Multimodel Predictive Control of Nonlinear Systems

Multimodel predictive control of nonlinear systems integrates multiple simplified representations of a complex process to anticipate and regulate its behaviour. Nonlinear systems often exhibit varying dynamics across operating regions, making single-model controllers insufficient to achieve robust performance. By constructing a bank of local models, each valid in a subset of conditions, and employing an optimisation algorithm over a finite horizon, multimodel predictive control reconciles competing objectives such as constraint satisfaction, stability and disturbance rejection. A supervisory mechanism evaluates the current state, selects or blends the most appropriate local models and computes control inputs that steer the process towards set-points while respecting physical and operational limits. This approach enhances feasibility and robustness in the face of parameter uncertainties, set-point changes and switching between distinct modes of operation. Practical realisations span chemical reactors, energy systems and structural vibration suppression, where rapid adaptation to evolving dynamics and stringent performance criteria are paramount. Recent advances have refined model distribution, switching strategies and optimisation routines, driving global interest in scalable, efficient control solutions for high-dimensional, nonlinear applications.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Multimodel Predictive Control of Nonlinear Systems publication trend

The graph below shows the total number of articles in multimodel predictive control of nonlinear systems across all publications each year (not limited to Nature Index journals).

Technical terms

Model Predictive Control (MPC): A control strategy that solves an optimisation problem over a finite prediction horizon to determine future control actions while respecting constraints.

Multimodel predictive control: An MPC variant that employs a set of local models to capture different operating regimes of a nonlinear system and switches or blends them to compute control inputs.

Switched system: A dynamical system characterised by a finite set of subsystems and a rule that governs transitions between them.

Mixed Logical Dynamical (MLD) model: A formalism that represents hybrid systems by combining continuous dynamics with logical (discrete) conditions, enabling mixed-integer optimisation.

Soft-switching: A technique that gradually transitions between controllers or models to avoid abrupt changes and maintain feasibility and stability.

References

1. On Feasibility and Asymptotically Stability of Switched Systems Using Adaptive Multi-Model Predictive Control. Complex System Modeling and Simulation (2025).
2. MLD-MPC Approach for Three-Tank Hybrid Benchmark Problem. Computers Materials & Continua (2023).
3. Robust multi‐model controllers design without impulse in switching times for non‐linear vibrations suppression of sandwich plate. IET Control Theory and Applications (2022).