Begin : t 0 iteration 0 (iteration ) m n ; Loop: ; for i 0 to n-1 do for k 1 to m do j; k j; j k end for tt1 end for L t t+1 0 iteration iteration+1

ration iteration+1 iteration< goto Loop End TSP C++ /* >->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->-> AntColonySystemTSP.cc, Heiko Stamer Ant Colony System (ACS) for the Traveling Salesman Problem (TSP) [The Ant System: Optimization by a colony of cooperating agents] by M. Dorigo, V. Maniezzo, A. Colorni IEEE Transactions on Systems, Man and Cybernetics - Part B, Vol.26-1 1996 [Ant Colony System: A Cooperative Learning Approach to the TSP] by M. Dorigo and L. M. Gambardella IEEE Transactions on Evolutionary Computation, Vol. 1, No. 1, 1997

http://stinfwww.informatik.uni-leipzig.de/~mai97ixb >->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->->-> Copyright (C) 2001 - until_the_end_of_the_ants This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ #include #include #include #include #include #include #define N 70 double C[N][2] = { {64 , 96} , {80 , 39} , {69 , 23} , {72 , 42} , {48 , 67} , {58 , 43} , {81 , 34} , {79 , 17} , {30 , 23} , {42 , 67} , {7 , 76} , {29 , 51} , {78 , 92} , {64 , 8} , {95 , 57} , {57 , 91} , {40 , 35} , {68 , 40} , {92 , 34} , {62 , 1} , {28 , 43} , {76 , 73} , {67 , 88} , {93 , 54} , {6 , 8} , {87 , 18} , {30 , 9} , {77 , 13} , {78 , 94} , {55 , 3} , {82 , 88} , {73 , 28} , {20 , 55} , {27 , 43} , {95 , 86} , {67 , 99} , {48 , 83} , {75 , 81} , {8 , 19} , {20 , 18} , {54 , 38} , {63 , 36} , {44 , 33} , {52 , 18} , {12 , 13} , {25 , 5} , {58 , 85} , {5 , 67} , {90 , 9} , {41 , 76} , {25 , 76} , {37 , 64} , {56 , 63} , {10 , 55} , {98 , 7} , {16 , 74} , {89 , 60} , {48 , 82} , {81 , 76} , {29 , 60} , {17 , 22} , {5 , 45} , {79 , 70} , {9 , 100} , {17 , 82} , {74 , 67} , {10 , 68} , {48 , 19} , {83 , 86} , {84 , 94} }; typedef int Tour[N][2]; typedef double doubleMatrix[N][N]; doubleMatrix D;

double dist(int i, int j) { return sqrt(pow((C[i][0]-C[j][0]), 2.0) + pow((C[i][1]-C[j][1]), 2.0)); } void calc_dist() { for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) D[i][j] = dist(i, j); } double max_dist() { double max_dist = 0.0; for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) if (dist(i, j) > max_dist) max_dist = dist(i, j); return max_dist; } double calc_length(Tour tour) { double l = 0.0; for (int n = 0; n < N; n++) { int i = tour[n][0]; int j = tour[n][1]; l += D[i][j]; } return (l); } void print_tour(Tour tour) { for (int n = 0; n < N; n++) printf("( %d , %d ) ", tour[n][0], tour[n][1]); printf("\n"); } int sum_sequence(int array[], int count) {

int sum = 0; for (int i = 0; i < count; i++) sum += array[i]; return (sum); } /******************************************************************************/ class Ant { protected: int START_CITY, CURRENT_CITY; int ALLOWED[N]; Tour CURRENT_TOUR; int CURRENT_TOUR_INDEX; public: inline Ant(int start_city) { START_CITY = start_city; } inline void moveTo(int to_city) { ALLOWED[to_city] = 0; CURRENT_TOUR[CURRENT_TOUR_INDEX][0] = CURRENT_CITY; CURRENT_TOUR[CURRENT_TOUR_INDEX][1] = to_city; CURRENT_TOUR_INDEX++; CURRENT_CITY = to_city; } }; class NNAnt : Ant { public: inline NNAnt(int start_city): Ant(start_city) { }; inline int choose() { double best_length = (double)N * max_dist(); int best_choose = -1; for (int j = 0; j < N; j++) { if ((ALLOWED[j] == 1) && (D[CURRENT_CITY][j] < best_length)) { best_choose = j;

