| 1 | /* |
| 2 | * schedulingExample.cpp |
| 3 | * @brief hard scheduling example |
| 4 | * @date March 25, 2011 |
| 5 | * @author Frank Dellaert |
| 6 | */ |
| 7 | |
| 8 | #define ENABLE_TIMING |
| 9 | #define ADD_NO_CACHING |
| 10 | #define ADD_NO_PRUNING |
| 11 | #include <gtsam_unstable/discrete/Scheduler.h> |
| 12 | #include <gtsam/base/debug.h> |
| 13 | #include <gtsam/base/timing.h> |
| 14 | |
| 15 | #include <algorithm> |
| 16 | |
| 17 | using namespace std; |
| 18 | using namespace gtsam; |
| 19 | |
| 20 | size_t NRSTUDENTS = 12; |
| 21 | |
| 22 | bool NonZero(size_t i) { |
| 23 | return i > 0; |
| 24 | } |
| 25 | |
| 26 | /* ************************************************************************* */ |
| 27 | void addStudent(Scheduler& s, size_t i) { |
| 28 | switch (i) { |
| 29 | case 0: |
| 30 | s.addStudent(studentName: "Young, Carol" , area1: "Controls" , area2: "Autonomy" , area3: "Mechanics" , advisor: "Fumin Zhang" ); |
| 31 | break; |
| 32 | case 1: |
| 33 | s.addStudent(studentName: "Erdogan, Can" , area1: "Controls" , area2: "AI" , area3: "Perception" , advisor: "Mike Stilman" ); |
| 34 | break; |
| 35 | case 2: |
| 36 | s.addStudent(studentName: "Arslan, Oktay" , area1: "Controls" , area2: "AI" , area3: "Mechanics" , advisor: "Panos Tsiotras" ); |
| 37 | break; |
| 38 | case 3: |
| 39 | s.addStudent(studentName: "Bhattacharjee, Tapomayukh" , area1: "Controls" , area2: "AI" , area3: "Mechanics" , advisor: "Charlie Kemp" ); |
| 40 | break; |
| 41 | case 4: |
| 42 | s.addStudent(studentName: "Grey, Michael" , area1: "Controls" , area2: "AI" , area3: "Mechanics" , advisor: "Wayne Book" ); |
| 43 | break; |
| 44 | case 5: |
| 45 | s.addStudent(studentName: "O'Flaherty, Rowland" , area1: "Controls" , area2: "AI" , area3: "Mechanics" , advisor: "Magnus Egerstedt" ); |
| 46 | break; |
| 47 | case 6: |
| 48 | s.addStudent(studentName: "Pickem, Daniel" , area1: "Controls" , area2: "AI" , area3: "Mechanics" , advisor: "Jeff Shamma" ); |
| 49 | break; |
| 50 | case 7: |
| 51 | s.addStudent(studentName: "Lee, Kimoon" , area1: "Controls" , area2: "Autonomy" , area3: "Mechanics" , advisor: "Henrik Christensen" ); |
| 52 | break; |
| 53 | case 8: |
| 54 | s.addStudent(studentName: "Melim, Andrew Lyon" , area1: "Controls" , area2: "AI" , area3: "Perception" , advisor: "Frank Dellaert" ); |
| 55 | break; |
| 56 | case 9: |
| 57 | s.addStudent(studentName: "Jensen, David" , area1: "Controls" , area2: "Autonomy" , area3: "HRI" , advisor: "Andrea Thomaz" ); |
| 58 | break; |
| 59 | case 10: |
| 60 | s.addStudent(studentName: "Nisbett, Jared" , area1: "Controls" , area2: "Perception" , area3: "Mechanics" , advisor: "Magnus Egerstedt" ); |
| 61 | break; |
| 62 | case 11: |
| 63 | s.addStudent(studentName: "Pan, Yunpeng" , area1: "Controls" , area2: "Perception" , area3: "Mechanics" , advisor: "Wayne Book" ); |
| 64 | break; |
| 65 | // case 12: |
| 66 | // s.addStudent("Grice, Phillip", "Controls", "None", "None", "Wayne Book"); |
| 67 | // break; |
| 68 | // case 13: |
| 69 | // s.addStudent("Robinette, Paul", "Controls", "None", "None", "Ayanna Howard"); |
| 70 | // break; |
| 71 | // case 14: |
