TrustRegionIteration stopping condition

[rjoomen]

In the original trajopt paper and in trajopt_sco, the trust region loop of the optimization is exited when the trust region is below a certain threshold (min_trust_box_size). In trajopt_sqp, however, the trust region loop is run up to a maximum number of iterations ( max_trust_region_expansions).

What is the reason for this change? Because I would expect something like this:

  int trust_region_iteration = -1;
  while (results_.box_size[0] >= params.min_trust_box_size)
  {
    trust_region_iteration++;

(On a side note: max_trust_region_expansions should be called max_trust_region_shrinkings or something like that, as an expansion immediately exits the loop.)

[Levi-Armstrong]

This was most likely an oversight in the original port. I have been going through it from time to time finding similar differences and fixing them. I would consider this a bug and it should be fixed.

[rjoomen]

Fixed by #349