[Back]


Talks and Poster Presentations (with Proceedings-Entry):

D. Phan, R. Grosu, N. Jansen, N. Paoletti, S. Smolka, S. Stoller:
"Neural simplex architecture";
Talk: NASA Formal Methods (NFM), virtuell; 2020-05-11 - 2020-05-15; in: "Neural simplex architecture", Springer, . (2020), ISBN: 978-3-030-55754-6; 97 - 114.



English abstract:
We present the Neural Simplex Architecture (NSA), a new
approach to runtime assurance that provides safety guarantees for neural
controllers (obtained e.g. using reinforcement learning) of autonomous
and other complex systems without unduly sacri cing performance. NSA
is inspired by the Simplex control architecture of Sha et al., but with
some signi cant di erences. In the traditional approach, the advanced
controller (AC) is treated as a black box; when the decision module
switches control to the baseline controller (BC), the BC remains in control
forever. There is relatively little work on switching control back to
the AC, and there are no techniques for correcting the AC's behavior
after it generates a potentially unsafe control input that causes a failover
to the BC. Our NSA addresses both of these limitations. NSA not only
provides safety assurances in the presence of a possibly unsafe neural
controller, but can also improve the safety of such a controller in an online
setting via retraining, without overly degrading its performance. To
demonstrate NSA's bene ts, we have conducted several signi cant case
studies in the continuous control domain. These include a target-seeking
ground rover navigating an obstacle eld, and a neural controller for an
arti cial pancreas system.


Electronic version of the publication:
https://publik.tuwien.ac.at/files/publik_293057.pdf


Created from the Publication Database of the Vienna University of Technology.