def _improve_with_3opt_star(self, problem, solution, strategy):
if len(problem) == 5:
N, pts, d, D, C = problem
L = None
else:
N, pts, d, D, C, L = problem
sol = list(solution)
if not PRINT_ONLY_FINAL_RESULT:
print("\nin", sol)
total_t = 0
while True:
start_t = time()
out_sol = do_3optstar_move(sol, D, d, C, L, strategy)
elapsed_t = time() - start_t
total_t += elapsed_t
if not PRINT_ONLY_FINAL_RESULT:
print("elapsed %.2f s" % elapsed_t)
print("out (%s)\n" % _strategy_to_str(strategy))
if out_sol[1] == None:
print("no more improvements found")
if PRINT_ONLY_FINAL_RESULT:
print("final (%s)" % _strategy_to_str(strategy), sol,
calculate_objective(sol, D))
print("total elapsed %.2f s" % total_t)
break
if not PRINT_ONLY_FINAL_RESULT:
print(out_sol, calculate_objective(out_sol[0], D))
sol = out_sol[0]
self.assertTrue(all(check_solution_feasibility(sol, D, d, C, L)),
"must be feasible")
def _random_init_lr3opt(pts, D, d, C, L, st, times):
best_sol = None
best_f = float('inf')
for t in range(times):
sol = lr3opt_init(D,
d,
C,
L,
initialization_algorithm=_init_with_random)
sol_f = calculate_objective(sol, D)
if sol_f < best_f:
best_sol = None
best_f = None
return best_sol
def setUp(self):
N = TestRandomStressOn3OptStarSolutionOperator.problem_size
problem = generate_CVRP(N, 50, 10, 5)
aN,pts,d,D,C = (problem.size, problem.coordinate_points,
problem.customer_demands, problem.distance_matrix,
problem.capacity_constraint)
pts = None
d = [int(dv) for dv in d]
D = D.astype(int)
problem = aN,pts,d,D,C
self.naive_sol = routes2sol( [[n] for n in range(1,aN+1)] )
self.nn_sol = nearest_neighbor_init(D, d, C)
self.L = max( calculate_objective(r,D) for r in sol2routes(self.nn_sol) )
self.N = aN
self.problem = (aN,pts,d,D,C)
def lr3opt_init(D,
d,
C,
L,
initial_lambda1_C=None,
initial_lambda1_L=None,
initialization_algorithm=_init_with_tsp,
postoptimize_with_3optstar=True,
max_concecutive_lamba_incs=None):
""" An implementation of the Stewart & Golden [1]_ 3-opt* heuristic
with Lagrangean relaxation.
The algorithm starts from a solution that can be either feasible or
infeasible and uses local search to move towards better and feasible
solutions. More specifically, it works by replacing the constraint checks
of the 3-opt* with a penalty that depends on how much the constraint was
violated. The 3-opt* that operates on the entire solution, that is, checks
for both intra and inter route moves on one pass, was used. The penalties
are iteratively doubled as the local search progresses and it is assumed
that this eventually forces the solutions to feasible region.
.. [1] Stewart, W. R. and Golden, B. L. (1984). A lagrangean relaxation
heuristic for vehicle routing. European Journal of Operational
Research, 15(1):84–88.
Parameters
----------
D : numpy.ndarray
is the full 2D distance matrix.
d : list
is a list of demands. d[0] should be 0.0 as it is the depot.
C : float
is the capacity constraint limit for the identical vehicles.
L : float
is the optional constraint for the maximum route length/duration/cost.
initial_lambda1_C : float
is the initial Langrange multiplier value for the capacity constraint C.
If left empty (None) the formula ``l1_C=average(d)/(20*max(D))`` is used.
The alternative value suggested by Stewart & Golden (1984) was 0.05.
initial_lambda1_L : float
is the initial Langrange multiplier value for the maximum route cost/
duration/length constraint. If left empty (None) the formula
``l1_L=average(distance to nearest neighbor)/(10*max(D))`` is used.
initialization_algorithm (function): is a function that retuns a TSP or VRP
solution and its objective function value. The default is to use LKH TSP
solution, but the function _init_with_random can be used to replicate the
results of Stewart & Golden (1984) where a random solution is used.
