket ascent problem, introducing fuel consumption, gravity, and drag effects.

  • Bryson, A. E. (1999). Dynamic Optimization. Addison Wesley Longman. pp. 392–394. Provides an introduction to singular arcs in optimal control with the Goddard rocket as a primary example.

  • Garfinkel, B. (1963). A Solution of the Goddard Problem. SIAM Journal on Control, 1(1), 1–20. doi:10.1137/0301020 Analytical and numerical solutions of the Goddard problem highlighting singular arc structures.

  • Tsiotras, P., & Kelley, H. J. (1992). Goddard Problem with Constrained Time of Flight. Journal of Guidance, Control, and Dynamics, 15(2), 394–399. doi:10.2514/3.20836 Study of time-constrained versions of the Goddard problem and associated optimal control strategies.

  • Bonnans, F., Martinon, P., & Trélat, E. (2007). Singular Arcs in the Generalized Goddard's Problem. arXiv:math/0703911 Analysis of singular arcs and numerical methods for the Goddard problem.

  • More, J., Garbow, B., Hillstrom, K., & Watson, L. (2001). COPS: Constrained Optimization Problem Set (COPS3). Argonne National Laboratory. Link Provides a benchmark implementation of the Goddard rocket problem for testing numerical optimization methods.

Parameters ​

ParameterDefault valueUnitMeaning
t00–initial time
h_t01–initial altitude, and lower bound of the altitude
v_t00–initial velocity, and lower bound of the velocity
m_t01–initial mass, and upper bound of the mass
g01–gravitational acceleration at the initial altitude
Tc3.5–maximum thrust, in units of the initial weight m_t0 g0
hc500–decay rate of the air density with the altitude, exp(−hc (h − h_t0) / h_t0)
vc620–drag coefficient factor, Dc = vc m_t0 / (2 g0)
mc0.6–final mass, as a fraction of the initial mass
T_l0–lower bound of the thrust T
tf_l0–lower bound of the final time

Every modeling takes other values with the keyword parameters, merged with these: rocket(backend; parameters=(t0=…,)).

Reference objective ​

The best objective among the solves that converged, in each modeling, on the default grid of the modeling: the modelings are independent, each has its own reference. See how the references are computed.

ModelingSchemeGridReference objectiveSolvesDistinct solutionsConfigurations agree
JuMP:trapeze500-1.01283498481yes, within 1.7e-06
OptimalControl:trapeze500-1.012834984181yes, within 1.7e-06
OptimalControl, scalarized:trapeze500-1.01283498431yes, within 0