Title
Robust Abstractions for Control Synthesis: Completeness via Robustness for Linear-Time Properties.
Abstract
We define robust abstractions for synthesizing provably correct and robust controllers for (possibly infinite) uncertain transition systems. It is shown that robust abstractions are sound in the sense that they preserve robust satisfaction of linear-time properties. We then focus on discrete-time control systems modelled by nonlinear difference equations with inputs and define concrete robust abstractions for them. While most abstraction techniques in the literature for nonlinear systems focus on constructing sound abstractions, we present computational procedures for constructing both sound and approximately complete robust abstractions for general nonlinear control systems without stability assumptions. Such procedures are approximately complete in the sense that, given a concrete discrete-time control system and an arbitrarily small perturbation of this system, there exists a finite transition system that robustly abstracts the concrete system and is abstracted by the slightly perturbed system simultaneously. A direct consequence of this result is that robust control synthesis for discrete-time nonlinear systems and linear-time specifications is robustly decidable. More specifically, if there exists a robust control strategy that realizes a given linear-time specification, we can algorithmically construct a (potentially less) robust control strategy that realizes the same specification. The theoretical results are illustrated with a simple motion planning example.
Year
DOI
Venue
2017
10.1145/3049797.3049826
HSCC
Keywords
Field
DocType
Nonlinear systems, control synthesis, abstraction, robustness, linear-time property, linear temporal logic, decidability
Transition system,Nonlinear system,Nonlinear control,Control theory,Decidability,Robustness (computer science),Control system,Robust control,Time complexity,Mathematics
Conference
Citations 
PageRank 
References 
4
0.43
18
Authors
1
Name
Order
Citations
PageRank
Jun Liu121520.63