| 1 | /* |
| 2 | * CSP.h |
| 3 | * @brief Constraint Satisfaction Problem class |
| 4 | * @date Feb 6, 2012 |
| 5 | * @author Frank Dellaert |
| 6 | */ |
| 7 | |
| 8 | #pragma once |
| 9 | |
| 10 | #include <gtsam/discrete/DiscreteFactorGraph.h> |
| 11 | #include <gtsam_unstable/discrete/AllDiff.h> |
| 12 | #include <gtsam_unstable/discrete/SingleValue.h> |
| 13 | |
| 14 | namespace gtsam { |
| 15 | |
| 16 | /** |
| 17 | * Constraint Satisfaction Problem class |
| 18 | * A specialization of a DiscreteFactorGraph. |
| 19 | * It knows about CSP-specific constraints and algorithms |
| 20 | */ |
| 21 | class GTSAM_UNSTABLE_EXPORT CSP : public DiscreteFactorGraph { |
| 22 | public: |
| 23 | using Values = DiscreteValues; ///< backwards compatibility |
| 24 | |
| 25 | /// Add a unary constraint, allowing only a single value |
| 26 | void addSingleValue(const DiscreteKey& dkey, size_t value) { |
| 27 | emplace_shared<SingleValue>(args: dkey, args&: value); |
| 28 | } |
| 29 | |
| 30 | /// Add a binary AllDiff constraint |
| 31 | void addAllDiff(const DiscreteKey& key1, const DiscreteKey& key2) { |
| 32 | emplace_shared<BinaryAllDiff>(args: key1, args: key2); |
| 33 | } |
| 34 | |
| 35 | /// Add a general AllDiff constraint |
| 36 | void addAllDiff(const DiscreteKeys& dkeys) { emplace_shared<AllDiff>(args: dkeys); } |
| 37 | |
| 38 | // /** return product of all factors as a single factor */ |
| 39 | // DecisionTreeFactor product() const { |
| 40 | // DecisionTreeFactor result; |
| 41 | // for(const sharedFactor& factor: *this) |
| 42 | // if (factor) result = (*factor) * result; |
| 43 | // return result; |
| 44 | // } |
| 45 | |
| 46 | // /* |
| 47 | // * Perform loopy belief propagation |
| 48 | // * True belief propagation would check for each value in domain |
| 49 | // * whether any satisfying separator assignment can be found. |
| 50 | // * This corresponds to hyper-arc consistency in CSP speak. |
| 51 | // * This can be done by creating a mini-factor graph and search. |
| 52 | // * For a nine-by-nine Sudoku, the search tree will be 8+6+6=20 levels |
| 53 | // deep. |
| 54 | // * It will be very expensive to exclude values that way. |
| 55 | // */ |
| 56 | // void applyBeliefPropagation(size_t maxIterations = 10) const; |
| 57 | |
| 58 | /* |
| 59 | * Apply arc-consistency ~ Approximate loopy belief propagation |
| 60 | * We need to give the domains to a constraint, and it returns |
| 61 | * a domain whose values don't conflict in the arc-consistency way. |
| 62 | * TODO: should get cardinality from DiscreteKeys |
| 63 | */ |
| 64 | Domains runArcConsistency(size_t cardinality, |
| 65 | size_t maxIterations = 10) const; |
| 66 | |
| 67 | /// Run arc consistency for all variables, return true if any domain changed. |
| 68 | bool runArcConsistency(const VariableIndex& index, Domains* domains) const; |
| 69 | |
| 70 | /* |
| 71 | * Create a new CSP, applying the given Domain constraints. |
| 72 | */ |
| 73 | CSP partiallyApply(const Domains& domains) const; |
| 74 | }; // CSP |
| 75 | |
| 76 | } // namespace gtsam |
| 77 | |