Pattern-moving based data-driven control for multi-input continuous-time non-Newtonian mechanical systems

A data-driven control scheme based on pattern-moving is proposed for the control problem of multi-input continuous-time non-Newtonian mechanical systems. A sampling data-driven dynamic linearisation method for pattern-moving is proposed by constructing pattern category variables with statistical characteristics to describe the output uncertainty of the system. The multi-input continuous-time non-Newtonian mechanical system is transformed into an affine Input/Output (I/O) model which includes pattern category variables and the statistical feature terms. And the statistical properties of the system are represented by the posterior probability of the pattern category variable. On this basis, the probability density evolution of the tracking error is obtained using the principle of probability density conservation, and a sampling data-driven control based on pattern-moving is proposed by constructing tracking error entropy. The algorithm for estimating controller related parameters is also provided. Then, the convergence of parameter estimation algorithms and control system tracking errors has been proven through the contraction mapping principle and the convergence theorem of iterative matrices. Finally, the effectiveness of the proposed method is verified through numerical simulation.