Predictive Control Methods for Time-Delay Systems

Predictive control methodologies provide a systematic approach to managing dynamic systems in which there is a significant delay between the application of a control input and the observable effect on the output. These methods use an explicit mathematical model to forecast future system behaviour over a finite horizon and determine control actions that optimise performance criteria such as set-point tracking, disturbance rejection and constraint adherence. Early schemes include the Smith predictor, which compensates for known constant delays by embedding a process model within the control loop, and Internal Model Control, which enhances robustness through filter design. More sophisticated frameworks such as Model Predictive Control and Generalised Predictive Control extend these concepts to multivariable and constrained environments, solving optimisation problems in real time. Recent developments integrate observer-based structures to estimate unmeasured disturbances and employ adaptive or nonlinear predictors—often realised with artificial neural networks or fuzzy logic—to cope with time-varying delays and model uncertainties. These advances have found application across process industries, power systems, networked control architectures and autonomous platforms, where precise delay compensation is essential for stability, efficiency and safety.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent practical implementations have validated classical predictor approaches. In an iron ore processing plant, a Smith predictor was applied to the feeder–conveyor control loop to offset sensor-induced measurement delays. Field results demonstrated a 55% reduction in flow oscillations and a 355 t/h increase in throughput, highlighting the method’s industrial viability.

Neural network–augmented prediction techniques have advanced delay compensation in nonlinear environments. In irrigation canal management, a controller combining a Smith predictor with a NARX neural network model achieved faster set-point tracking and improved disturbance rejection across varying hydraulic regimes, outperforming conventional linear predictors under unsteady flow conditions.

Observer-based repetitive control has emerged as an effective solution for periodic disturbance rejection in delayed, uncertain systems. A predictive extended state observer (ESO)–based repetitive controller was tested on a brushless DC servo motor with significant input delay. By concurrently estimating disturbances and computing periodic compensation signals, the scheme delivered enhanced tracking accuracy without sacrificing robust stability.

Predictive Control Methods for Time-Delay Systems publication trend

The graph below shows the total number of articles in predictive control methods for time-delay systems across all publications each year (not limited to Nature Index journals).

Technical terms

Time-delay system: A dynamic system in which the effect of an input or disturbance is observed only after a finite time lag.

Predictive control: A control strategy that uses a model to forecast future outputs and optimises control inputs over a receding time horizon.

Smith predictor: A delay-compensation scheme that embeds a process model within the feedback loop to predict and offset known constant delays.

Model Predictive Control (MPC): An optimisation-based framework that handles multivariable systems and constraints by solving finite-horizon control problems online.

Extended State Observer (ESO): An observer design that estimates both the system states and unknown disturbances to enhance robustness against model uncertainties.

NARX neural network predictor: A nonlinear autoregressive predictor with exogenous inputs, using a neural network to model and forecast future system outputs.

Repetitive controller: A control structure tailored to reject or track periodic signals by embedding a periodic inversion mechanism within the loop.

References

1. Delay Compensation in a Feeder–Conveyor System Using the Smith Predictor: A Case Study in an Iron Ore Processing Plant. Sensors (2024).
2. Design of a NARX-ANN-Based SP Controller for Control of an Irrigation Main Canal Pool. Applied Sciences (2022).
3. Predictive extended state observer-based repetitive controller for uncertain systems with input delay. Automatika (2021).