[WIP]: Add OSQP solver in Proxsuite #415

Lucas-Haubert · 2025-08-22T10:32:57Z

This (draft) PR aims at adding the OSQP solver in the Proxsuite library. Information about the solver can be found in the related paper and the related website.

Contents of the PR:

Implementation of the OSQP algorithm (within the Proxsuite PrimalDualLDLT dense backend),
API in C++ and Python for OSQP, based on the one of ProxQP,
Code refactoring to put the common tools (enum, classes, utility functions) in the namespace proxsuite::common instead of proxsuite::proxqp,
Calibration tools in Python to compare our implementation of OSQP Proxsuite with OSQP source code (conda install conda-forge::osqp) on the same problems, given some precision criteria on the number of iterations, values of the variables and residuals, etc.

Comparison of results (OSQP Proxsuite vs source):

Calibration: Problem instances inspired by the ones from proxsuite/common/utils/random_qp_problems.hpp are generated with dimensions going from dim=10 to dim=1000. Then the two solvers are compared.
Benchmarks: The Proxsuite OSQP solver is added in the ProxQP benchmark repository.
Results: Proxsuite OSQP and OSQP source are close in terms of behavior and performance in the majority of test cases. Yet, the implementation in Proxsuite remains perfectible in order to make its behavior (in terms of number of iterations, updates of mu, values of variables, etc) strictly identical and its performances (timings) better.

Remaining work:

Strict calibration of the Proxsuite OSQP solver given OSQP source (algorithmic behavior),
Profling to reduce run_time (timings),
Implement the DenseBackend::PrimalLDLT and SparseBackend options.

…cting QP

…f qpwork.dirty

…stacks from the one in admm

…rce to explain failing tests

…_calibration_dense.py

…in due to source code implementation

Pure inequality degenerate QP: Set eps_primal_inf and eps_dual_inf equal to 1e-15 to reproduce benchmark setting in Simple-Robotics/proxqp_benchmark Unit tests: Fix unit tests in osqp/dense_qp_wrapper

… tests

…_meszaros functions

…ExposeParallel

…_qp_solve.cpp

…_wrapper (1e-3, low precision)

jorisv · 2025-09-03T14:19:11Z

bindings/python/src/helpers.hpp

+// Copyright (c) 2025 INRIA
+//
+
+#include "pytypedefs.h"


This header doesn't exists in Python < 3.11. Why is it mandatory ? It only declare forward declaration as stated in the header file:

// Forward declarations of types of the Python C API. // Declare them at the same place since redefining typedef is a C11 feature. // Only use a forward declaration if there is an interdependency between two // header files.

jorisv · 2025-09-03T14:49:04Z

bindings/python/src/helpers.hpp

+{
+  namespace nb = nanobind;
+  assert(h.is_type());
+  return nb::steal<nb::str>(PyType_GetName((PyTypeObject*)h.ptr()));


PyType_GetName is only available in Python 3.11

test/CMakeLists.txt

examples/python/CMakeLists.txt

examples/cpp/CMakeLists.txt

jorisv · 2025-09-03T15:27:03Z

test/src/osqp/dense_qp_wrapper.cpp

This test fail with the following error:

226: Traceback (most recent call last): 226: File "/home/jvaillan/dev/inria/proxsuite/proxsuite/test/src/osqp/dense_qp_wrapper.py", line 675, in test_case_cold_start_with_previous_result 226: proxsuite.osqp.InitialGuess.EQUALITY_CONSTRAINED_INITIAL_GUESS 226: ^^^^^^^^^^^^^^^^^^^^^^^^^^^ 226: AttributeError: module 'proxsuite.proxsuite_pywrap_avx2.osqp' has no attribute 'InitialGuess'

jorisv · 2025-09-04T10:17:19Z

include/proxsuite/osqp/dense/solver.hpp

+
+  for (i64 iter = 0; iter < qpsettings.max_iter; ++iter) {
+
+    common::dense::compute_residuals(qpsettings,


On first run, this method will access qpwork.active_set_up that is not initialized.

