New software applications for system identification

New software applications for linear multivariable system identification are presented. The incorporated algorithms use subspace-based techniques (MOESP, N4SID, or their combination) to find a standard discrete-time state-space description, and optionally the covariance matrices and Kalman predictor gain, using input and output (I/O) trajectories. For flexibility, separate applications are offered for obtaining the processed triangular factor of the structured, block-Hankel-block matrix of I/O data (using fast or standard QR factorization algorithms), for computing the system and predictor matrices, for estimating the initial state of the system, and for simulating and evaluating the model. The applications are encapsulated in Docker containers, which are managed by the Kubernetes platform on a Linux machine. This ensures greater flexibility, enhanced security, and fast execution. The services to be implemented are part of a cloud-based open platform for process control applications.