Returns
-------
list
The solution as a list of node indices to visit.
.. todo:: due to how the algorithm works, introducing minimize_K would require
balancing between penalizing constraint violations and penalizing new
routes with an additional multipiler. This was not implemented.
"""
sol = None
try:
## STEP 1: Generate an initial solution
sol, initial_f = initialization_algorithm(D, d, C, L)
max_D = None
lambdas = [initial_lambda1_C, initial_lambda1_L]
if C and lambdas[0] == None:
max_D = _get_max(D, sol)
lambdas[0] = np.average(d) / (20 * max_D)
if L and lambdas[1] == None:
# Stewart & Golden (1984) did not propose an extension for the maximum
# route duration/length/cost constraint, but here we have something
# similar to L than they used for C constraint relaxation.
max_D = _get_max(D, sol) if (max_D is None) else max_D
closest_neighbor_D = D.copy()
np.fill_diagonal(closest_neighbor_D, max_D)
lambdas[1] = np.average(
closest_neighbor_D.min(axis=0)) / (10 * max_D)
if __debug__:
log(
DEBUG,
"Start from initial solution %s (%.2f), and with l1=%.2f, l2=%.2f"
% (sol, calculate_objective(sol, D),
(0 if lambdas[0] is None else lambdas[0]),
(0 if lambdas[1] is None else lambdas[1])))
checker_function = partial(_check_lr3opt_move, lambdas=lambdas)
# STEP 2: Solve the relaxed problem using 3-opt*
c_lambda_incs = 0
while True:
# Make sure there is an empty route (for giving the 3-opt* procedure
# the option of adding vehicles)
while not (sol[-1] == 0 and sol[-2] == 0):
sol += [0]
if __debug__:
log(
DEBUG - 2, "Finding a LR3OPT move for %s (%.2f)" %
(sol, calculate_objective(sol, D)))
new_sol, delta = do_3optstar_move(sol,
D,
d,
C,
L,
strategy=LSOPT.FIRST_ACCEPT,
move_checker=checker_function)
# local optima reached, tighten the relaxation
# TODO: it should not happen that the sol==new_sol. However it happens and as a quickfix check for it.
if delta is None or sol == new_sol:
# return the first feasible solution (note: does not check for covering)
if fast_constraint_check(sol, D, d, C, L):
if __debug__:
log(
DEBUG, "Reached feasible solution %s (%.2f)" %
(sol, calculate_objective(sol, D)))
while postoptimize_with_3optstar:
opt_sol, delta = do_3optstar_move(
sol, D, d, C, L, strategy=LSOPT.FIRST_ACCEPT)
if delta is None:
return normalize_solution(
sol) # remove any [0,0]'s
else:
sol = opt_sol
#print("REMOVEME improved with post-optimization 3-opt*")
log(
DEBUG,
"Found improving 3-opt* move leading to %s (%.2f)"
% (sol, calculate_objective(sol, D)))
return normalize_solution(sol) # remove any [0,0]'s
else:
# STEP 3: Update lambdas
lambda_at_inf = False
if lambdas[0] is not None:
lambdas[0] = lambdas[0] * 2
lambda_at_inf = lambdas[0] == float('inf')
if lambdas[1] is not None:
lambdas[1] = lambdas[1] * 2
lambda_at_inf = lambda_at_inf or lambdas[0] == float(
'inf')
if __debug__:
log(
DEBUG - 1,
"No improving moves left, increasing lambda to l1=%.2f, l2=%.2f"
% ((0 if lambdas[0] is None else lambdas[0]),
(0 if lambdas[1] is None else lambdas[1])))
#print("No improving moves left, increasing lambda to l1=%.2f, l2=%.2f"%
# ((0 if lambdas[0] is None else lambdas[0]),
# (0 if lambdas[1] is None else lambdas[1])))
#TODO: if penalty >> cost, break (stuck on a infeasible region)
