Hi,
could you please explain what the parameter obj.opts_struct.qp_solver_ric_alg can be used for? How does it influence the behavior of HPIPM and what are possible values?
BR
Martin
Hi,
could you please explain what the parameter obj.opts_struct.qp_solver_ric_alg can be used for? How does it influence the behavior of HPIPM and what are possible values?
BR
Martin
Hi Martin,
qp_solver_cond_ric_alg
Possible values: {0,1}
0: dont factorize hessian in the condensing;
1: factorize
From here:
Cheers,
Jonathan
Hi Jonathan,
I asked about qp_solver_ric_alg and not qp_solver_cond_ric_alg. The inline documentation says only “HPIPM specific”.
BR
Martin
Right, sorry.
I dont know then, hopefully @giaf can explain.
These options directly trigger HPIPM options.
Specifically, the qp_solver_ric_alg controls whether the IPM solver uses the classic (0) or square-root (1) Riccati recursion for factorization of the KKT system.
The former requires more flops, the latter requires the full Hessian to be positive definite.
Regarding qp_solver_cond_ric_alg, the default state condensing algorithm is based on an algorithm analogous to the Riccati recursion, and identical considerations apply.