220
Neural Controller for Incremental Stability of Unknown Continuous-time Systems
arXiv:2504.18330v2 Announce Type: replace
Abstract: This work primarily focuses on synthesizing a controller that guarantees an unknown continuous-time system to be incrementally input-to-state stable ($\delta$-ISS). In this context, the notion of $\delta$-ISS control Lyapunov function ($\delta$-ISS-CLF) for the continuous-time system is introduced. Combined with the controller, the $\delta$-ISS-CLF guarantees that the system is incrementally stable. As the paper deals with unknown dynamical systems, the controller as well as the $\delta$-ISS-CLF are parametrized using neural networks. The data set used to train the neural networks is generated from the state space of the system by proper sampling. Now, to give a formal guarantee that the controller makes the system incrementally stable, we develop a validity condition by having some Lipschitz continuity assumptions and incorporate the condition into the training framework to ensure a provable correctness guarantee at the end of the training process. Finally, we demonstrate the effectiveness of the proposed approach through several case studies: a scalar system with a non-affine, non-polynomial structure, a one-link manipulator system, a nonlinear Moore-Greitzer model of a jet engine, a magnetic levitator system and a rotating rigid spacecraft model.
Abstract: This work primarily focuses on synthesizing a controller that guarantees an unknown continuous-time system to be incrementally input-to-state stable ($\delta$-ISS). In this context, the notion of $\delta$-ISS control Lyapunov function ($\delta$-ISS-CLF) for the continuous-time system is introduced. Combined with the controller, the $\delta$-ISS-CLF guarantees that the system is incrementally stable. As the paper deals with unknown dynamical systems, the controller as well as the $\delta$-ISS-CLF are parametrized using neural networks. The data set used to train the neural networks is generated from the state space of the system by proper sampling. Now, to give a formal guarantee that the controller makes the system incrementally stable, we develop a validity condition by having some Lipschitz continuity assumptions and incorporate the condition into the training framework to ensure a provable correctness guarantee at the end of the training process. Finally, we demonstrate the effectiveness of the proposed approach through several case studies: a scalar system with a non-affine, non-polynomial structure, a one-link manipulator system, a nonlinear Moore-Greitzer model of a jet engine, a magnetic levitator system and a rotating rigid spacecraft model.
Anonymous
No comments yet.