eduzhai > Applied Sciences > Engineering >

Neural Certificates for Safe Control Policies

  • Save

... pages left unread,continue reading

Document pages: 13 pages

Abstract: This paper develops an approach to learn a policy of a dynamical system thatis guaranteed to be both provably safe and goal-reaching. Here, the safetymeans that a policy must not drive the state of the system to any unsaferegion, while the goal-reaching requires the trajectory of the controlledsystem asymptotically converges to a goal region (a generalization ofstability). We obtain the safe and goal-reaching policy by jointly learning twoadditional certificate functions: a barrier function that guarantees the safetyand a developed Lyapunov-like function to fulfill the goal-reachingrequirement, both of which are represented by neural networks. We show theeffectiveness of the method to learn both safe and goal-reaching policies onvarious systems, including pendulums, cart-poles, and UAVs.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×