Multi-Sudoku AI approach

Question

I'm conceptualizing a solver for a variant of sudoku called multi-sudoku, where multiple boards overlap like so:

If I understand the game correctly, you must solve each grid in such a way that the overlap between any two or more grids is part of the solution for each.

I'm unsure as to how I should be thinking about this. Anybody got any hints/conceptual clues? Additionally, if any topics in artificial intelligence come to mind, I'd like to hear those too.

Answer 1

Evidently, there are other AI techniques to solve problems. A technique that is commonly used to tackle hard search and optimization problems in evolutionary computing: one starts by filling in the sudoku allowing some values to be wrong, then at each time step they aim to fix one or more fields. Sometimes they introduce new errors in order to find a valid solution eventually.

Solving the problem with ECLiPSe:

On the ECLiPSe website, you can find a model for sudoku. Given you read some documentation about ECLiPSe, you can turn this file into a model for multi-sudoku.

Here you can find the sample code for the same:

http://eclipseclp.org/examples/sudoku.ecl.txt

And here's a MiniZinc model using the same approach as the ECLiPSe model: hakank.org/minizinc/sudoku_multi.mzn