Research papers · Posted 2019.07.08 · Preprint
Robust Guarantees for Perception-Based Control
Robust Guarantees for Perception-Based Control
Overview
The method designs a safe set and robust controller for linear dynamics observed through learned perception. Under smoothness and sampling assumptions, it analyzes generalization and demonstrates the perception–control loop in synthetic and CARLA examples.
Author-supplied source information
Author-supplied information on arXiv.
This revision includes reframing the local generalization problem, with relaxed the assumptions so that the robust problem depends on a local slope bound rather than a Lipschitz constant, and provide a method for learning the slope bound from data. We also include additional experiments with a CNN perception module
Version: v2