Stefan Ratschan

Safety verification of hybrid systems by constraint propagation-based abstraction refinement

This paper deals with the problem of safety verification of nonlinear hybrid systems. We start from a classical method that uses interval arithmetic to check whether trajectories can move over the boundaries in a rectangular grid. We put this method into an abstraction refinement framework and improve it by developing an additional refinement step that employs interval-constraint propagation to add information to the abstraction without introducing new grid elements. Moreover, the resulting method allows switching conditions, initial states, and unsafe states to be described by complex constraints, instead of sets that correspond to grid elements. Nevertheless, the method can be easily implemented, since it is based on a well-defined set of constraints, on which one can run any constraint propagation-based solver. Tests of such an implementation are promising.

Efficient solving of quantified inequality constraints over the real numbers

Let a quantified inequality constraint over the reals be a formula in the first-order predicate language over the structure of the real numbers, where the allowed predicate symbols are ≤ and <. Solving such constraints is an undecidable problem when allowing function symbols such sin or cos. In this article, we give an algorithm that terminates with a solution for all, except for very special, pathological inputs. We ensure the practical efficiency of this algorithm by employing constraint programming techniques.

Safety verification of hybrid systems by constraint propagation based abstraction refinement

This paper deals with the problem of safety verification of non-linear hybrid systems. We start from a classical method that uses interval arithmetic to check whether trajectories can move over the boundaries in a rectangular grid. We put this method into an abstraction refinement framework and improve it by developing an additional refinement step that employs constraint propagation to add information to the abstraction without introducing new grid elements. Moreover, the resulting method allows switching conditions, initial states and unsafe states to be described by complex constraints instead of sets that correspond to grid elements. Nevertheless, the method can be easily implemented since it is based on a well-defined set of constraints, on which one can run any constraint propagation based solver. First tests of such an implementation are promising.

Providing a Basin of Attraction to a Target Region of Polynomial Systems by Computation of Lyapunov-Like Functions

In this paper, we present a method for computing a basin of attraction to a target region for polynomial ordinary differential equations. This basin of attraction is ensured by a Lyapunov-like polynomial function that we compute using an interval based branch-and-relax algorithm. This algorithm relaxes the necessary conditions on the coefficients of the Lyapunov-like function to a system of linear interval inequalities that can then be solved exactly. It iteratively refines these relaxations in order to ensure that, whenever a nondegenerate solution exists, it will eventually be found by the algorithm. Application of an implementation to a range of benchmark problems shows the usefulness of the approach.

Continuous First-Order Constraint Satisfaction

This paper shows how to use constraint programming techniques for solving first-order constraints over the reals (i.e., formulas in the first-order predicate language over the structure of the real numbers). More specifically, based on a narrowing operator that implements an arbitrary notion of consistency for atomic constraints over the reals (e.g., box-consistency), the paper provides a narrowing operator for first-order constraints that implements a corresponding notion of first-order consistency, and a solver based on such a narrowing operator. As a consequence, this solver can take over various favorable properties from the field of constraint programming.

GUARANTEED TERMINATION IN THE VERIFICATION OF LTL PROPERTIES OF NON-LINEAR ROBUST DISCRETE TIME HYBRID SYSTEMS

We present a novel approach to the automatic verification and falsification of LTL requirements of non-linear discrete-time hybrid systems. The verification tool uses an interval-based constraint solver for non-linear robust constraints to compute incrementally refined abstractions. Although the problem is in general undecidable, we prove termination of abstraction refinement based verification and falsification of such properties for the class of non-linear robust discrete-time hybrid systems. We argue, that—in industrial practice—safety critical control applications give rise to hybrid systems that are robust. We give first results on the application of this approach to a variant of an aircraft collision avoidance protocol.

Quantified constraints under perturbation

Quantified constraints (i.e. first-order formulae over the real numbers) are often exposed to perturbations: usually constants that come from measurements are only known up to certain precision, and numerical methods only compute with approximations of real numbers. In this paper we study the behavior of quantified constraints under perturbation by showing that one can formulate the problem of solving quantified constraints as a nested parametric optimization problem followed by one sign computation. Using the fact that minima and maxima are stable under perturbation, but the sign of a real number is stable only for non-zero inputs, we derive practically useful conditions for the stability of quantified constraints under perturbation.

Approximate Quantified Constraint Solving by Cylindrical Box Decomposition

This paper applies interval methods to a classical problem in computer algebra. Let a quantified constraint be a first-order formula over the real numbers. As shown by A. Tarski in the 1930s, such constraints, when restricted to the predicate symbols <, = and function symbols +, ×, are in general solvable. However, the problem becomes undecidable, when we add function symbols like sin. Furthermore, all exact algorithms known up to now are too slow for big examples, do not provide partial information before computing the total result, cannot satisfactorily deal with interval constants in the input, and often generate huge output. As a remedy we propose an approximation method based on interval arithmetic. It uses a generalization of the notion of cylindrical decomposition—as introduced by G. Collins. We describe an implementation of the method and demonstrate that, for quantified constraints without equalities, it can efficiently give approximate information on problems that are too hard for current exact methods.

Search Heuristics for Box Decomposition Methods

In this paper we study search heuristics for box decomposition methods that solve problems such as global optimization, minimax optimization, or quantified constraint solving. For this we unify these methods under a branch-and-bound framework, and develop a model that is more convenient for studying heuristics for such algorithms than the traditional models from Artificial Intelligence. We use the result to prove various theorems about heuristics and apply the outcome to the box decomposition methods under consideration. We support the findings with timings for the method of quantified constraint solving developed by the author.

Guaranteed termination in the verification of LTL properties of non-linear robust discrete time hybrid systems

We present a novel approach to the automatic verification and falsification of LTL requirements of non-linear discrete-time hybrid systems. The verification tool uses an interval-based constraint solver for non-linear robust constraints to compute incrementally refined abstractions. Although the problem is in general undecidable, we prove termination of abstraction refinement based verification and falsification of such properties for the class of robust non-linear hybrid systems, thus significantly extending previous semi-decidability results. We argue, that safety critical control applications are robust hybrid systems. We give first results on the application of this approach to a variant of an aircraft collision avoidance protocol.