Solve a Nonlinear Program with Ipopt — ipopt

This is a low-level wrapper over Ipopt's C interface. Derivative callbacks are supplied as R functions. Indices in jacobian_structure and hessian_structure are one-based R indices and are converted to Ipopt's zero-based C index style internally.

Usage

ipopt_solve(
  x0,
  lower,
  upper,
  constraint_lower = numeric(),
  constraint_upper = numeric(),
  eval_f,
  eval_grad_f,
  eval_g = NULL,
  eval_jac_g = NULL,
  jacobian_structure = matrix(integer(), ncol = 2),
  eval_h = NULL,
  hessian_structure = matrix(integer(), ncol = 2),
  options = list()
)

Arguments

x0: Numeric starting point.
lower, upper: Numeric variable bounds, length length(x0).
constraint_lower, constraint_upper: Numeric constraint bounds.
eval_f: Function f(x) returning the scalar objective value.
eval_grad_f: Function grad_f(x) returning a numeric vector.
eval_g: Function g(x) returning a numeric vector of constraints.
eval_jac_g: Function returning the Jacobian nonzero values in the order specified by jacobian_structure.
jacobian_structure: Integer two-column matrix of one-based row/column positions for Jacobian nonzeros.
eval_h: Function (x, obj_factor, lambda) returning Hessian nonzero values in the order specified by hessian_structure. A structurally empty Hessian (no rows in hessian_structure, the function returning numeric()) is valid and corresponds to a linear objective with linear/bound constraints. Pass NULL only when no exact Hessian is available, in which case a Hessian approximation must be requested via options (for example hessian_approximation = "limited-memory"); eval_h = NULL without such an option is an error.
hessian_structure: Integer two-column matrix of one-based lower triangular row/column positions for Hessian nonzeros.
options: Named list of Ipopt options. Numeric, integer, logical, and character values are supported.

Value

A list with solution, objective, constraints, multipliers, and status.