| 72 | // s.addStudent("Huaman, Ana", "Autonomy", "None", "None", "Mike Stilman"); |
| 73 | // break; |
| 74 | } |
| 75 | } |
| 76 | |
| 77 | /* ************************************************************************* */ |
| 78 | Scheduler largeExample(size_t nrStudents = NRSTUDENTS, bool addStudents=true) { |
| 79 | string path("../../../gtsam_unstable/discrete/examples/" ); |
| 80 | Scheduler s(nrStudents, path + "Doodle2013.csv" ); |
| 81 | |
| 82 | s.addArea(facultyName: "Harvey Lipkin" , areaName: "Mechanics" ); |
| 83 | s.addArea(facultyName: "Jun Ueda" , areaName: "Mechanics" ); |
| 84 | s.addArea(facultyName: "Mike Stilman" , areaName: "Mechanics" ); |
| 85 | // s.addArea("Frank Dellaert", "Mechanics"); |
| 86 | s.addArea(facultyName: "Wayne Book" , areaName: "Mechanics" ); |
| 87 | // s.addArea("Charlie Kemp", "Mechanics"); |
| 88 | |
| 89 | s.addArea(facultyName: "Patricio Vela" , areaName: "Controls" ); |
| 90 | s.addArea(facultyName: "Magnus Egerstedt" , areaName: "Controls" ); |
| 91 | s.addArea(facultyName: "Jun Ueda" , areaName: "Controls" ); |
| 92 | s.addArea(facultyName: "Panos Tsiotras" , areaName: "Controls" ); |
| 93 | s.addArea(facultyName: "Fumin Zhang" , areaName: "Controls" ); |
| 94 | s.addArea(facultyName: "Ayanna Howard" , areaName: "Controls" ); |
| 95 | s.addArea(facultyName: "Jeff Shamma" , areaName: "Controls" ); |
| 96 | |
| 97 | s.addArea(facultyName: "Frank Dellaert" , areaName: "Perception" ); |
| 98 | s.addArea(facultyName: "Henrik Christensen" , areaName: "Perception" ); |
| 99 | |
| 100 | s.addArea(facultyName: "Mike Stilman" , areaName: "AI" ); |
| 101 | // s.addArea("Henrik Christensen", "AI"); |
| 102 | // s.addArea("Ayanna Howard", "AI"); |
| 103 | s.addArea(facultyName: "Charles Isbell" , areaName: "AI" ); |
| 104 | // s.addArea("Tucker Balch", "AI"); |
| 105 | s.addArea(facultyName: "Andrea Thomaz" , areaName: "AI" ); |
| 106 | |
| 107 | s.addArea(facultyName: "Ayanna Howard" , areaName: "Autonomy" ); |
| 108 | s.addArea(facultyName: "Charlie Kemp" , areaName: "Autonomy" ); |
| 109 | |
| 110 | // s.addArea("Andrea Thomaz", "HRI"); |
| 111 | s.addArea(facultyName: "Karen Feigh" , areaName: "HRI" ); |
| 112 | // s.addArea("Charlie Kemp", "HRI"); |
| 113 | |
| 114 | // add students |
| 115 | if (addStudents) |
| 116 | for (size_t i = 0; i < nrStudents; i++) |
| 117 | addStudent(s, i); |
| 118 | |
| 119 | return s; |
| 120 | } |
| 121 | |
| 122 | /* ************************************************************************* */ |
| 123 | void runLargeExample() { |
| 124 | |
| 125 | Scheduler scheduler = largeExample(); |
| 126 | scheduler.print(); |
| 127 | |
| 128 | // BUILD THE GRAPH ! |
| 129 | size_t addMutex = 3; |
| 130 | SETDEBUG("Scheduler::buildGraph" , true); |
| 131 | scheduler.buildGraph(mutexBound: addMutex); |
| 132 | |
| 133 | // Do brute force product and output that to file |
| 134 | if (scheduler.nrStudents() == 1) { // otherwise too slow |
| 135 | DecisionTreeFactor product = |
| 136 | *std::dynamic_pointer_cast<DecisionTreeFactor>(r: scheduler.product()); |
| 137 | product.dot(name: "scheduling-large" , keyFormatter: DefaultKeyFormatter, showZero: false); |
