Skip to content

Bryson denham ​

  • 2 state
  • 1 control
  • Lagrange cost
  • fixed final time
  • 1 state box
  • 4 boundary
  • OptimalControl valid
  • OptimalControl, scalarized valid
  • JuMP valid
  • ModelingToolkit valid

The Bryson–Denham problem is a classic optimal control benchmark involving a state inequality constraint. It models the movement of a unit mass on a frictionless plane, where the goal is to relocate the mass over a fixed time interval with minimum control effort, while ensuring the position never exceeds a certain limit. Originally introduced in Bryson et al. 1963, it is a standard test case for evaluating the ability of numerical solvers to handle state-constrained trajectories.

Mathematical formulation ​

The problem can be stated as

\begin{aligned} \min_{x,u} \quad & J(x,u) = \frac{1}{2} \int_0^1 u^2(t) ,\mathrm{d}t \ \text{s.t.} \quad & \dot{x}_1(t) = x_2(t), \quad \dot{x}_2(t) = u(t), \ & x_1(0) = 0, ; x_1(1) = 0, ; x_2(0) = 1, ; x_2(1) = -1, \ & x_1(t) \le a. \end

where represents position, velocity, acceleration (control), and is the maximum allowable displacement (e.g.,  ).

Qualitative behaviour ​

The problem features a second-order state constraint because the control only appears in the second derivative of the constrained state :

\ddot{x}_1(t) = u(t).

When the trajectory is on a boundary arc ( ), the derivatives and must be zero. This implies that while the constraint is active, the velocity is zero and the control is zero:

u(t) = 0.

The solution structure changes based on the value of the constraint parameter :

  • Unconstrained Case: If is sufficiently large ( ), the state constraint is never active.

  • Touch Point: For a critical value of , the trajectory just touches the boundary   at a single point  .

  • Boundary Arc: For smaller values (e.g.,  ), the trajectory remains on the boundary for a finite time interval. During this interval, the control vanishes.

Characteristics ​

  • Linear–quadratic dynamics with a second-order state inequality constraint.

  • Demonstrates a "bang-off-bang" control structure where the control is zero on the constraint boundary.

  • Used to test the accuracy of state constraint handling and the detection of switching times.

References ​

  • Bryson, A.E., Denham, W.F., & Dreyfus, S.E. (1963). Optimal programming problems with inequality constraints I: necessary conditions for extremal solutions. AIAA Journal. doi.org/10.2514/3.2107

  • Dymos Documentation: Bryson-Denham Problem. dymos/examples/bryson_denham

Parameters ​

ParameterDefault valueUnitMeaning
t00.0–initial time
tf1.0–final time
x1_t00.0–initial position x₁
x2_t01.0–initial velocity x₂
x1_tf0.0–final position x₁
x2_tf-1.0–final velocity x₂
x1_max0.1111111111111111–upper bound of the position x₁ (the state constraint)

Every modeling takes other values with the keyword parameters, merged with these: bryson_denham(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:trapeze5004.00014342281yes, within 1.1e-07
OptimalControl:trapeze5004.000142996181yes, within 1.4e-07
OptimalControl, scalarized:trapeze5004.00014349831yes, within 2.8e-09

Analytic solution. The continuous problem has a known solution, of objective 4 (Bryson, Denham & Dreyfus (1963), https://doi.org/10.2514/3.2107). The reference objective above is the one of the discretized problem: the gap between the two is the discretization error.