Hashi Puzzle Solver
Hashi is a logic puzzle: numbered islands sit on a grid, and you join them with horizontal and vertical bridges, at most two between any pair, until every island has exactly as many bridges as its number and no two bridges cross. This AI course project models the puzzle as a constraint satisfaction problem and solves it with backtracking search. The solver always works on the island with the fewest remaining options, tries a single bridge and then a double in a fixed north, east, south, west order, checks bridge limits and crossings before every move, and undoes moves when a branch dead-ends. A second solver based on conflict-driven clause learning was started and never finished.
Attention: Coursework. Only the backtracking solver works; the planned second solver, based on conflict-driven clause learning, was never finished, as the repo's own README says.
Built
2025
Who built it
Two-person project with Ben Smith. I wrote the backtracking solver, most of the puzzle and bridge logic, the test puzzles, and the experiment runner; Ben set up the puzzle classes and wrote the unfinished constraint-learning solvers.
Repository
BenSmith1202/AIResearchProject1 (private)
Built with
- Python (standard library only)
Not a hosted project
Source only
A Python script that prints its results to the console. There is nothing to host.
The title above is my own descriptive name for this project. The repository name is BenSmith1202/AIResearchProject1.