piqp: PIQP Solver object

In piqp: R Interface to Proximal Interior Point Quadratic Programming Solver

PIQP Solver object

Description

PIQP Solver object

Usage

piqp(
  P = NULL,
  c = NULL,
  A = NULL,
  b = NULL,
  G = NULL,
  h_l = NULL,
  h_u = NULL,
  x_l = NULL,
  x_u = NULL,
  settings = list(),
  backend = c("auto", "sparse", "dense")
)

Arguments

P	dense or sparse matrix of class dgCMatrix or coercible into such, must be positive semidefinite
c	numeric vector
A	dense or sparse matrix of class dgCMatrix or coercible into such
b	numeric vector
G	dense or sparse matrix of class dgCMatrix or coercible into such
h_l	numeric vector of lower inequality bounds, default NULL indicating -Inf for all inequality constraints
h_u	numeric vector of upper inequality bounds, default NULL indicating Inf for all inequality constraints
x_l	a numeric vector of lower variable bounds, default NULL indicating -Inf for all variables
x_u	a numeric vector of upper variable bounds, default NULL indicating Inf for all variables
settings	list with optimization parameters, empty by default; see piqp_settings() for a comprehensive list of parameters that may be used
backend	which backend to use, if auto and P, A or G are sparse then sparse backend is used ("auto", "sparse" or "dense") ("auto")

Details

Allows one to solve a parametric problem with for example warm starts between updates of the parameter, c.f. the examples. The object returned by piqp contains several methods which can be used to either update/get details of the problem, modify the optimization settings or attempt to solve the problem.

Value

An S7 object of class "piqp_model" with methods solve(), update(), get_settings(), get_dims(), update_settings() which can be used to solve the problem with updated settings / parameters.

Usage

model = piqp(P = NULL, c = NULL, A = NULL, b = NULL, G = NULL,
             h_l = NULL, h_u = NULL, x_l = NULL, x_u = NULL,
             settings = piqp_settings(),
             backend = c("auto", "sparse", "dense"))

solve(model)
update(model, P = NULL, c = NULL, A = NULL, b = NULL, G = NULL,
       h_l = NULL, h_u = NULL, x_l = NULL, x_u = NULL)
get_settings(model)
get_dims(model)
update_settings(model, new_settings = piqp_settings())

See Also

solve_piqp(), piqp_settings()

Examples

## example, adapted from PIQP documentation
library(piqp)
library(Matrix)

P <- Matrix(c(6., 0.,
              0., 4.), 2, 2, sparse = TRUE)
c <- c(-1., -4.)
A <- Matrix(c(1., -2.), 1, 2, sparse = TRUE)
b <- c(1.)
G <- Matrix(c(1., 2., -1., 0.), 2, 2, sparse = TRUE)
h_u <- c(0.2, -1.)
x_l <- c(-1., -Inf)
x_u <- c(1., Inf)

settings <- list(verbose = TRUE)

model <- piqp(P, c, A, b, G, h_u = h_u, x_l = x_l, x_u = x_u, settings = settings)

# Solve
res <- solve(model)
res$x

# Define new data
A_new <- Matrix(c(1., -3.), 1, 2, sparse = TRUE)
h_u_new <- c(2., 1.)

# Update model and solve again
update(model, A = A_new, h_u = h_u_new)
res <- solve(model)
res$x

piqp documentation built on Oct. 6, 2026, 9:07 a.m.