def test_lm_cut_hmax_simple():
# define a simple test task
task1 = _get_simple_task()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts['var1'].hmax_value == 0.
assert heuristic.relaxed_facts['var2'].hmax_value == 1.
assert heuristic.relaxed_facts['var3'].hmax_value == float('inf')
assert heuristic.relaxed_facts['GOAL'].hmax_value == 1.
def test_lm_cut_hmax_simple():
# define a simple test task
task1 = _get_simple_task()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts["var1"].hmax_value == 0.0
assert heuristic.relaxed_facts["var2"].hmax_value == 1.0
assert heuristic.relaxed_facts["var3"].hmax_value == float("inf")
assert heuristic.relaxed_facts["GOAL"].hmax_value == 1.0
def test_lm_cut_hmax_intermediate():
task1 = _get_intermediate_task()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts["v1"].hmax_value == 0.0
assert heuristic.relaxed_facts["v2"].hmax_value == 1.0
assert heuristic.relaxed_facts["v3"].hmax_value == 2.0
assert heuristic.relaxed_facts["v6"].hmax_value == 2.0
assert heuristic.relaxed_facts["v4"].hmax_value == 3.0
assert heuristic.relaxed_facts["v5"].hmax_value == 3.0
assert heuristic.relaxed_facts["g"].hmax_value == 4.0
assert heuristic.relaxed_ops["op1"].hmax_supporter.name == "v1"
assert heuristic.relaxed_ops["op2"].hmax_supporter.name == "v2"
assert heuristic.relaxed_ops["op3"].hmax_supporter.name == "v3"
assert heuristic.relaxed_ops["op4"].hmax_supporter.name in {"v4", "v5"}
assert heuristic.relaxed_ops["op5"].hmax_supporter.name == "v2"
assert heuristic.relaxed_ops["op6"].hmax_supporter.name == "v6"
def test_lm_cut_hmax_intermediate():
task1 = _get_intermediate_task()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts['v1'].hmax_value == 0.
assert heuristic.relaxed_facts['v2'].hmax_value == 1.
assert heuristic.relaxed_facts['v3'].hmax_value == 2.
assert heuristic.relaxed_facts['v6'].hmax_value == 2.
assert heuristic.relaxed_facts['v4'].hmax_value == 3.
assert heuristic.relaxed_facts['v5'].hmax_value == 3.
assert heuristic.relaxed_facts['g'].hmax_value == 4.
assert heuristic.relaxed_ops['op1'].hmax_supporter.name == 'v1'
assert heuristic.relaxed_ops['op2'].hmax_supporter.name == 'v2'
assert heuristic.relaxed_ops['op3'].hmax_supporter.name == 'v3'
assert heuristic.relaxed_ops['op4'].hmax_supporter.name in {'v4', 'v5'}
assert heuristic.relaxed_ops['op5'].hmax_supporter.name == 'v2'
assert heuristic.relaxed_ops['op6'].hmax_supporter.name == 'v6'
def test_lm_cut_hmax_intermediate_two_paths():
task1 = _get_intermediate_task_two_paths()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts['v1'].hmax_value == 0.
assert heuristic.relaxed_facts['v2'].hmax_value == 1.
assert heuristic.relaxed_facts['v3'].hmax_value == 2.
assert heuristic.relaxed_facts['v6'].hmax_value == float('inf')
assert not heuristic.relaxed_facts['v6'] in heuristic.reachable
assert heuristic.relaxed_facts['v4'].hmax_value == 2.
assert heuristic.relaxed_facts['v5'].hmax_value == 2.
assert heuristic.relaxed_facts['v7'].hmax_value == 3.
assert heuristic.relaxed_facts['g'].hmax_value == 4.
