def add_atom(self, atom):
if not is_atom(atom):
return False
head = atom.head
if head in self.atoms:
return False
self.atoms.add(head)
# TODO: doing this in a way that will eventually allow constants
for i, stream in enumerate(self.streams):
for j, domain_atom in enumerate(stream.domain):
if get_prefix(head) != get_prefix(domain_atom):
continue
if len(head.args) != len(get_args(domain_atom)):
raise ValueError(head, domain_atom)
if any(
isinstance(b, Object) and (a != b)
for (a, b) in zip(head.args, get_args(domain_atom))):
continue
self.atoms_from_domain[(i, j)].append(head)
values = [
self.atoms_from_domain[(i, k)] if j != k else [head]
for k in range(len(stream.domain))
]
domain = list(map(head_from_fact, stream.domain))
#domain = stream.domain
for combo in product(*values):
mapping = get_mapping(domain, combo)
if mapping is None:
continue
input_objects = tuple(mapping[p] for p in stream.inputs)
self._add_instance(stream, input_objects)
return True
def get_mapping(atoms1, atoms2, initial={}):
assert len(atoms1) == len(atoms2)
mapping = initial.copy()
for a1, a2 in zip(atoms1, atoms2):
assert (get_prefix(a1) == get_prefix(a2))
#for arg1, arg2 in zip(get_args(a1), a2.args): # TODO: this is because eval vs predicate
for arg1, arg2 in zip(a1.args, a2.args):
if mapping.get(arg1, arg2) == arg2:
mapping[arg1] = arg2
else:
return None
return mapping
def compile_to_exogenous_axioms(evaluations, domain, streams):
import pddl
fluent_predicates = get_fluents(domain)
domain_predicates = {get_prefix(a) for s in streams for a in s.domain}
if not (domain_predicates & fluent_predicates):
return
certified_predicates = {get_prefix(a) for s in streams for a in s.certified}
future_map = {p: 'f-{}'.format(p) for p in certified_predicates}
augment_evaluations(evaluations, future_map)
rename_future = lambda a: rename_atom(a, future_map)
derived_map = {p: 'd-{}'.format(p) for p in certified_predicates}
rename_derived = lambda a: rename_atom(a, derived_map)
for action in domain.actions:
action.precondition = replace_predicates(derived_map, action.precondition)
for effect in action.effects:
assert(isinstance(effect, pddl.Effect))
effect.condition = replace_predicates(derived_map, effect.condition)
for axiom in domain.axioms:
axiom.condition = replace_predicates(derived_map, axiom.condition)
#fluent_predicates.update(certified_predicates)
for stream in list(streams):
if not isinstance(stream, Stream):
raise NotImplementedError(stream)
streams.append(create_static_stream(stream, evaluations, fluent_predicates, rename_future))
stream_atom = streams[-1].certified[0]
domain.predicate_dict[get_prefix(stream_atom)] = pddl.Predicate(get_prefix(stream_atom), get_args(stream_atom))
precondition = pddl.Conjunction(tuple(map(fd_from_fact, (stream_atom,) + tuple(map(rename_derived, stream.domain)))))
for fact in stream.certified:
derived_fact = fd_from_fact(rename_derived(fact))
external_params = derived_fact.args
internal_params = tuple(p for p in (stream.inputs + stream.outputs)
if p not in derived_fact.args)
parameters = tuple(pddl.TypedObject(p, 'object')
for p in (external_params + internal_params))
#precondition = pddl.Conjunction(tuple(map(fd_from_fact, [stream_atom] +
# list(map(rename_derived, stream.domain)))))
#precondition = pddl.Disjunction([fd_from_fact(fact), precondition]) # TODO: quantifier
domain.axioms.extend([
pddl.Axiom(name=derived_fact.predicate, parameters=parameters,
num_external_parameters=len(external_params),
condition=precondition),
pddl.Axiom(name=derived_fact.predicate, parameters=parameters[:len(external_params)],
num_external_parameters=len(external_params),
condition=fd_from_fact(fact))])
stream.certified = tuple(set(stream.certified) |
set(map(rename_future, stream.certified)))
def fd_from_fact(fact):
