OR-Tools  9.1
probing.h
Go to the documentation of this file.
1// Copyright 2010-2021 Google LLC
2// Licensed under the Apache License, Version 2.0 (the "License");
3// you may not use this file except in compliance with the License.
4// You may obtain a copy of the License at
5//
6// http://www.apache.org/licenses/LICENSE-2.0
7//
8// Unless required by applicable law or agreed to in writing, software
9// distributed under the License is distributed on an "AS IS" BASIS,
10// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
11// See the License for the specific language governing permissions and
12// limitations under the License.
13
14#ifndef OR_TOOLS_SAT_PROBING_H_
15#define OR_TOOLS_SAT_PROBING_H_
16
17#include "absl/types/span.h"
18#include "ortools/sat/clause.h"
20#include "ortools/sat/integer.h"
21#include "ortools/sat/model.h"
23#include "ortools/sat/util.h"
25
26namespace operations_research {
27namespace sat {
28
29class Prober {
30 public:
31 explicit Prober(Model* model);
32
33 // Fixes Booleans variables to true/false and see what is propagated. This
34 // can:
35 //
36 // - Fix some Boolean variables (if we reach a conflict while probing).
37 //
38 // - Infer new direct implications. We add them directly to the
39 // BinaryImplicationGraph and they can later be used to detect equivalent
40 // literals, expand at most ones clique, etc...
41 //
42 // - Tighten the bounds of integer variables. If we probe the two possible
43 // values of a Boolean (b=0 and b=1), we get for each integer variables two
44 // propagated domain D_0 and D_1. The level zero domain can then be
45 // intersected with D_0 U D_1. This can restrict the lower/upper bounds of a
46 // variable, but it can also create holes in the domain! This will detect
47 // common cases like an integer variable in [0, 10] that actually only take
48 // two values [0] or [10] depending on one Boolean.
49 //
50 // Returns false if the problem was proved INFEASIBLE during probing.
51 //
52 // TODO(user): For now we process the Boolean in their natural order, this is
53 // not the most efficient.
54 //
55 // TODO(user): This might generate a lot of new direct implications. We might
56 // not want to add them directly to the BinaryImplicationGraph and could
57 // instead use them directly to detect equivalent literal like in
58 // ProbeAndFindEquivalentLiteral(). The situation is not clear.
59 //
60 // TODO(user): More generally, we might want to register any literal => bound
61 // in the IntegerEncoder. This would allow to remember them and use them in
62 // other part of the solver (cuts, lifting, ...).
63 //
64 // TODO(user): Rename to include Integer in the name and distinguish better
65 // from FailedLiteralProbing() below.
66 bool ProbeBooleanVariables(double deterministic_time_limit);
67
68 // Same as above method except it probes only on the variables given in
69 // 'bool_vars'.
70 bool ProbeBooleanVariables(double deterministic_time_limit,
71 absl::Span<const BooleanVariable> bool_vars);
72
73 bool ProbeOneVariable(BooleanVariable b);
74
75 private:
76 bool ProbeOneVariableInternal(BooleanVariable b);
77
78 // Model owned classes.
79 const Trail& trail_;
80 const VariablesAssignment& assignment_;
81 IntegerTrail* integer_trail_;
82 ImpliedBounds* implied_bounds_;
83 SatSolver* sat_solver_;
84 TimeLimit* time_limit_;
85 BinaryImplicationGraph* implication_graph_;
86
87 // To detect literal x that must be true because b => x and not(b) => x.
88 // When probing b, we add all propagated literal to propagated, and when
89 // probing not(b) we check if any are already there.
91
92 // Modifications found during probing.
93 std::vector<Literal> to_fix_at_true_;
94 std::vector<IntegerLiteral> new_integer_bounds_;
95 std::vector<std::pair<Literal, Literal>> new_binary_clauses_;
96
97 // Probing statistics.
98 int num_new_holes_ = 0;
99 int num_new_binary_ = 0;
100 int num_new_integer_bounds_ = 0;
101
102 // Logger.
103 SolverLogger* logger_;
104};
105
106// Try to randomly tweak the search and stop at the first conflict each time.
107// This can sometimes find feasible solution, but more importantly, it is a form
108// of probing that can sometimes find small and interesting conflicts or fix
109// variables. This seems to work well on the SAT14/app/rook-* problems and
110// do fix more variables if run before probing.
111//
112// If a feasible SAT solution is found (i.e. all Boolean assigned), then this
113// abort and leave the solver with the full solution assigned.
114//
115// Returns false iff the problem is UNSAT.
116bool LookForTrivialSatSolution(double deterministic_time_limit, Model* model);
117
118// Options for the FailedLiteralProbing() code below.
119//
120// A good reference for the algorithms involved here is the paper "Revisiting
121// Hyper Binary Resolution" Marijn J. H. Heule, Matti Jarvisalo, Armin Biere,
122// http://www.cs.utexas.edu/~marijn/cpaior2013.pdf
124 // The probing will consume all this deterministic time or stop if nothing
125 // else can be deduced and everything has been probed until fix-point. The
126 // fix point depend on the extract_binay_clauses option:
127 // - If false, we will just stop when no more failed literal can be found.
