A constraint solver written completely in Julia. It is heavily inspired by MiniCP. The intent of this project is for pedagogical purposes. However, as development has continued, it seems like in time, it shall be released for use in development & production systems.
This is a very very high level view of the components that make up the Engine. The problems the solver aims to provide a means to solve are Discrete and Deterministic. So far, there's only support for Integer variables - though not or long - rich variables will be introduced in time.
This work will be accompanies with an online book describing everything in it.
The current implementation includes:
- Domains
- AbstractDomain functions
- Domain Listener implementation
- StateBitSet
- StateSparseSet
- SparseSetDomain
Here's an overview of the domain representaion:
- Variables
- AbstractVariable functions
- IntVar implementation
- IntVarArray implementation
- Variable views:
- IntVarMultView
- IntVarOffsetView
- Constraint shell definition
- AbstractConstraint functions
- ConstraintClosure
- A set of shared interfaces
- AbstractConstraint
- AbstractDomain{T}
- AbstractDomainListener
- AbstractVariable{T}
- AbstractSolver
- AbstractObjective
- Objectives
- Minimize
- Solver
- LearnieCP the core solver
- InnerCore - module that export the core functions & objects
-
AbstractSolverException
- InconsistencyException
- NotImplementedException
- EmptyBackUpException
-
Exceptions are still experimental. The aim is to put them in distinct categories & sub-categories that will be handled in precise ways by the solver while also providing exact messages, where necessary, to the user.
- Refined Constraints
- DFSearch
- SearchStatistics
- Think through adding more features
- ConstNotEqual
- ConstEqual
- NotEqual
- Equal
- Global Constraints
- Sum
- Element2D
- Element1D
- Element1DVar
- Table Constraint
- AllDifferentDC constraint
- Tested
- Circuit Constraint
- Tested
- Scheduling constraint
- Cumulative Constraint
- Tested
- Disjunctive Constraint
- Tested
- Cumulative Constraint
- SelectMin
- FirstFail
- And
- BackUp representation
- StateManagers
- AbstractStateManager
- Copier
- Trailer
- AbstractStateManager
- AbstractState
- Copy
- Trail
- StateEntry (used as entries into the backup store)
- StateStack (for constraints in the domain listener)
- StateInt
A sample representation of the copier state manager.
- The Book Repo is here
- backtracking
- NQueens
- SEND + MORE = MONEY
- Oughts to change based on the changes to the sum constraint interface
- Sudoku
- Sum Test (testing the sum constraint)
- NQueens
- Quadratic Assignment Problem
- Stable Matching Problem
- Eternity Puzzle
- TSP
Here's a small implementation of the NQueens problem implementation in LearnieCP.
Some of these examples are accompanied by plots, especially the ones that have an objective value to be minimized. All these plots will be included in the book. A sample plot can be seen here:
This Ciruit constraint was used here on a 15x15 grid. However, these results are from an exact search algorithm.
TAs with course and block details dataframe:
Courses with respective TAs dataframe:
- Constraints
- More localized constraints
- Branching Schema
- Using ML
- Large Neighbourhood search
- Local Search Heuristics
- Simulated Annaeling
- Random Search
- Tabu Search
- More examples
- VRP
- A Book on this approach
- Comparison with JuMP solvers.
- Continuous Variables
- Simplex or methods alike
- A smoother interface - probably link with
JuMP - Return the objective value of an optimization problem using
objectiveValuefunction - Solver state of constraints, variables and the objective
- Printing of solver state to show the model.
Contributions are welcome.