L. Markolf, A. Siewert and O. Stursberg: "Approximations of Optimal Control Laws for Constrained Piecewise Affine Systems by Deep Neural Networks",Proc. of the 2024 European Control Conference, pp. 946-952, 2024

Abstract

The paper on hand considers the optimal control of piecewise affine systems subject to polytopic constraints. While this problem can be addressed by receding horizon control, the approach is known to be computationally demanding. This paper considers the approximation of receding horizon control laws by deep artificial feed-forward neural networks. The concept of projecting inadmissible inputs onto regions derived from feasible sets is extended to the considered problem setup in order to achieve deterministic guarantees on feasibility and constraint satisfaction. Two approaches are proposed and illustrated in numerical examples.

BibTex

@article{markolf2024approximations,
 abstract = {The paper on hand considers the optimal control of piecewise affine systems subject to polytopic constraints. While this problem can be addressed by receding horizon control, the approach is known to be computationally demanding. This paper considers the approximation of receding horizon control laws by deep artificial feed-forward neural networks. The concept of projecting inadmissible inputs onto regions derived from feasible sets is extended to the considered problem setup in order to achieve deterministic guarantees on feasibility and constraint satisfaction. Two approaches are proposed and illustrated in numerical examples.},
 author = {Markolf, L. and Siewert, A. and Stursberg, O.},
 journal = {Proc. of the 2024 European Control Conference},
 pages = {946-952},
 title = {Approximations of Optimal Control Laws for Constrained Piecewise Affine Systems by Deep Neural Networks},
 year = 2024
}

URL