assert heuristic.relaxed_ops['op1'].hmax_supporter.name == 'v1'
assert heuristic.relaxed_ops['op2'].hmax_supporter.name == 'v2'
assert heuristic.relaxed_ops['op3'].hmax_supporter.name == 'v3'
assert heuristic.relaxed_ops['op4'].hmax_supporter.name == 'v7'
assert heuristic.relaxed_ops['op5'].hmax_supporter.name == 'v2'
def test_lm_cut_hmax_intermediate_two_paths():
task1 = _get_intermediate_task_two_paths()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
assert heuristic.relaxed_facts["v1"].hmax_value == 0.0
assert heuristic.relaxed_facts["v2"].hmax_value == 1.0
assert heuristic.relaxed_facts["v3"].hmax_value == 2.0
assert heuristic.relaxed_facts["v6"].hmax_value == float("inf")
assert not heuristic.relaxed_facts["v6"] in heuristic.reachable
assert heuristic.relaxed_facts["v4"].hmax_value == 2.0
assert heuristic.relaxed_facts["v5"].hmax_value == 2.0
assert heuristic.relaxed_facts["v7"].hmax_value == 3.0
assert heuristic.relaxed_facts["g"].hmax_value == 4.0
assert heuristic.relaxed_ops["op1"].hmax_supporter.name == "v1"
assert heuristic.relaxed_ops["op2"].hmax_supporter.name == "v2"
assert heuristic.relaxed_ops["op3"].hmax_supporter.name == "v3"
assert heuristic.relaxed_ops["op4"].hmax_supporter.name == "v7"
assert heuristic.relaxed_ops["op5"].hmax_supporter.name == "v2"
def test_lm_cut_compute_single_cut():
task1 = _get_intermediate_task_two_paths()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
heuristic.goal_plateau.clear()
heuristic.compute_goal_plateau(heuristic.explicit_goal)
cut = heuristic.find_cut(task1.initial_state)
assert [op.name for op in cut] == ["op4"]
def test_lm_cut_heuristic_value_simple_task_always_true():
task1 = _get_simple_task_always_true()
heuristic = LmCutHeuristic(task1)
h_val = heuristic(make_root_node(task1.initial_state))
# print('Printing goal plateau')
# for f in heuristic.goal_plateau:
# print(repr(f))
# print('Printing cut')
# for op in cut:
# print(repr(op))
assert h_val == 1.0
def test_lm_cut_mark_single_goal():
task = _get_intermediate_task()
heuristic = LmCutHeuristic(task)
heuristic.compute_hmax(task.initial_state)
heuristic.compute_goal_plateau(heuristic.explicit_goal)
assert {f.name
for f in heuristic.goal_plateau} == {heuristic.explicit_goal, 'g'}
def test_lm_cut_blocksworld_initial_state():
parser = Parser("")
parser.domInput = blocks_dom
parser.probInput = blocks_problem_1
domain = parser.parse_domain(False)
problem = parser.parse_problem(domain, False)
task = grounding.ground(problem)
heuristic = LmCutHeuristic(task)
h_val = heuristic(make_root_node(task.initial_state))
assert h_val == 6.0
def test_lm_cut_compute_single_cut():
task1 = _get_intermediate_task_two_paths()
heuristic = LmCutHeuristic(task1)
heuristic.compute_hmax(task1.initial_state)
heuristic.goal_plateau.clear()
heuristic.compute_goal_plateau(heuristic.explicit_goal)
cut = heuristic.find_cut(task1.initial_state)
assert [op.name for op in cut] == ['op4']
def test_lm_cut_mark_multiple_goal():
task = _get_intermediate_task()
heuristic = LmCutHeuristic(task)
heuristic.compute_hmax(task.initial_state)
# artificially alter operator costs to get a larger goal plateau
heuristic.relaxed_ops['op4'].cost = 0.
heuristic.compute_goal_plateau(heuristic.explicit_goal)
assert ({f.name for f in heuristic.goal_plateau} in
[{heuristic.explicit_goal, 'v4', 'g'},
{heuristic.explicit_goal, 'v5', 'g'}])
def test_lm_cut_relaxed_operators():
# define a simple test task
task1 = _get_simple_task()
# test the relaxed operator generation
heuristic = LmCutHeuristic(task1)
assert heuristic.relaxed_facts['GOAL']
assert heuristic.relaxed_facts['var1']
assert heuristic.relaxed_facts['var2']
assert heuristic.relaxed_facts['var3']
assert 'GOALOP' in [o.name
for o in heuristic.relaxed_facts['GOAL'].effect_of]
assert not heuristic.relaxed_facts['GOAL'].precondition_of
assert ([o.name
for o in heuristic.relaxed_facts['var1'].precondition_of] ==
['op1', 'op2', 'GOALOP'])
assert ([o.name for o in heuristic.relaxed_facts['var1'].effect_of] ==
['op3'])
assert ([o.name
for o in heuristic.relaxed_facts['var2'].precondition_of] ==
['op3', 'GOALOP'])
assert ([o.name for o in heuristic.relaxed_facts['var2'].effect_of] ==
['op1'])
assert heuristic.relaxed_facts['var3'].precondition_of == []
assert heuristic.relaxed_facts['var3'].effect_of == []
