Issue: we sometimes get numerical errors (usually for trigonometric functions) on some problems, which are currently not caught and abort the execution
Improvement: use a try-catch on the NLP solver call, and in case of errors pursue the execution by returning a dummy solution with successful=false and maybe some information in the text message.
This will allow better testing of the difficult problems.
Issue: we sometimes get numerical errors (usually for trigonometric functions) on some problems, which are currently not caught and abort the execution
Improvement: use a
try-catchon the NLP solver call, and in case of errors pursue the execution by returning a dummy solution withsuccessful=falseand maybe some information in the text message.This will allow better testing of the difficult problems.