The MiniZinc Challenge compares different solvers on a set of MiniZinc models and problems instances. MiniZ- inc1 (Nethercote et al. 2007) was our response to the
call for a standard constraint-programming modeling language. MiniZinc is high level enough to express most combinatorial optimization problems easily and in a largely solver-independent way; for example, it supports sets, arrays, and
user-defined predicates, some overloading, and some automatic coercions. However, MiniZinc is low level enough that
it can be mapped easily onto many solvers. For example, it is
first order, and it only supports decision variable types that
are supported by most existing constraint-programming
solvers: integers, floats, Booleans, and sets of integers. MiniZinc also allows separation of a model from its data; provides
a library containing declarative definitions of many global
constraints; and has a system of annotations that allows nondeclarative information (such as search strategies) and solver-specific information (such as variable representations) to be
layered on top of declarative models.
The MiniZinc Challenge
Peter J. Stuckey, Thibaut Feydy, Andreas Schutt,
Guido Tack, Julien Fischer
n MiniZinc is a solver-agnostic modeling language for defining and solving
combinatorial satisfaction and optimization problems. MiniZinc provides a
solver-independent modeling language
that is now supported by constraint-programming solvers, mixed integer programming solvers, SAT and SAT modulo theory solvers, and hybrid solvers.
Every year since 2008 we have run the
MiniZinc Challenge, which compares
and contrasts the different strengths of
different solvers and solving technologies on a set of MiniZinc models. Here
we report on what we have learned from
running the competition for 6 years.