assert heuristic.relaxed_ops['GOALOP']
assert heuristic.relaxed_ops['op1']
assert heuristic.relaxed_ops['op2']
assert heuristic.relaxed_ops['op3']
assert heuristic.relaxed_ops['GOALOP'].cost == 0
assert (sorted(f.name
for f in heuristic.relaxed_ops['GOALOP'].precondition) ==
['var1', 'var2'])
assert [f.name
for f in heuristic.relaxed_ops['GOALOP'].effects] == ['GOAL']
assert heuristic.relaxed_ops['op1'].cost == 1
assert [f.name
for f in heuristic.relaxed_ops['op1'].precondition] == ['var1']
assert [f.name for f in heuristic.relaxed_ops['op1'].effects] == ['var2']
assert heuristic.relaxed_ops['op2'].cost == 1
assert [f.name
for f in heuristic.relaxed_ops['op2'].precondition] == ['var1']
assert heuristic.relaxed_ops['op2'].effects == []
assert heuristic.relaxed_ops['op3'].cost == 1
assert [f.name
for f in heuristic.relaxed_ops['op3'].precondition] == ['var2']
assert [f.name for f in heuristic.relaxed_ops['op3'].effects] == ['var1']
assert not 'ALWAYSTRUE' in heuristic.relaxed_facts
def test_two_times_hmax_same_result():
task = _get_intermediate_task()
heuristic = LmCutHeuristic(task)
heuristic.compute_hmax(task.initial_state)
heuristic.compute_hmax(task.initial_state)
# artificially alter operator costs to get a larger goal plateau
heuristic.relaxed_ops["op4"].cost = 0.0
heuristic.compute_goal_plateau(heuristic.explicit_goal)
assert {f.name for f in heuristic.goal_plateau} in [
{heuristic.explicit_goal, "v4", "g"},
{heuristic.explicit_goal, "v5", "g"},
]
def test_lm_cut_relaxed_operators():
# define a simple test task
task1 = _get_simple_task()
# test the relaxed operator generation
heuristic = LmCutHeuristic(task1)
assert heuristic.relaxed_facts["GOAL"]
assert heuristic.relaxed_facts["var1"]
assert heuristic.relaxed_facts["var2"]
assert heuristic.relaxed_facts["var3"]
assert "GOALOP" in [o.name for o in heuristic.relaxed_facts["GOAL"].effect_of]
assert not heuristic.relaxed_facts["GOAL"].precondition_of
assert [o.name for o in heuristic.relaxed_facts["var1"].precondition_of] == [
"op1",
"op2",
"GOALOP",
]
assert [o.name for o in heuristic.relaxed_facts["var1"].effect_of] == ["op3"]
assert [o.name for o in heuristic.relaxed_facts["var2"].precondition_of] == [
"op3",
"GOALOP",
]
assert [o.name for o in heuristic.relaxed_facts["var2"].effect_of] == ["op1"]
assert heuristic.relaxed_facts["var3"].precondition_of == []
assert heuristic.relaxed_facts["var3"].effect_of == []
assert heuristic.relaxed_ops["GOALOP"]
assert heuristic.relaxed_ops["op1"]
assert heuristic.relaxed_ops["op2"]
assert heuristic.relaxed_ops["op3"]
assert heuristic.relaxed_ops["GOALOP"].cost == 0
assert sorted(f.name for f in heuristic.relaxed_ops["GOALOP"].precondition) == [
"var1",
"var2",
]
assert [f.name for f in heuristic.relaxed_ops["GOALOP"].effects] == ["GOAL"]
assert heuristic.relaxed_ops["op1"].cost == 1
assert [f.name for f in heuristic.relaxed_ops["op1"].precondition] == ["var1"]
assert [f.name for f in heuristic.relaxed_ops["op1"].effects] == ["var2"]
assert heuristic.relaxed_ops["op2"].cost == 1
assert [f.name for f in heuristic.relaxed_ops["op2"].precondition] == ["var1"]
assert heuristic.relaxed_ops["op2"].effects == []
assert heuristic.relaxed_ops["op3"].cost == 1
assert [f.name for f in heuristic.relaxed_ops["op3"].precondition] == ["var2"]
assert [f.name for f in heuristic.relaxed_ops["op3"].effects] == ["var1"]
assert not "ALWAYSTRUE" in heuristic.relaxed_facts
def test_lm_cut_relaxed_operators2():
# define a simple test task
task1 = _get_simple_task_always_true()
# test the relaxed operator generation
heuristic = LmCutHeuristic(task1)
assert "ALWAYSTRUE" in heuristic.relaxed_facts
def test_lm_cut_at_goal():
task = _get_simple_task_at_goal()
heuristic = LmCutHeuristic(task)
h_val = heuristic(make_root_node(task.initial_state))
heuristic.compute_goal_plateau(heuristic.explicit_goal)
assert h_val == 0.
def test_lm_cut_heuristic_value_two_initial_facts():
task1 = _get_intermediate_task_two_initial_facts()
heuristic = LmCutHeuristic(task1)
h_val = heuristic(make_root_node(task1.initial_state))
assert h_val == 4.0
def test_lm_cut_at_goal():
task = _get_simple_task_at_goal()
heuristic = LmCutHeuristic(task)
h_val = heuristic(make_root_node(task.initial_state))
heuristic.compute_goal_plateau(heuristic.explicit_goal)
assert h_val == 0.0
