The backstory is that I'm a big fan of the game and was curious if I could model its logic using constraint programming. I used MiniZinc to build it. The project treats each puzzle as a constraint satisfaction problem—it defines the rules of the game (how tracks connect, how switches work, etc.) and then searches for a layout that solves the level.
It's different from other solvers because it's a declarative model rather than an imperative search algorithm. It was a fun challenge to translate mechanics like timed gates, dynamic switches, and decoy cars into formal constraints.
You can run it on your own machine. The repo has instructions, but the basics are: 1. Install the MiniZinc IDE/toolchain. 2. Clone the repo. 3. Run a command like minizinc --solver or-tools main.mzn data/1/1-1.dzn
The project is open source and currently solves the first 8 worlds of the game.
Repo: https://github.com/Th1nhNg0/railbound_cp
I'm here to answer any questions. I'd love to hear your feedback.
mzl•3mo ago
Have you considered using records for some of the data definitions? For example, a record `(row: int, col: int, piece: Piece)` for pre-placed pieces to avoid having .1, .2 and so on in the models. If you are concerned with the size of the data files, one way to handle that is to have the data as a list of tuples that are mapped via a function from a tulle to the record.
th1nhng0•3mo ago
In my local tests, Chuffed is often faster on the simpler instances. However, I've noticed on the harder puzzles, Chuffed can take over 10 minutes, while OR-Tools (with just 4 threads) solves them in about 5.
You're also right about the records. I started with tuples for simplicity, but as the model grew more complex, it's clear that records would be a much better approach for readability.
mzl•3mo ago
You might find it interesting to install and try both Huub (https://github.com/huub-solver/huub) and Pumpkin (https://github.com/consol-lab/pumpkin) as well. I have both in https://zayenz.se/blog/post/benchmarking-linkedin-queens/ with good results for the small problems I tested there. Both also did very well in the MiniZinc Challenge this year.
One thing that might be interesting (and might be useful for some solvers) is to change from checking if there is a solution and requiring each part to be used in the solution, to find solutions and minimize the number of unused parts.
One additional constraint that you could add that I don't think I've seen is distance to goal. In essence, for each grid position (without considering any tracks) there is a shortest time it would take from that position to get to the final position. If a train is in (a, b) and would take at least time t to get to the exit, there is no way to solve the problem if it is in (a, b) at time MAX_TIME-t. Might be useful for some solvers, worth testing.
Finally, adding some simple heuristic might help some solvers (I'm thinking of Gecode in particular here). My initial guess would be to add an interleaving over time steps of the train_row, train_col, and train dir for each train and search in order. That would as a basic heuristic represent a search from the start snaking out. Value choice is trickier to specify in MiniZinc of course.
th1nhng0•3mo ago
The "distance to goal" constraint is a clever idea, though it's tricky to implement because the tunnels act as teleporters, so a simple distance metric doesn't apply. I'd likely need to pre-calculate a shortest-path graph first.
I'm benchmarking Huub and Pumpkin now. My main goal is raw performance, as a friend's imperative brute-force solver (https://github.com/FoxtrotOnce/RailboundSolver) solves most levels in under 2 seconds. It's a high bar to meet with a declarative model, but I'm keen to see how close I can get with these optimizations.
I've also just pushed some initial benchmarks for Chuffed vs. OR-Tools to the repo if you're interested (https://github.com/Th1nhNg0/railbound_cp/tree/master/docs/be...). Thanks again for the great feedback!