Conclusion

In this dossier, we have presented an overview of methods for the direct identification of continuous-time black-box linear models from sampled data. These approaches have long been overshadowed by traditional discrete-time model identification methods. Direct continuous-time model identification approaches offer the advantages of providing a model that can be directly interpreted physically by the user, of facilitating the implementation of the complete system identification procedure, and of being able to handle the case of variable-step sampled data. The CONTSID software toolbox has helped to disseminate and facilitate the use of these approaches. However, these approaches should not be seen as competitors, but rather as complementary to discrete-time model identification techniques.