-
Mapping back and forth between model predictive control and neural networks
Authors:
Ross Drummond,
Pablo R Baldivieso-Monasterios,
Giorgio Valmorbida
Abstract:
Model predictive control (MPC) for linear systems with quadratic costs and linear constraints is shown to admit an exact representation as an implicit neural network. A method to "unravel" the implicit neural network of MPC into an explicit one is also introduced. As well as building links between model-based and data-driven control, these results emphasize the capability of implicit neural networ…
▽ More
Model predictive control (MPC) for linear systems with quadratic costs and linear constraints is shown to admit an exact representation as an implicit neural network. A method to "unravel" the implicit neural network of MPC into an explicit one is also introduced. As well as building links between model-based and data-driven control, these results emphasize the capability of implicit neural networks for representing solutions of optimisation problems, as such problems are themselves implicitly defined functions.
△ Less
Submitted 18 April, 2024;
originally announced April 2024.
-
Two-layer nonlinear control of DC-DC buck converters with meshed network topology
Authors:
Pablo R. Baldivieso-Monasterios,
Mahdieh Sadabadi,
George C. Konstantopoulos
Abstract:
In this paper, we analyse the behaviour of a buck converter network that contains arbitrary, up to mild regularity assumptions, loads. Our analysis of the network begins with the study of the current dynamics; we propose a novel Lyapunov function for the current in closed-loop with a bounded integrator. We leverage on these results to analyse the interaction properties between voltages and bounded…
▽ More
In this paper, we analyse the behaviour of a buck converter network that contains arbitrary, up to mild regularity assumptions, loads. Our analysis of the network begins with the study of the current dynamics; we propose a novel Lyapunov function for the current in closed-loop with a bounded integrator. We leverage on these results to analyse the interaction properties between voltages and bounded currents as well as between node voltages and to propose a two-layer optimal controller that keeps network voltages within a compact neighbourhood of the nominal operational voltage. We analyse the stability of the closed loop system in two ways: one considering the interconnection properties which yields a weaker ISS type property and a second that contemplates the network in closed loop with a distributed optimal controller. For the latter, we propose a novel distributed way of controlling a Laplacian network using neighbouring information which results in asymptotic stability. We demonstrate our results in a meshed topology network containing 6 power converters, each converter feeding an individual constant power load with values chaning arbitrarily within a pre-specified range.
△ Less
Submitted 26 December, 2021;
originally announced December 2021.
-
Coalitional predictive control: consensus-based coalition forming with robust regulation
Authors:
Pablo R Baldivieso-Monasterios,
Paul A Trodden
Abstract:
This paper is concerned with the problem of controlling a system of constrained dynamic subsystems in a way that balances the performance degradation of decentralized control with the practical cost of centralized control. We propose a coalitional control scheme in which controllers of subsystems may, as the need arises, group together into coalitions and operate as a single entity. The scheme emp…
▽ More
This paper is concerned with the problem of controlling a system of constrained dynamic subsystems in a way that balances the performance degradation of decentralized control with the practical cost of centralized control. We propose a coalitional control scheme in which controllers of subsystems may, as the need arises, group together into coalitions and operate as a single entity. The scheme employs a robust form of distributed model predictive control for which recursive feasibility and stability are guaranteed, yet---uniquely---the reliance on robust invariant sets is merely implicit, thus enabling applicability to higher-order systems. The robust control algorithm is combined with an algorithm for coalition forming based on consensus theory and potential games; we establish conditions under which controllers reach a consensus on the sets of coalitions. The recursive feasibility and closed-loop stability of the overall time-varying coalitional control scheme are established under a sufficient dwell time, the existence of which is guaranteed.
△ Less
Submitted 28 January, 2020;
originally announced January 2020.
-
Model predictive control of linear systems with preview information: feasibility, stability and inherent robustness
Authors:
Pablo R Baldivieso-Monasterios,
Paul A. Trodden
Abstract:
The use of available disturbance predictions within a nominal model predictive control formulation is studied. The main challenge that arises is the loss of recursive feasibility and stability guarantees when a persistent disturbance is present in the model and on the system. We show how standard stabilizing terminal conditions may be modified to account for the use of disturbances in the predicti…
▽ More
The use of available disturbance predictions within a nominal model predictive control formulation is studied. The main challenge that arises is the loss of recursive feasibility and stability guarantees when a persistent disturbance is present in the model and on the system. We show how standard stabilizing terminal conditions may be modified to account for the use of disturbances in the prediction model. Robust stability and feasibility are established under the assumption that the disturbance change across sampling instances is limited.
△ Less
Submitted 30 July, 2018;
originally announced July 2018.