| 138 | } |
| 139 | |
| 140 | // Do exact inference |
| 141 | // SETDEBUG("timing-verbose", true); |
| 142 | SETDEBUG("DiscreteConditional::DiscreteConditional" , true); |
| 143 | //#define SAMPLE |
| 144 | #ifdef SAMPLE |
| 145 | gttic(large); |
| 146 | DiscreteBayesNet::shared_ptr chordal = scheduler.eliminate(); |
| 147 | gttoc(large); |
| 148 | tictoc_finishedIteration(); |
| 149 | tictoc_print(); |
| 150 | for (size_t i=0;i<100;i++) { |
| 151 | auto assignment = sample(*chordal); |
| 152 | vector<size_t> stats(scheduler.nrFaculty()); |
| 153 | scheduler.accumulateStats(assignment, stats); |
| 154 | size_t max = *max_element(stats.begin(), stats.end()); |
| 155 | size_t min = *min_element(stats.begin(), stats.end()); |
| 156 | size_t nz = count_if(stats.begin(), stats.end(), NonZero); |
| 157 | // cout << min << ", " << max << ", " << nz << endl; |
| 158 | if (nz >= 13 && min >=1 && max <= 4) { |
| 159 | cout << "======================================================\n" ; |
| 160 | scheduler.printAssignment(assignment); |
| 161 | } |
| 162 | } |
| 163 | #else |
| 164 | gttic(large); |
| 165 | auto MPE = scheduler.optimize(); |
| 166 | gttoc(large); |
| 167 | tictoc_finishedIteration(); |
| 168 | tictoc_print(); |
| 169 | scheduler.printAssignment(assignment: MPE); |
| 170 | #endif |
| 171 | } |
| 172 | |
| 173 | /* ************************************************************************* */ |
| 174 | // Solve a series of relaxed problems for maximum flexibility solution |
| 175 | void solveStaged(size_t addMutex = 2) { |
| 176 | |
| 177 | bool debug = false; |
| 178 | |
| 179 | // super-hack! just count... |
| 180 | SETDEBUG("DiscreteConditional::COUNT" , true); |
| 181 | SETDEBUG("DiscreteConditional::DiscreteConditional" , debug); // progress |
| 182 | |
| 183 | // make a vector with slot availability, initially all 1 |
| 184 | // Reads file to get count :-) |
| 185 | vector<double> slotsAvailable(largeExample(nrStudents: 0).nrTimeSlots(), 1.0); |
| 186 | |
| 187 | // now, find optimal value for each student, using relaxed mutex constraints |
| 188 | for (size_t s = 0; s < NRSTUDENTS; s++) { |
| 189 | // add all students first time, then drop last one second time, etc... |
| 190 | Scheduler scheduler = largeExample(nrStudents: NRSTUDENTS - s); |
| 191 | // scheduler.print(str(boost::format("Scheduler %d") % (NRSTUDENTS-s))); |
| 192 | |
| 193 | // only allow slots not yet taken |
| 194 | scheduler.setSlotsAvailable(slotsAvailable); |
| 195 | |
| 196 | // BUILD THE GRAPH ! |
| 197 | scheduler.buildGraph(mutexBound: addMutex); |
| 198 | |
| 199 | // Do EXACT INFERENCE |
| 200 | gttic_(eliminate); |
| 201 | DiscreteBayesNet::shared_ptr chordal = scheduler.eliminate(); |
| 202 | gttoc_(eliminate); |
| 203 | |
| 204 | // find root node |
| 205 | DiscreteConditional::shared_ptr root = chordal->back(); |
| 206 | if (debug) |
| 207 | root->print(s: "" /*scheduler.studentName(s)*/); |
| 208 | |
| 209 | // solve root node only |
| 210 | size_t bestSlot = root->argmax(); |
| 211 | |
| 212 | // get corresponding count |
| 213 | DiscreteKey dkey = scheduler.studentKey(i: NRSTUDENTS - 1 - s); |
