Data-Driven Filtered Smith Predictor for SISO Dead-Time Processes

Abstract This paper introduces an innovative data-driven filtered Smith predictor (DD-FSP) for controlling single-input, single-output (SISO) processes with dead-time. The proposed approach is applicable to open-loop stable, integrative, and unstable systems. The controller design is based on a mean-square minimization algorithm that directly maps input-output data, allowing the automatic generation of an estimated impulse response sequence. This data-driven methodology enables the control strategy to address the well-known trade-off between robustness and performance without requiring an explicit parametric model for system prediction. By leveraging the mean-square optimization framework, the proposed DD-FSP provides the desired closed-loop response while adapting to the system dynamics through available data rather than predefined mathematical models. This flexibility enhances the applicability of the method across a wide range of process dynamics, improving system performance without compromising robustness. To validate the effectiveness of the proposed approach, simulation studies are carried out, demonstrating the practical benefits of the data-driven strategy. The results highlight the capability of the DD-FSP to balance performance and robustness efficiently, reinforcing its potential as a reliable alternative for controlling SISO processes with dead-time.