Improved model-free adaptive control method using recursive least-squares estimation algorithm

Model-free control is a data-based control approach which can be applied when a dynamic model of the controlled system is difficult to obtain or identify. To realize the controller, a linearized data model is often established locally by using only input-output information from the closed-loop system. The data-based model contains a time-varying parameter matrix called pseudo-jacobian matrix (PJM) which can be estimated and updated recursively at every time instant of the system operation using different on-line parameter estimation methods. As a traditional estimation approach, projection algorithm is generally used only for single-input single-output (SISO) systems. In this contribution, the novel idea of using recursive least-squares (RLS) algorithm is discussed and implemented to improve parameter estimation accuracy. A modified model-free adaptive controller (MMFAC) is designed by considering minimization of the output error and its derivative for tracking control improvement. The proposed method is applied to reduce vibration of a ship-mounted crane representing a multi-input multi-output (MIMO) system. Simulation results indicate that the RLS-based MMFAC has significant improvement in vibration control of the elastic boom and the payload position compared to other traditional approaches.