| 214 | DiscreteValues values; |
| 215 | values[dkey.first] = bestSlot; |
| 216 | double count = (*root)(values); |
| 217 | |
| 218 | // remove this slot from consideration |
| 219 | slotsAvailable[bestSlot] = 0.0; |
| 220 | cout << scheduler.studentName(i: NRSTUDENTS - 1 - s) << " = " << scheduler.slotName(s: bestSlot) << " (" << bestSlot |
| 221 | << "), count = " << count << endl; |
| 222 | } |
| 223 | tictoc_print_(); |
| 224 | } |
| 225 | |
| 226 | /* ************************************************************************* */ |
| 227 | // Sample from solution found above and evaluate cost function |
| 228 | DiscreteBayesNet::shared_ptr createSampler(size_t i, |
| 229 | size_t slot, vector<Scheduler>& schedulers) { |
| 230 | Scheduler scheduler = largeExample(nrStudents: 1,addStudents: false); |
| 231 | addStudent(s&: scheduler, i); |
| 232 | cout << " creating sampler for " << scheduler.studentName(i: 0) << endl; |
| 233 | SETDEBUG("Scheduler::buildGraph" , false); |
| 234 | // scheduler.print(); |
| 235 | scheduler.addStudentSpecificConstraints(i: 0, slot); |
| 236 | DiscreteBayesNet::shared_ptr chordal = scheduler.eliminate(); |
| 237 | schedulers.push_back(x: scheduler); |
| 238 | return chordal; |
| 239 | } |
| 240 | |
| 241 | void sampleSolutions() { |
| 242 | |
| 243 | size_t nrFaculty = 17; // Change to correct number ! |
| 244 | |
| 245 | vector<Scheduler> schedulers; |
| 246 | vector<DiscreteBayesNet::shared_ptr> samplers(NRSTUDENTS); |
| 247 | |
| 248 | // Given the time-slots, we can create NRSTUDENTS independent samplers |
| 249 | vector<size_t> slots{12,11,13, 21,16,1, 3,2,6, 7,22,4}; // given slots |
| 250 | for (size_t i = 0; i < NRSTUDENTS; i++) |
| 251 | samplers[i] = createSampler(i, slot: slots[i], schedulers); |
| 252 | |
| 253 | // now, sample schedules |
| 254 | for (size_t n = 0; n < 10000; n++) { |
| 255 | vector<size_t> stats(nrFaculty, 0); |
| 256 | vector<DiscreteValues> samples; |
| 257 | for (size_t i = 0; i < NRSTUDENTS; i++) { |
| 258 | samples.push_back(x: samplers[i]->sample()); |
| 259 | schedulers[i].accumulateStats(assignment: samples[i], stats); |
| 260 | } |
| 261 | size_t max = *max_element(first: stats.begin(), last: stats.end()); |
| 262 | size_t min = *min_element(first: stats.begin(), last: stats.end()); |
| 263 | size_t nz = count_if(first: stats.begin(), last: stats.end(), pred: NonZero); |
| 264 | if (nz >= 16 && max <= 3) { |
| 265 | cout << "Sampled schedule " << n + 1 << ", min = " << min << ", nz = " << nz << ", max = " << max << endl; |
| 266 | for (size_t i = 0; i < NRSTUDENTS; i++) { |
| 267 | cout << schedulers[i].studentName(i: 0) << " : " << schedulers[i].slotName( |
| 268 | s: slots[i]) << endl; |
| 269 | schedulers[i].printSpecial(assignment: samples[i]); |
| 270 | } |
| 271 | } |
| 272 | } |
| 273 | } |
| 274 | |
| 275 | /* ************************************************************************* */ |
| 276 | int main() { |
| 277 | // runLargeExample(); |
| 278 | // solveStaged(3); |
| 279 | sampleSolutions(); |
| 280 | return 0; |
| 281 | } |
| 282 | /* ************************************************************************* */ |
| 283 | |
| 284 | |