We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Systems and Control

Title: Synthesis of Stabilizing Recurrent Equilibrium Network Controllers

Abstract: We propose a parameterization of a nonlinear dynamic controller based on the recurrent equilibrium network, a generalization of the recurrent neural network. We derive constraints on the parameterization under which the controller guarantees exponential stability of a partially observed dynamical system with sector bounded nonlinearities. Finally, we present a method to synthesize this controller using projected policy gradient methods to maximize a reward function with arbitrary structure. The projection step involves the solution of convex optimization problems. We demonstrate the proposed method with simulated examples of controlling nonlinear plants, including plants modeled with neural networks.
Comments: Submitted to IEEE CDC 2022. arXiv admin note: text overlap with arXiv:2109.03861
Subjects: Systems and Control (eess.SY); Artificial Intelligence (cs.AI); Robotics (cs.RO)
Cite as: arXiv:2204.00122 [eess.SY]
  (or arXiv:2204.00122v3 [eess.SY] for this version)

Submission history

From: Neelay Junnarkar [view email]
[v1] Thu, 31 Mar 2022 22:27:51 GMT (10254kb,D)
[v2] Wed, 25 May 2022 16:32:57 GMT (10329kb,D)
[v3] Tue, 13 Sep 2022 00:49:06 GMT (10348kb,D)

Link back to: arXiv, form interface, contact.