Interval Constraint Propagation in SMT Compliant Decision Procedures

There is a wide range of decision procedures available for solving the existential fragment of first order theory of linear real algebra (QFLRA). However, for formulas of the theory of quantifier-free nonlinear real arithmetic (QFNRA), which are much harder to solve, there are only few decision procedures (the lower bound for complete solvers is exponential). The context this thesis is settled in is the software project SMT-RAT, a software framework for SAT Modulo Theories (SMT) solving. SMT solving is a combination of a SAT solver, which checks the Boolean skeleton of a given input formula and a theory solver, which handles the involved theory constraints. SMT-RAT maintains different complete and incomplete solving modules and allows to combine several modules to operate as a theory solver. Interval constraint propagation (ICP) is an incomplete decision procedure to efficiently reduce the domain of a set of variables with respect to a conjunction of polynomial constraints. The goal of this thesis is to present a module based on ICP for SMT-RAT. This module takes a conjunction of polynomial constraints as well as an initial set of boundaries, represented by intervals, for the variables occurring in the constraints as an input. It utilizes an interval extension of Newton’s method to reduce the domain of the given variables up to a certain bound and passes the set of constraints as well as the new boundaries to complete solvers, such as the module based on the cylindrical algebraic decomposition (CAD).

[1]  Frédéric Benhamou,et al.  Algorithm 852: RealPaver: an interval solver using constraint satisfaction techniques , 2006, TOMS.

[2]  Bruno Dutertre,et al.  Integrating Simplex with DPLL(T ) , 2006 .

[3]  Stefan Herbort,et al.  Improving the Efficiency of a Nonlinear-System-Solver Using a Componentwise Newton Method , 1997 .

[4]  Ramon E. Moore.  A Test for Existence of Solutions to Nonlinear Systems , 1977 .

[5]  Cesare Tinelli,et al.  DPLL( T): Fast Decision Procedures , 2004, CAV.

[6]  Martin Fränzle,et al.  Efficient Solving of Large Non-linear Arithmetic Constraint Systems with Complex Boolean Structure , 2007, J. Satisf. Boolean Model. Comput..

[7]  Erika Ábrahám,et al.  Virtual Substitution for SMT-Solving , 2011, FCT.

[8]  George E. Collins,et al.  Quantifier elimination for real closed fields by cylindrical algebraic decomposition , 1975 .

[9]  David A. McAllester,et al.  Solving Polynomial Systems Using a Branch and Prune Approach , 1997 .

[10]  L. G. H. Cijan.  A polynomial algorithm in linear programming , 1979 .

[11]  Heinz Kredel,et al.  Gröbner Bases: A Computational Approach to Commutative Algebra , 1993 .

[12]  Frédéric Goualard,et al.  A Reinforcement Learning Approach to Interval Constraint Propagation , 2008, Constraints.

[13]  George B. Dantzig,et al.  Linear programming and extensions , 1965 .

[14]  Volker Weispfenning,et al.  The Complexity of Linear Problems in Fields , 1988, Journal of symbolic computation.

[15]  L. Khachiyan.  Polynomial algorithms in linear programming , 1980 .

[16]  Sebastian Junges,et al.  SMT-RAT: An SMT-Compliant Nonlinear Real Arithmetic Toolbox - (Tool Presentation) , 2012, SAT.

[17]  Eldon R. Hansen,et al.  Interval forms of Newtons method , 1978, Computing.

[18]  Ulrich W. Kulisch,et al.  Complete Interval Arithmetic and Its Implementation on the Computer , 2009, Numerical Validation in Current Hardware Architectures.

[19]  Sriram Sankaranarayanan,et al.  Integrating ICP and LRA solvers for deciding nonlinear real arithmetic problems , 2010, Formal Methods in Computer Aided Design.