Project : popart
Section: New Results
Automatic generation of correct controllers
The control of multi-mode multi-tasking systems
Work in the last few years has produced a methodology for the automatic generation of correct controllers for multi-task systems . The model of commonly found task control patterns is proposed in terms of labeled transition systems, representing idle, waiting, or active states, and transitions in reaction to requests, authorizations and termination events. Quantitative weights can be associated to active states, representing costs (time, power consumption) or quality level. Standard properties of the interactions between such components are formulated, in terms of invariants or configurations that should be always reachable. When a system is modeled by composing instances of such patterns, discrete controller synthesis is applied to obtain automatically (if it exists) the controller of activations such that the properties are satisfied, and the weights are optimized. This work is done in cooperation with VERIMAG (Synchronous team) and IRISA/Inria-Rennes (VERTECS( http://www.irisa.fr/vertecs)).
We have begun considering the possible complementarities between the application of controller synthesis on the global model, and the use of composition ¨glue¨, also in terms of an automaton, that would enforce a given property between components. The idea is that, for some simple properties, this technique can avoid the costly synthesis, which would however be necessary for others. This issue raises considerations of compositionality as in Section 6.4.
Automated generation of property-enforcing layers
A generalization of the methodology has been formalized and defined as the automated generation of a property-enforcing layer . The component automata model the local constraints; the product of these automata is a first approximation of the set of constraints that should be respected. The constraints that involve several components are expressed as temporal logic properties of this product. We then use general controller synthesis techniques and tools in order to combine the set of communicating parallel automata with the global constraint.
In order to obtain automatically fault-tolerant real-time systems, we investigate a new solution based on the application of discrete controller synthesis. The real-time systems we consider consist of a set of tasks, and a set of distributed, heterogenous processors. The latter are fail-silent, and an environment model can detail actual fault patterns. We apply controller synthesis, with objectives w.r.t. consistent execution, functionality fulfillment, and some optimizations. We construct a manager that ensures fault-tolerance by migrating the tasks automatically, upon occurrence of a failure, according to the policy given by the objectives. The advantage is that, once the system is modeled, it becomes possible to study several fault-tolerance policies .
We also approach controller synthesis for fault-tolerance from another angle, related to production systems, in cooperation with LAG( http://www.lag.ensieg.inpg.fr) (H. Alla). The case study concerned a set of machine-tools, which could operate at full speed in their nominal mode, and at a lower speed when a failure occurred or some tool became worn off .