Loading...
 
[Show/Hide Left Column]

Compositional Controller Synthesis for Hybrid systems

  • Dans le cadre du projet émergence CODECSYS

Axis : Scilex
Subject : compositional controller synthesis for hybrid systems
Thesis Supervisor : Antoine GIRARD (L2S), Laurent FRIBOURG (LSV).
Institutions :Laboratoire des signaux et systèmes (L2S), Laboratoire de spécification et vérification (LSV).
PhD : Adnane SAOUD
Beginning : October 2016
Scientific productions :
  • Adnane Saoud and Antoine Girard, Multirate symbolic models for incrementally stable switched systems. IFAC World Congress, Toulouse, France, 2017.
  • Adnane Saoud, Antoine Girard, Optimal multirate sampling in symbolic models for incrementally stable switched systems. Automatica, To appear.
  • Adnane Saoud, Antoine Girard, Laurent Fribourg, Contract based design of symbolic controllers for interconnected multiperiodic sampled-data systems. IEEE Conference on Decision and Control, Miami, USA, 2018.
  • Zohra Kader, Antoine Girard and Adnane Saoud, Symbolic models for incrementally stable switched systems with aperiodic time sampling. IFAC Conference on Analysis and Design of Hybrid Systems, Oxford, UK, 2018.
  • Adnane Saoud, Pushpak Jagtap, Majid Zamani and Antoine Girard, Compositional abstraction-based synthesis for cascade discrete-time control systems. IFAC Conference on Analysis and Design of Hybrid Systems, Oxford, UK, 2018.
  • Adnane Saoud, Antoine Girard and Laurent Fribourg, On the composition of discrete and continuous-time assume-guarantee contracts for invariance. European Control Conference, Limassol, Cyprus, 2018. Finalist of the Best Student Paper Award.
  • Adnane Saoud and Antoine Girard, Multirate symbolic models for incrementally stable switched systems. IFAC World Congress, Toulouse, France, 2017.

Context :
Cyber-physical systems (CPS) result from integrations of computational devices with physical processes and are to become ubiquitous in modern societies (autonomous vehicles, smart buildings, robots, etc.). The development of rigorous model based approaches to the design of CPS therefore constitutes a major challenge for the future years. Hybrid systems are natural models of CPS enabling to capture the tight interactions between "discrete" computing devices with the "continuous" physical world. Despite considerable progress in the field, current techniques apply to hybrid systems of moderate complexity.Thus, the design of complex CPS requires to divide large design problems in smaller sub-problems that can be solved using existing tools. The CODECSYS project aims at developing such approaches by decomposing a complex CPS into components which are designed independently. Each component is assigned a contract, which specifies guarantees that the component must fulfill under assumptions on the behavior of other components. For a given desired behavior of the global system, the decomposition into contracts to be satisfed by components is generally not unique: some contracts may be infeasible by components, resulting in an unsuccessful overall design; and even when all contracts can be satisfed, their choice may impact the robustness of the overall system.The CODECSYS project will contribute to contract based design of CPS by developing novel approaches, which explore systematically the space of possible design contracts. For that purpose, we consider contracts that are given by parametric assumptions and guarantees. For each component, we characterize a feasible region of parameter values for which the corresponding contract can be satisfied. The intersection of these feasible regions provides parameter values, which guarantee the correct behavior of the overall system. Moreover, one can also search for a particular parameter value, which optimizes the robustness of the design. Efficient computational techniques for contract parameter synthesis will be developed for design contracts that enjoy some monotonicity property with respect to parameters, using multiobjective optimization algorithms. Connections with control theoretic techniques such as small-gain theorems will be investigated. The development of our approach will be motivated by and validated on two applications: the first one deals with compositional controller synthesis for hybrid systems, the second one deals with digital implementation and scheduling of embedded controllers.