Here the valgrind trace:

==58952== Conditional jump or move depends on uninitialised value(s) ==58952== at 0x26D64C: Eigen::internal::evaluator<Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > >::coeff(long, long) const (CoreEvaluators.h:1264) ==58952== by 0x25D322: Eigen::internal::binary_evaluator<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const>, Eigen::internal::IndexBased, Eigen::internal::IndexBased, double, double>::coeff(long, long) const (CoreEvaluators.h:769) ==58952== by 0x241104: Eigen::internal::redux_evaluator<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> >::coeffByOuterInner(long, long) const (Redux.h:381) ==58952== by 0x22B738: double Eigen::internal::redux_impl<Eigen::internal::scalar_sum_op<double, double>, Eigen::internal::redux_evaluator<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> >, 0, 0>::run<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> >(Eigen::internal::redux_evaluator<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> > const&, Eigen::internal::scalar_sum_op<double, double> const&, Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> const&) (Redux.h:202) ==58952== by 0x21A92C: double Eigen::DenseBase<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> >::redux<Eigen::internal::scalar_sum_op<double, double> >(Eigen::internal::scalar_sum_op<double, double> const&) const (Redux.h:418) ==58952== by 0x20B4D9: Eigen::DenseBase<Eigen::CwiseBinaryOp<Eigen::internal::scalar_conj_product_op<double, double>, Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > const> >::sum() const (Redux.h:463) ==58952== by 0x1F8DBE: Eigen::internal::dot_nocheck<Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > >, Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > >, false>::run(Eigen::MatrixBase<Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > > const&, Eigen::MatrixBase<Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > > const&) (Dot.h:37) ==58952== by 0x1E84B3: Eigen::ScalarBinaryOpTraits<double, Eigen::internal::traits<Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > >::Scalar, Eigen::internal::scalar_product_op<double, Eigen::internal::traits<Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > >::Scalar> >::ReturnType Eigen::MatrixBase<Eigen::Select<Eigen::Block<Eigen::Matrix<bool, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::Block<Eigen::Matrix<double, -1, 1, 0, -1, 1>, -1, 1, false>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > >::dot<Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > >(Eigen::MatrixBase<Eigen::Select<Eigen::CwiseBinaryOp<Eigen::internal::scalar_cmp_op<double, double, (Eigen::internal::ComparisonName)1>, Eigen::ArrayWrapper<Eigen::Matrix<double, -1, 1, 0, -1, 1> const> const, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Array<double, -1, 1, 0, -1, 1> > const>, Eigen::Matrix<double, -1, 1, 0, -1, 1>, Eigen::CwiseNullaryOp<Eigen::internal::scalar_constant_op<double>, Eigen::Matrix<double, -1, 1, 0, -1, 1> > > > const&) const (Dot.h:84) ==58952== by 0x1D7792: void proxsuite::common::dense::global_dual_residual<double>(proxsuite::common::Results<double>&, proxsuite::common::dense::Workspace<double>&, proxsuite::common::dense::Model<double> const&, bool, proxsuite::common::dense::preconditioner::RuizEquilibration<double> const&, double&, double&, double&, double&, double&, double&, proxsuite::common::HessianType const&) (utils.hpp:433) ==58952== by 0x1D038B: void proxsuite::common::dense::compute_residuals<double>(proxsuite::common::Settings<double> const&, proxsuite::common::dense::Model<double> const&, proxsuite::common::Results<double>&, proxsuite::common::dense::Workspace<double>&, bool, proxsuite::common::HessianType const&, proxsuite::common::dense::preconditioner::RuizEquilibration<double>&, double&, double&, double&, double&, double&, double&, double&, double&, double&, double&, double&) (utils.hpp:551) ==58952== by 0x1C5ACF: void proxsuite::osqp::dense::qp_solve<double>(proxsuite::common::Settings<double> const&, proxsuite::common::dense::Model<double> const&, proxsuite::common::Results<double>&, proxsuite::common::dense::Workspace<double>&, bool, proxsuite::common::DenseBackend const&, proxsuite::common::HessianType const&, proxsuite::common::dense::preconditioner::RuizEquilibration<double>&) (solver.hpp:677) ==58952== by 0x1B40F3: proxsuite::osqp::dense::QP<double>::solve_implem() (wrapper.hpp:237) ==58952== Uninitialised value was created by a heap allocation ==58952== at 0x4848899: malloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so) ==58952== by 0x1A2F17: Eigen::internal::aligned_malloc(unsigned long) (Memory.h:182) ==58952== by 0x1E980A: void* Eigen::internal::conditional_aligned_malloc<true>(unsigned long) (Memory.h:241) ==58952== by 0x1EEF70: bool* Eigen::internal::conditional_aligned_new_auto<bool, true>(unsigned long) (Memory.h:404) ==58952== by 0x1DDE27: Eigen::DenseStorage<bool, -1, -1, 1, 0>::resize(long, long, long) (DenseStorage.h:639) ==58952== by 0x1CD801: Eigen::PlainObjectBase<Eigen::Matrix<bool, -1, 1, 0, -1, 1> >::resize(long) (PlainObjectBase.h:311) ==58952== by 0x1BF746: proxsuite::common::dense::Workspace<double>::Workspace(long, long, long, bool, proxsuite::common::DenseBackend) (workspace.hpp:283) ==58952== by 0x1B26D6: proxsuite::common::dense::QPBase<proxsuite::osqp::dense::QP<double>, double>::QPBase(long, long, long, bool, proxsuite::common::HessianType, proxsuite::common::DenseBackend) (wrapper.hpp:109) ==58952== by 0x1ABDBA: proxsuite::osqp::dense::QP<double>::QP(long, long, long, bool) (wrapper.hpp:116) ==58952== by 0x195F5A: DOCTEST_ANON_FUNC_92() (dense_qp_wrapper.cpp:7146) ==58952== by 0x2EF68D: doctest::Context::run() (doctest.hpp:8578) ==58952== by 0x2F0086: main (doctest.hpp:8682)