# how much bigger can be determined by finding such a
# pathological problem instance?
# safeguard for getting stuck
c_lambda_incs += 1
#print("REMOVEME: c_lambda_incs", c_lambda_incs)
if lambda_at_inf or (
max_concecutive_lamba_incs is not None
and c_lambda_incs > max_concecutive_lamba_incs):
return _force_feasible(sol, D, d, C, L)
else:
if __debug__:
log(
DEBUG,
"Found improving LR3OPT move leading to %s (%.2f)" %
(new_sol, calculate_objective(new_sol, D)))
log(
DEBUG - 2, "However, routes %s remain infeasible." % [
r for r in sol2routes(new_sol)
if not fast_constraint_check(r, D, d, C, L)
])
sol = new_sol
c_lambda_incs = 0
except KeyboardInterrupt: # or SIGINT
# Pass on the current solution forced feasbile by splitting routes
# according to the constraints.
raise KeyboardInterrupt(_force_feasible(sol, D, d, C, L))
return without_empty_routes(sol)
def _init_with_random(D, d, C, L):
customers = list(range(1, len(D)))
shuffle(customers)
random_sol = [0] + customers + [0]
return random_sol, calculate_objective(random_sol, D)
def main(overridden_args=None):
## 1. parse arguments
parser = ArgumentParser(
description=
"Solve some .vrp problems with the algorithms built into VeRyPy.")
parser.add_argument('-l',
dest='list_algorithms',
help="List the available heuristics and quit",
action="store_true")
parser.add_argument(
'-v',
dest='verbosity',
help=
"Set the verbosity level (to completely disable debug output, run this script with 'python -O')",
type=int,
default=-1)
parser.add_argument(
'-a',
dest='active_algorithms',
help=
"Algorithm to apply (argument can be set multiple times to enable multiple algorithms, or one can use 'all' or 'classical')",
action='append')
parser.add_argument(
'-b',
dest='objective',
choices=['c', 'cost', 'K', 'vehicles'],
help="Primary optimization oBjective (default is cost)",
default="cost")
parser.add_argument(
'-m',
dest='minimal_output',
help=
"Overrides the output options and prints only one line CSV report per solved instance",
action="store_true")
parser.add_argument(
'-t',
dest='print_elapsed_time',
help="Print elapsed wall time for each solution attempt",
action="store_true")
parser.add_argument(
'-c',
dest='show_solution_cost',
help="Display solution cost instead of solution length",
action="store_true")
parser.add_argument('-D',
dest='dist_weight_format',
choices=['ROUND', 'EXACT', 'TRUNCATE'],
help="Force distance matrix rounding")
parser.add_argument(
'-1',
dest='use_single_iteration',
help="Force the algorithms to use only single iteration.",
action="store_true")
parser.add_argument(
'--iinfo',
dest='print_instance_info',
help="Print the instance info in the collected results",
action="store_true")
parser.add_argument(
'--routes',
dest='print_route_stat',
help="Print per route statistics of the final solution",
action="store_true")
parser.add_argument('--vrph',
dest='print_vrph_sol',
help="Print the final solution in the VRPH format",
action="store_true")
parser.add_argument(
'--forbid',
dest='forbid_algorithms',
help=
"Forbid applying algorithms (argument can set multiple times to forbid multiple algorithms)",
action='append')
parser.add_argument(
'--simulate',
dest='simulate',
help=
"Do not really invoke algorithms, can be used e.g. to test scripts",
action="store_true")
#TODO: consider adding more LS opts e.g. 2optstart, 3optstart
parser.add_argument(
'--post-optimize',
