qpmad-> https://github.com/asherikov/qpmadqpOASES-> https://github.com/coin-or/qpOASESeiquadprog-fast-> https://github.com/stack-of-tasks/eiquadprog
- All problems are dense and positive-definite.
qpOASESis used with MPC option preset, this gives ~50% performance boost compared to default preset.- Time measurements include only the time required to solve a problem; other operations, such as initialization of the solver or reshaping of the constraints, are not taken into account.
-
The problem set includes 'standalone' problems distributed with qpOASES -> https://github.com/coin-or/qpOASES/tree/misc/testingdata/cpp/problems, positive-semidefinite problems are skipped.
-
Hotstarting is not used.
Time (second) required to solve each of the problems:
The same plot with logarithmic scale:
-
The problem set includes MPC problems distributed with qpOASES -> https://github.com/coin-or/qpOASES/tree/misc/testingdata/cpp/oqp, all of them have constant Hessian and varying constraints.
-
qpOASESis hotstarted starting from the second problem in the sequence, factorization of the Hessian is reused byqpmadin the same way.
eiQuadProgdoes not produce the exact answer (up to 1e-9 precision) on some of the problems.
Plots correspond to specific MPC problems and demonstrate time (second) taken
by the solvers on each MPC iteration:
