Approximate Robust Control of Uncertain Dynamical Systems

Edouard Leurent (1, 2, 3) , Yann Blanco (3) , Denis Efimov (2) , Odalric-Ambrym Maillard (1)

1 SEQUEL - Sequential Learning
2 NON-A-POST - Non-Asymptotic estimation for online systems
3 RENAULT

This work studies the design of safe control policies for large-scale non-linear systems operating in uncertain environments. In such a case, the robust control framework is a principled approach to safety that aims to maximize the worst-case performance of a system. However, the resulting optimization problem is generally intractable for non-linear systems with continuous states. To overcome this issue, we introduce two tractable methods that are based either on sampling or on a conservative approximation of the robust objective. The proposed approaches are applied to the problem of autonomous driving.

Domaines: Intelligence artificielle [cs.AI]; Robotique [cs.RO]

Edouard Leurent, Yann Blanco, Denis Efimov, Odalric-Ambrym Maillard. Approximate Robust Control of Uncertain Dynamical Systems. Proc. MLITS Workshop at NeurIPS, Dec 2018, Montreal, Canada. ⟨hal-01931744v2⟩