best_length = D[CURRENT_CITY][j]; } } return best_choose; } inline Tour *search() { CURRENT_CITY = START_CITY; CURRENT_TOUR_INDEX = 0; for (int i = 0; i < N; i++) ALLOWED[i] = 1; ALLOWED[CURRENT_CITY] = 0; while (sum_sequence(ALLOWED, N) > 0) moveTo(choose()); ALLOWED[START_CITY] = 1; moveTo(START_CITY); return &CURRENT_TOUR; } }; class AntColonySystem; class ACSAnt : Ant { private: AntColonySystem *ACS; public: ACSAnt(AntColonySystem *acs, int start_city): Ant(start_city) { ACS = acs; } inline int choose(); inline Tour *search(); }; class AntColonySystem { private: double ALPHA, BETA, RHO, TAU0; doubleMatrix TAU, dTAU; static const int M = 420; ACSAnt *ANTS[M]; public: double Q0;

AntColonySystem(double alpha, double beta, double rho, double q0); inline double calc_tau0(); inline void init_tau_by_value(double value); inline void init_tau_by_matrix(doubleMatrix matrix); inline void init_uniform(); inline void init_random(); inline void init_randomMOAPC(); inline double ETA(int i, int j); inline double transition(int i, int j); inline double sum_transition(int i, int allowed[]); inline void local_update_rule(int i, int j); inline void clear_global_update(); inline void add_global_update(Tour tour, double length); inline void global_update_rule(); inline doubleMatrix *get_tau(); inline Tour *search(int T); }; inline int ACSAnt::choose() { double q = rand() / (double)RAND_MAX; if (q <= ACS->Q0) { double best_value = -1.0; int best_choose = -1; for (int j = 0; j < N; j++) { if ((ALLOWED[j] == 1) && (ACS->transition(CURRENT_CITY, j) > best_value)) { best_choose = j; best_value = ACS->transition(CURRENT_CITY, j); } } return best_choose; } double sum = ACS->sum_transition(CURRENT_CITY, ALLOWED); double p = rand() / (double)RAND_MAX; double p_j = 0.0; for (int j = 0; j < N; j++) {

if (ALLOWED[j] == 1) p_j += ACS->transition(CURRENT_CITY, j) / sum; if ((p < p_j) && (ALLOWED[j] == 1)) return j; } return -1; } inline Tour *ACSAnt::search() { CURRENT_CITY = START_CITY; CURRENT_TOUR_INDEX = 0; for (int i = 0; i < N; i++) ALLOWED[i] = 1; ALLOWED[CURRENT_CITY] = 0; while (sum_sequence(ALLOWED, N) > 0) { int LAST_CITY = CURRENT_CITY; moveTo(choose()); ACS->local_update_rule(LAST_CITY, CURRENT_CITY); } ALLOWED[START_CITY] = 1; ACS->local_update_rule(CURRENT_CITY, START_CITY); moveTo(START_CITY); return &CURRENT_TOUR; } /******************************************************************************/ AntColonySystem::AntColonySystem(double alpha, double beta, double rho, double q0) { ALPHA = alpha; BETA = beta; RHO = rho; Q0 = q0; } inline double AntColonySystem::calc_tau0() { double best_length = (double)N * max_dist(); for (int n = 0; n < N; n++) { NNAnt *nnANT = new NNAnt(n); Tour tour;

