This benchmark dataset contains input-output data from the CubeSpec Fine Steering Mirror, a multi-input multi-output high-precision control platform used in a small satellite. Voltages applied to three piezo-actuators serve as inputs, while the mirror displacements measured at three non-collocated reference points serve as outputs. The available dataset can be downloaded here, you can also find a detailed description of the system setup through this link.
The excitation signals are orthogonal random-phase multisines spanning a wide frequency range, and are applied at three different amplitude levels. The system behaves mostly linearly, but the presence of hysteresis in the piezo-actuators introduces dynamic nonlinearities, making the dataset well-suited for benchmarking nonlinear identification methods.
The dataset is particularly well suited for frequency-domain identification, as it uses periodic multisine excitations and contains no transients. Key challenges mainly arise from its high dimensionality and include, e.g., model order selection, the decoupling of multivariate nonlinearities, and maintaining computational tractability during both identification and deployment.
Previously published results on the Fine Steering Mirror benchmark can be found below. You can submit your own results through this form. Note that the reported results are curated, only complete submissions with meaningful contributions will be included. Candidate entries should make use of the Python dataloader functionalities and figure of merit calculation functions provided through this link.
Please refer to the Fine Steering Mirror dataset as:
Floren, M., Peri, L., De Maeyer, J., De Munter W., Vandepitte, D., and Noël, J.-P. Data-driven state-space identification and nonlinearity assessment of the CubeSpec Fine Steering Mirror. In Conference Proceedings of ISMA-USD2024, (2024), pp. 2042--2052.