Compositional Synthesis for Symbolic Control

Antoine Girard (1)

1 L2S - Laboratoire des signaux et systèmes

Symbolic control aims at designing "correct by construction" controllers for continuous dynamical systems, by using algorithmic discrete synthesis techniques. The key concept in symbolic control is that of symbolic model (also called finite abstraction), which is a finite-state dynamical system, obtained by abstracting continuous trajectories over a finite set of symbols. When the symbolic and the continuous dynamics are formally related by some behavioral relationship (e.g. simulation or bisimulation relations), controllers synthesized for the symbolic model using discrete synthesis techniques can be refined to certified controllers for the original continuous system. Computation of finite abstractions is often based on dis-cretization of the state and input spaces and therefore the symbolic control approach suffers from scalability issues. However, the design of large systems can still be tackled by means of compositional techniques. In this talk, we will present some recent results on compositional synthesis in the symbolic control approach. Firstly, we will present an approach to compute abstractions of systems made of several, possibly overlapping components. Secondly, we will show how to synthesize decentralized (and possibly asynchro-nous) controllers for invariance properties, by combining these overlapping abstractions and assume-guarantee contracts. In the last part of the talk, motivated by the use of parametric assume-guarantee contracts for stability properties, we will show recent developments on abstraction-based quantitative synthesis.

Mots clés: Symbolic control; abstraction; quantitative synthesis; compositional synthesis

Domaines: Optimisation et contrôle [math.OC]; Automatique / Robotique; Systèmes et contrôle [cs.SY]

Antoine Girard. Compositional Synthesis for Symbolic Control. 21st International Conference on Hybrid Systems: Computation and Control, Apr 2018, Porto, Portugal. pp.136-136, ⟨10.1145/3178126.3196957⟩. ⟨hal-02053194⟩