dest='local_search_operators',
choices=['2opt', '3opt'],
help=
"Do post-optimization with local search operator(s) (can set multiple)",
action='append')
parser.add_argument(
"problem_file",
help=
"a path of a .vrp problem file, a directory containing .vrp files, or a text file of paths to .vrp files",
action='append')
if overridden_args:
app_args = parser.parse_args(overridden_args)
elif "-l" in sys.argv:
print("Select at least one algorithm (with -a) from the list:",
file=sys.stderr)
print(_build_algorithm_help())
sys.exit(1)
elif len(sys.argv) == 1:
print("Give at least one .vrp file and use -h to get help.",
file=sys.stderr)
sys.exit(1)
else:
app_args = parser.parse_args()
# some further argument validation
if not app_args.active_algorithms or app_args.list_algorithms:
print("Select at least one algorithm (with -a) from the list:",
file=sys.stderr)
print(_build_algorithm_help())
exit()
if len(app_args.problem_file) == 0:
print("Provide at least one .vrp file to solve", file=sys.stderr)
exit()
# get .vrp file list
files_to_solve = shared_cli.get_a_problem_file_list(app_args.problem_file)
# get algorithms
algos = get_algorithms(app_args.active_algorithms)
if app_args.forbid_algorithms:
forbidden_algos = [
algo_name_aliases[algo_name]
for algo_name in app_args.forbid_algorithms
if (algo_name in app_args.forbid_algorithms)
]
algos = [a for a in algos if (a[0] not in forbidden_algos)]
# get primary objective
minimize_K = False
if app_args.objective == 'K' or app_args.objective == 'vehicles':
minimize_K = True
run_single_iteration = False
if app_args.use_single_iteration:
run_single_iteration = True
# get post-optimization local search move operators
ls_ops = []
ls_algo_names = []
if app_args.local_search_operators:
ls_algo_names = app_args.local_search_operators
for ls_op_name in ls_algo_names:
if ls_op_name == "2opt":
ls_ops.append(do_2opt_move)
if ls_op_name == "3opt":
ls_ops.append(do_3opt_move)
# verbosity
if app_args.verbosity >= 0:
shared_cli.set_logger_level(app_args.verbosity)
# minimal header
if app_args.minimal_output:
print("algo;problem;is_feasible;f;K;t")
## 2. solve
results = defaultdict(lambda: defaultdict(float))
instance_data = dict()
interrupted = False
for pfn in files_to_solve:
bn = path.basename(pfn).replace(".vrp",
"").replace(".tsp",
"").replace(".pickle", "")
try:
N, points, dd_points, d, D, C, ewt, K, L, st = pickle.load(
open(pfn, "rb"))
except:
N, points, dd_points, d, D, C, ewt = cvrp_io.read_TSPLIB_CVRP(pfn)
K, L, st = cvrp_io.read_TSBLIB_additional_constraints(pfn)
# We do not have point coodrinates, but we have D!
if points is None:
if dd_points is not None:
points = dd_points
else:
points, ewt = cvrp_ops.generate_missing_coordinates(D)
if app_args.dist_weight_format == "TRUNCATE":
D = np.floor(D)
ewt = "FLOOR_2D"
if app_args.dist_weight_format == "ROUND":
D = np.int(D)
ewt = "EUC_2D"
# Bake service time to D (if needed)
D_c = cvrp_ops.D2D_c(D, st) if st else D
for algo_abbreviation, algo_name, _, algo_f in algos:
if not app_args.minimal_output:
print("Solving %s with %s" % (bn, algo_name))
start_t = time()
sol = None
try:
if not app_args.simulate:
sol = algo_f(points, D_c, d, C, L, st, ewt,
run_single_iteration, minimize_K)
except (KeyboardInterrupt, Exception) as e:
if type(e) is KeyboardInterrupt:
interrupted = True
