Abstract
Solving constraint satisfaction problems (CSP) efficiently depends on the solver configuration and the search strategy.
However, it is difficult to customize the constraint solvers because they are not modular enough, and it is hard to create new search strategies by composition.
To solve these problems, we propose spacetime programming, a paradigm based on lattices and synchronous process calculi that views search strategies as processes working collaboratively towards the resolution of a CSP.
We implement the compiler of the language and use it to replace the search module of Choco, a state of the art constraint solver, with an efficient spacetime program that offers better modularity and compositionality of search strategies.
Keywords
- Artificial Intelligence: computer science
- Artificial Intelligence: constraints
- Artificial Intelligence: search and constraint satisfaction