128 // - If true, we will do more work and stop when all failed literal have been
129 // found and all hyper binary resolution have been performed.
130 //
131 // TODO(user): We can also provide a middle ground and probe all failed
132 // literal but do not extract all binary clauses.
133 //
134 // Note that the fix-point is unique, modulo the equivalent literal detection
135 // we do. And if we add binary clauses, modulo the transitive reduction of the
136 // binary implication graph.
137 //
138 // To be fast, we only use the binary clauses in the binary implication graph
139 // for the equivalence detection. So the power of the equivalence detection
140 // changes if the extract_binay_clauses option is true or not.
141 //
142 // TODO(user): The fix point is not yet reached since we don't currently
143 // simplify non-binary clauses with these equivalence, but we will.
145
146 // This is also called hyper binary resolution. Basically, we make sure that
147 // the binary implication graph is augmented with all the implication of the
148 // form a => b that can be derived by fixing 'a' at level zero and doing a
149 // propagation using all constraints. Note that we only add clauses that
150 // cannot be derived by the current implication graph.
151 //
152 // With these extra clause the power of the equivalence literal detection
153 // using only the binary implication graph with increase. Note that it is
154 // possible to do exactly the same thing without adding these binary clause
155 // first. This is what is done by yet another probing algorithm (currently in
156 // simplification.cc).
157 //
158 // TODO(user): Note that adding binary clause before/during the SAT presolve
159 // is currently not always a good idea. This is because we don't simplify the
160 // other clause as much as we could. Also, there can be up to a quadratic
161 // number of clauses added this way, which might slow down things a lot. But
162 // then because of the deterministic limit, we usually cannot add too much
163 // clauses, even for huge problems, since we will reach the limit before that.
165
166 // Use a version of the "Tree look" algorithm as explained in the paper above.
167 // This is usually faster and more efficient. Note that when extracting binary
168 // clauses it might currently produce more "redundant" one in the sense that a
169 // transitive reduction of the binary implication graph after all hyper binary
170 // resolution have been performed may need to do more work.
171 bool use_tree_look = true;
172
173 // There is two sligthly different implementation of the tree-look algo.
174 //
175 // TODO(user): Decide which one is better, currently the difference seems
176 // small but the queue seems slightly faster.
177 bool use_queue = true;
178
179 // If we detect as we probe that a new binary clause subsumes one of the
180 // non-binary clause, we will replace the long clause by the binary one. This
181 // is orthogonal to the extract_binary_clauses parameters which will add all
182 // binary clauses but not neceassirly check for subsumption.
184
185 // We assume this is also true if --v 1 is activated.
186 bool log_info = false;
187
188 std::string ToString() const {
189 return absl::StrCat("deterministic_limit: ", deterministic_limit,
190 " extract_binary_clauses: ", extract_binary_clauses,
191 " use_tree_look: ", use_tree_look,
192 " use_queue: ", use_queue);
193 }
194};
195
196// Similar to ProbeBooleanVariables() but different :-)
197//
198// First, this do not consider integer variable. It doesn't do any disjunctive
199// reasoning (i.e. changing the domain of an integer variable by intersecting
200// it with the union of what happen when x is fixed and not(x) is fixed).
201//
202// However this should be more efficient and just work better for pure Boolean
203// problems. On integer problems, we might also want to run this one first,
204// and then do just one quick pass of ProbeBooleanVariables().
205//
206// Note that this by itself just do one "round", look at the code in the
207// Inprocessing class that call this interleaved with other reductions until a
208// fix point is reached.
209//
210// This can fix a lot of literals via failed literal detection, that is when
211// we detect that x => not(x) via propagation after taking x as a decision. It
212// also use the strongly connected component algorithm to detect equivalent
213// literals.
214//
215// It will add any detected binary clause (via hyper binary resolution) to
216// the implication graph. See the option comments for more details.
217bool FailedLiteralProbingRound(ProbingOptions options, Model* model);
218
219} // namespace sat
220} // namespace operations_research
221
222#endif // OR_TOOLS_SAT_PROBING_H_
A simple class to enforce both an elapsed time limit and a deterministic time limit in the same threa...
Definition: time_limit.h:105
Class that owns everything related to a particular optimization model.
Definition: sat/model.h:38
bool ProbeOneVariable(BooleanVariable b)
Definition: probing.cc:181
bool ProbeBooleanVariables(double deterministic_time_limit)
Definition: probing.cc:44
int64_t b
GRBmodel * model
bool LookForTrivialSatSolution(double deterministic_time_limit, Model *model)
Definition: probing.cc:276
bool FailedLiteralProbingRound(ProbingOptions options, Model *model)
Definition: probing.cc:354
Collection of objects used to extend the Constraint Solver library.