I use acados with ROS for trajectory planning in presence of obstacles. I initialize the solver in a loop with rate 10HZ. I see that the solver not always successes in finding the solution i.e. the returned status sometimes not equals to zero. However in this case i ignore the resulted solution. I can say that the percentage of success is more than 85% But i need to know if this is normal or it should be 100% if the solver is initialized well?
“well” is not really defined.
If you start really close to the solution, an SQP method, as implemented in acados, should converge (or not even iterate).
It might be that in a closed loop setting with hard constraints, you run into infeasible instances of the OCP.
If you use SQP with a limited number of iterations, it can also be that acados returns “ACADOS_MAXITER” (solution status 2), which is not an issue in general.