Initial state values ignored

User Questions
1

Hi,

I am trying to set up an NMPC for obstacle avoidance, i implemented it in python and now i am trying to use it my ROS setup in c++ by using the c generated code. I am using an SQP with partial condensing HPIPM.

In the c generated code, i am bounding my initial value for my states using the following the commands:

ocp_nlp_constraints_model_set(nlp_config, nlp_dims, nlp_in, 0, "lbx", lbx0);
ocp_nlp_constraints_model_set(nlp_config, nlp_dims, nlp_in, 0, "ubx", ubx0);

However, the resulting initial values from the optimization are drifting more and more after each close loop iteration. Is there any reason why these constraints would not be respected ? Is there any reason why it might get worst across the simulation ?

Thanks for the help,

Hope this makes sense

Hugo

2

Hi,

in the snippet you posted, everything seems correct.
How far is x0 value in the solution off from the constraint?
More than the tolerance?
Do you check what solver status acados returns?