.gitmodules

Lucas-Haubert added 30 commits August 22, 2025 11:46

osqp/: Create dense, solver, wrapper files

508bd28

osqp/dense/wrapper: Function init_osqp_settings to call after constru…

4fd1c6e

…cting QP

osqp/utils/prints: Add print preambule function

7b0fe6f

proxqp/dense/workspace: Add OSQP material

97204a7

Copyright date in osqp/utils/prints

29d794a

bindings: Add OSQP

8c3eee6

Moved zeta_eq and zeta_in in results

4498ae0

osqp/dense/solver: Structure of the function qp_solve

256892f

osqp/dense/solver: Coded ADMM with no mu update

70b660a

test: Add unit tests for ADMM only

fe18192

Add remove columns step in setup_factorisation_complete_kkt i ncase o…

7f971c9

…f qpwork.dirty

Renamed setup_factorisation_somplete_kkt into factorization_complete_kkt

739dbe2

Removed assrt on type_check in exposeAndExportValues

0ced984

Mu update

cc7e3ff

Polishing

5e3b145

Tests: Set precision 1e-3 in unit tests

5097cfc

Polishing: Perform polishing factorization and solve in separate ldl …

392f988

…stacks from the one in admm

Fix unit tests before calibration with osqp source

a48d218

Test osqp_dense_maros_meszaros: Comparaison of OSQP proxsuite and sou…

9e5b068

…rce to explain failing tests

Renamed examples/python/osqp_calibration.py into examples/python/osqp…

41996c7

…_calibration_dense.py

tests osqp: print iter_ext instead of iter

1c0a5e4

unit test: degenerate: File to compare proxsuite vs source

d29e7cf

osqp solver header: Documentation of the code blocks

e33f103

unit test: Add tests for solution polishing

23cd995

Failing tests in wrapper: document them

61228de

calibration: Compare the rho_estimate values instead of mu_eq and mu_…

19773c8

…in due to source code implementation

solver: Correction in ADMM update

8f7745a

Setup of calibration tests

ca5d60a

Calibration strongly convex qp, not strongly convex, degenerate

311bb5e

examples/python/osqp_calibration: Refactor code

eb465d4

Lucas-Haubert and others added 15 commits August 22, 2025 11:46

unit test: Commented failed tests in osqp/dense_qp_wrapper

1d3737c

Pure inequality degenerate QP: Set eps_primal_inf and eps_dual_inf equal to 1e-15 to reproduce benchmark setting in Simple-Robotics/proxqp_benchmark Unit tests: Fix unit tests in osqp/dense_qp_wrapper

