We propose an efficient technique for performing data-driven optimal control of discrete-time systems. In particular, we show that log-sum-exp ($lse$) neural networks, which are smooth and convex universal approximators of convex functions, can be efficiently used to approximate Q-factors arising from finite-horizon optimal control problems with continuous state space. The key advantage of these networks over classical approximation techniques is that they are convex and hence readily amenable to efficient optimization.

Efficient model-free Q-factor approximation in value space via log-sum-exp neural networks / Calafiore, GIUSEPPE CARLO; Possieri, Corrado. - ELETTRONICO. - (2020). ((Intervento presentato al convegno European Control Conference (ECC2020) tenutosi a Saint Petersburg, Russia nel 12-15 May, 2020 [10.23919/ECC51009.2020.9143765].

Efficient model-free Q-factor approximation in value space via log-sum-exp neural networks

Giuseppe Calafiore;Corrado Possieri
2020

Abstract

We propose an efficient technique for performing data-driven optimal control of discrete-time systems. In particular, we show that log-sum-exp ($lse$) neural networks, which are smooth and convex universal approximators of convex functions, can be efficiently used to approximate Q-factors arising from finite-horizon optimal control problems with continuous state space. The key advantage of these networks over classical approximation techniques is that they are convex and hence readily amenable to efficient optimization.
File in questo prodotto:
File Dimensione Formato  
reinLSE_final.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 239.36 kB
Formato Adobe PDF
239.36 kB Adobe PDF Visualizza/Apri
Calafiore-Efficient.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 272.44 kB
Formato Adobe PDF
272.44 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11583/2837797