Browsing by Author "THIAGARAJAN, P. S."

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    The Discrete Time Behaviour of Restricted Linear Hybrid Automata
    (2008-12-15T01:59:32Z) AGRAWAL, Manindra; STEPHAN, Frank; THIAGARAJAN, P. S.; YANG, Shaofa
    We summarize results from [2, 3, 1] on the discrete time behaviour of a class of restricted linear hybrid automata. Specifically, we show the regularity of the discrete time behaviour of hybrid automata in which the rates of continuous vari- ables are governed by linear operators in a diagonal form and in which the values of the continuous variables can be observed only with finite precision. Crucially, we do not demand—as is usually done—that the values of the continuous vari- ables be reset during mode changes. We can cope with polynomial guards and we can tolerate bounded delays both in sampling the values of the continuous variables and in effecting changes in their rates required by mode switchings. We also show that if the rates are governed by diagonalizable linear operators with rational eigenvalues and there is no delay in effecting rate changes, the discrete time behaviour of the hybrid automaton is recursive. However, the control state reachability problem in this setting is undecidable.
  • No Thumbnail Available
    Item
    Interacting Process Classes
    (2005-09-07T06:47:59Z) GOEL, Ankit; SUN, Meng; ROYCHOUDHURY, Abhik; THIAGARAJAN, P. S.
    Many reactive control systems consist of a large number of similar interacting objects; these objects can be often grouped into classes. Such interacting process classes appear in telecommunication, transportation and avionics domains. In this paper, we propose a modeling and simulation technique for interacting process classes. Our modeling style uses well-known UML notations to capture behavior. In particular, the control flow of a process class is captured by a state diagram, unit interactions between process objects by sequence diagrams and the structural relations are captured via class diagrams. The key feature of our approach is that our simulation is symbolic. We dynamically group together objects of the same class based on their past behavior. This leads to a simulation strategy that is both time and memory efficient and we demonstrate this on well-studied non-trivial examples of reactive systems. We also use our simulator for debugging realistic designs such as NASA's CTAS weather monitoring system.