Coded check_termination option inspired from OSQP source code

850ec14

Rename PolishStatus into PolishOutput

0e7e07b

test: Updates comments on option primal_infeasibility_solving in OSQP…

451ac4d

… tests

osqp/calibration: Deactivate run of calibration tests in ctest

6b5c8f8

README: Add WIP: OSQP section

febb65f

ci: run_test option in test_calibration_qp and test_calibration_maros…

545ace5

…_meszaros functions

ci: Remove calibration from tests in CI and ctest

b8c26e1

osqp/dense/solver.hpp: Removed unsuded variable sqrt_mu_update

0feedd2

ci: expose-all.cpp: changed dense_module into proxqp_dense_module in …

88eee84

…ExposeParallel

ci: Add osqp in environment.yml

705914d

ci: Management of proxqp and common namespace in test proxqp/parallel…

a8bc5b1

…_qp_solve.cpp

Unit tests OSQP: eps_abs = 1e-5 (high precision), except for dense_qp…

243d078

…_wrapper (1e-3, low precision)

changelog: Add entry

4af963a

cmake: Use patched jrl-cmakemodules

1d0928c

jorisv reviewed Sep 3, 2025

View reviewed changes

python: Remove useless include

2391a00

jorisv reviewed Sep 3, 2025

View reviewed changes

python: Use nanobind function instead of CPython API

e0bcfab

jorisv requested changes Sep 3, 2025

View reviewed changes

test/CMakeLists.txt Outdated Show resolved Hide resolved

examples/python/CMakeLists.txt Outdated Show resolved Hide resolved

examples/cpp/CMakeLists.txt Outdated Show resolved Hide resolved

jorisv reviewed Sep 3, 2025

View reviewed changes

jorisv added 3 commits September 4, 2025 10:54

python: Fix type_name_short to return type name without module

204f70f

ci: Remove deprecated ROS version

2ea0269

packaging: Fix build

77ddd62

jorisv force-pushed the topic/osqp branch from fb463c0 to 77ddd62 Compare September 4, 2025 09:36

jorisv reviewed Sep 4, 2025

View reviewed changes

jorisv reviewed Sep 5, 2025

View reviewed changes

.gitmodules Outdated Show resolved Hide resolved

jorisv and others added 3 commits September 8, 2025 13:17

cmake: jrl-cmakemodules patch is merged

c68f351

cmake-module: update

569c87f

Removed useless REGEX in string(REGEX REPLACE

0af2ddb

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[WIP]: Add OSQP solver in Proxsuite #415

[WIP]: Add OSQP solver in Proxsuite #415

Uh oh!

Lucas-Haubert commented Aug 22, 2025 •

edited

Loading

Uh oh!

jorisv Sep 3, 2025

Uh oh!

jorisv Sep 3, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jorisv Sep 3, 2025

Uh oh!

jorisv Sep 4, 2025

Uh oh!

Uh oh!

Uh oh!


		for (i64 iter = 0; iter < qpsettings.max_iter; ++iter) {

		common::dense::compute_residuals(qpsettings,

[WIP]: Add OSQP solver in Proxsuite #415

Are you sure you want to change the base?

[WIP]: Add OSQP solver in Proxsuite #415

Uh oh!

Conversation

Lucas-Haubert commented Aug 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jorisv Sep 3, 2025

Choose a reason for hiding this comment

Uh oh!

jorisv Sep 3, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jorisv Sep 3, 2025

Choose a reason for hiding this comment

Uh oh!

jorisv Sep 4, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Lucas-Haubert commented Aug 22, 2025 •

edited

Loading