# if interrupted on initial sol gen, return the best of those
if len(e.args) > 0 and type(e.args[0]) is list:
sol = e.args[0]
if not app_args.minimal_output:
print("WARNING: Interrupted solving %s with %s" %
(bn, algo_abbreviation),
file=sys.stderr)
else:
if not app_args.minimal_output:
print("ERROR: Failed to solve %s with %s because %s" %
(bn, algo_abbreviation, str(e)),
file=sys.stderr)
sol = None
if sol:
sol = cvrp_ops.normalize_solution(sol)
if app_args.show_solution_cost:
sol_q = cvrp_ops.calculate_objective(sol, D_c)
else:
sol_q = cvrp_ops.calculate_objective(sol, D)
sol_K = sol.count(0) - 1
if app_args.local_search_operators:
if not app_args.minimal_output:
print("Postoptimize with %s ..." %
", ".join(app_args.local_search_operators),
end="")
sol = do_local_search(ls_ops, sol, D, d, C, L)
sol = cvrp_ops.normalize_solution(sol)
if app_args.show_solution_cost:
ls_sol_q = cvrp_ops.calculate_objective(sol, D_c)
else:
ls_sol_q = cvrp_ops.calculate_objective(sol, D)
if ls_sol_q < sol_q:
if not app_args.minimal_output:
print(" improved by %.2f%%." %
(1 - ls_sol_q / sol_q))
sol_q = ls_sol_q
sol_K = sol.count(0) - 1
else:
if not app_args.minimal_output:
print(" did not find improving moves.")
else:
sol_q = float('inf')
elapsed_t = time() - start_t
if app_args.minimal_output:
print("%s;%s" % (algo_abbreviation, bn), end="")
timecap_symbol = "*" if interrupted else ""
if sol:
feasible = all(
cvrp_ops.check_solution_feasibility(
sol, D_c, d, C, L, st))
print(";%s;%.2f;%d;%.2f%s" % (str(feasible), sol_q, sol_K,
elapsed_t, timecap_symbol))
else:
print(";False;inf;inf;%.2f%s" %
(elapsed_t, timecap_symbol))
elif sol:
# Minimal output is not enabled, print like crazy :)
if app_args.print_elapsed_time:
print("Algorithm produced a solution in %.2f s\n" %
(elapsed_t))
else:
#just a newline
print()
tightness = None
if C and sol_K:
tightness = (sum(d) / (C * sol_K))
if not bn in instance_data or sol_K < instance_data[bn][1]:
#"N K C tightness L st"
instance_data[bn] = (N, sol_K, C, "%.3f" % tightness, L,
st)
shared_cli.print_problem_information(
points,
D_c,
d,
C,
L,
st,
tightness,
verbosity=app_args.verbosity)
solution_print_verbosity = 3 if app_args.print_route_stat else 1
shared_cli.print_solution_statistics(sol, D, D_c, d, C, L, st,
solution_print_verbosity)
if app_args.print_vrph_sol:
print("SOLUTION IN VRPH FORMAT:")
print(" ".join(
str(n) for n in cvrp_io.as_VRPH_solution(sol)))
print("\n")
short_algo_name = algo_name
results[bn][short_algo_name] = sol_q
if interrupted:
break # algo loop
if interrupted:
break # problem file loop
## Print collected results
sys.stdout.flush()
sys.stderr.flush()
if not app_args.minimal_output and (len(results) > 1 or len(algos) > 1):
print("\n")
print_title = True
ls_label = "+".join(ls_algo_names)
for problem, algo_results in sorted(results.items()):
algo_names = [ "%s+%s"%(algo_name,ls_label) if ls_algo_names else (algo_name)\
for algo_name in sorted(algo_results.keys())]
if print_title:
instance_fields = "instance\t"
if PRINT_INSTANCE_DATA:
#"N K C tightness L st"
instance_fields += "N\tK*\tC\ttightness\tL\tst\t"
print(instance_fields + "\t".join(algo_names))
print_title = False
print(problem, end="")
if PRINT_INSTANCE_DATA:
print("\t", end="")
print("\t".join(str(e) for e in instance_data[problem]),
end="")
for _, result in sorted(algo_results.items()):
print("\t", result, end="")
print()
sys.stdout.flush()
sys.stderr.flush()