tour = *(nnANT->search()); double tour_length = calc_length(tour); if (tour_length < best_length) best_length = tour_length; delete nnANT; } return 1.0 / ((double)N * best_length); } inline void AntColonySystem::init_tau_by_value(double value) { TAU0 = value; for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) TAU[i][j] = TAU0; } inline void AntColonySystem::init_tau_by_matrix(doubleMatrix matrix) { for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) TAU[i][j] = matrix[i][j]; } inline void AntColonySystem::init_uniform() { // uniformly distributed for (int k = 0; k < M; k++) ANTS[k] = new ACSAnt(this, (k % N)); } inline void AntColonySystem::init_random() { // randomly distributed for (int k = 0; k < M; k++) ANTS[k] = new ACSAnt(this, (int)((double)N * (rand() / (double)RAND_MAX))); } inline void AntColonySystem::init_randomMOAPC() { // randomly distributed with MOAPC (most one ant per city) bool MOAPCarray[N]; assert(M <= N);

for (int n = 0; n < N; n++) MOAPCarray[n] = false; for (int k = 0; k < M; k++) { int c; do { c = (int)((double)N * (rand() / (double)RAND_MAX)); } while (MOAPCarray[c]); MOAPCarray[c] = true; ANTS[k] = new ACSAnt(this, c); } } inline double AntColonySystem::ETA(int i, int j) { return ( 1.0 / D[i][j] ); } inline double AntColonySystem::transition(int i, int j) { if (i != j) return ( TAU[i][j] * pow( ETA(i, j), BETA ) ); else return(0.0); } inline double AntColonySystem::sum_transition(int i, int allowed[]) { double sum = 0.0; for (int j = 0; j < N; j++) sum += ((double)allowed[j] * transition(i, j)); return (sum); } inline void AntColonySystem::local_update_rule(int i, int j) { TAU[i][j] = (1.0 - RHO) * TAU[i][j] + RHO * TAU0; // symmetric TSP TAU[j][i] = TAU[i][j]; }

inline void AntColonySystem::clear_global_update() { for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) dTAU[i][j] = 0.0; }

inline void AntColonySystem::add_global_update(Tour tour, double length) { for (int n = 0; n < N; n++) { int i = tour[n][0]; int j = tour[n][1]; dTAU[i][j] += (1.0 / length); // symmetric TSP dTAU[j][i] += (1.0 / length); } } inline void AntColonySystem::global_update_rule() { for (int i = 0; i < N; i++) for (int j = 0; j < N; j++) TAU[i][j] = (1.0 - ALPHA) * TAU[i][j] + ALPHA * dTAU[i][j]; } inline doubleMatrix *AntColonySystem::get_tau() { return &TAU; } inline Tour *AntColonySystem::search(int T) { Tour best_tour, tour; double best_length = (double)N * max_dist(), tour_length; clear_global_update(); // do T iterations of ACS algorithm int t; for (t = 0; t < T; t++) { for (int k = 0; k < M; k++)

{ tour = *(ANTS[k]->search()); tour_length = calc_length(tour); if (tour_length < best_length) { best_tour = tour; best_length = tour_length; clear_global_update(); add_global_update(tour, tour_length); //printf("[%d / %d]: %lf \n", t, T, tour_length); } } global_update_rule(); } //printf("[%d/%d] best tour (length = %f):\n", t, T, best_length); //print_tour(best_tour); //printf("[%d/%d] iterations done\n", t, T); printf("%f\n", best_length); return (&best_tour); } int main(int argc, char* argv[]) { // PRNG initalisieren time_t timer; time(&timer); pid_t pid = getpid() + getppid(); unsigned long seed = (timer * pid); if (seed == 0) { time(&timer); seed = 7 * timer * pid; if (seed == 0) seed = pid; else seed = seed % 56000; } else seed = seed % 56000; srand((unsigned int)seed); // EUC2D calc_dist(); // Ant Colony System AntColonySystem *acs = new AntColonySystem(0.1, 2.0, 0.1, 0.9); double tau0 = acs->calc_tau0(); acs->init_tau_by_value(tau0);

acs->init_uniform(); acs->search(1000); return(0); }