Event

Proposes a verifiable safe synthesis framework for stochastic neural control barrier functions

1 reports · 1 sources

Hongchao Zhang et al. propose a verifiable safe synthesis framework for stochastic neural control barrier functions (SNCBFs). For the two cases of smooth activation functions and ReLU, they respectively present a verification-free synthesis framework and a verification-in-the-loop synthesis framework, and complete validation on three models: the inverted pendulum, Darboux, and the unicycle.