# TODO: convert to evaluation?
prefix = get_prefix(fact)
if prefix == NOT:
return fd_from_fact(fact[1]).negate()
if prefix == EQ:
_, head, value = fact
predicate = get_prefix(head)
args = map(pddl_from_object, get_args(head))
fluent = pddl.f_expression.PrimitiveNumericExpression(symbol=predicate,
args=args)
expression = pddl.f_expression.NumericConstant(value)
return pddl.f_expression.Assign(fluent, expression)
args = map(pddl_from_object, get_args(fact))
return pddl.Atom(prefix, args)
def plan_functions(functions, externals):
external_from_function = {}
for external in filter(lambda e: isinstance(e, Function), externals):
assert external.function not in external_from_function
external_from_function[external.function] = external
function_plan = set()
for term in functions:
if get_prefix(term) not in external_from_function:
raise ValueError('{} is not implemented'.format(get_prefix(term)))
external = external_from_function[get_prefix(term)]
instance = external.get_instance(get_args(term))
[result] = instance.next_optimistic()
function_plan.add(result)
print('Function plan:', str_from_object(function_plan))
return function_plan
def create_static_stream(stream, evaluations, fluent_predicates, get_future):
def static_fn(*input_values):
instance = stream.get_instance(tuple(map(Object.from_value, input_values)))
if all(evaluation_from_fact(f) in evaluations for f in instance.get_domain()):
return None
return tuple(FutureValue(stream.name, input_values, o) for o in stream.outputs)
#opt_evaluations = None
def static_opt_gen_fn(*input_values):
instance = stream.get_instance(tuple(map(Object.from_value, input_values)))
if all(evaluation_from_fact(f) in evaluations for f in instance.get_domain()):
return
for output_values in stream.opt_gen_fn(*input_values):
yield output_values
# TODO: need to replace regular opt_gen_fn to update opt_evaluations
# if I want to prevent switch from normal to static in opt
# Focused algorithm naturally biases against using future because of axiom layers
stream_name = 'future-{}'.format(stream.name)
gen_fn = from_fn(static_fn)
static_domain = list(filter(lambda a: get_prefix(a) not in fluent_predicates, stream.domain))
new_domain = list(map(get_future, static_domain))
stream_atom = ('{}-result'.format(stream.name),) + tuple(stream.inputs + stream.outputs)
new_certified = [stream_atom] + list(map(get_future, stream.certified))
static_stream = Stream(stream_name, gen_fn, stream.inputs, new_domain,
stream.outputs, new_certified, stream.info)
static_stream.opt_gen_fn = static_opt_gen_fn
return static_stream
def create_domain(goal_facts):
domain = make_domain()
for fact in goal_facts: # TODO: consider removing this annoying check
name = get_prefix(fact)
parameters = ['?x{}'.format(i) for i in range(len(get_args(fact)))]
add_predicate(domain, make_predicate(name, parameters))
return domain
def compile_to_exogenous_actions(evaluations, domain, streams):
import pddl
# TODO: automatically derive fluents
# TODO: version of this that operates on fluents of length one?
# TODO: better instantiation when have full parameters
# TODO: conversion from stream cost to real cost units?
# TODO: any predicates derived would need to be replaced as well
fluent_predicates = get_fluents(domain)
domain_predicates = {get_prefix(a) for s in streams for a in s.domain}
if not (domain_predicates & fluent_predicates):
return
certified_predicates = {get_prefix(a) for s in streams for a in s.certified}
future_map = {p: 'f-{}'.format(p) for p in certified_predicates}
augment_evaluations(evaluations, future_map)
rename_future = lambda a: rename_atom(a, future_map)
for stream in list(streams):
if not isinstance(stream, Stream):
raise NotImplementedError(stream)
# TODO: could also just have conditions asserting that one of the fluent conditions fails
streams.append(create_static_stream(stream, evaluations, fluent_predicates, rename_future))
stream_atom = streams[-1].certified[0]
parameters = [pddl.TypedObject(p, 'object') for p in get_args(stream_atom)]
# TODO: add to predicates as well?
domain.predicate_dict[get_prefix(stream_atom)] = pddl.Predicate(get_prefix(stream_atom), parameters)
precondition = pddl.Conjunction(tuple(map(fd_from_fact, (stream_atom,) + tuple(stream.domain))))
effects = [pddl.Effect(parameters=[], condition=pddl.Truth(),
literal=fd_from_fact(fact)) for fact in stream.certified]
effort = 1 # TODO: use stream info
#effort = 1 if unit_cost else result.instance.get_effort()
#if effort == INF:
# continue
fluent = pddl.PrimitiveNumericExpression(symbol=TOTAL_COST, args=[])
expression = pddl.NumericConstant(int_ceil(effort)) # Integer
cost = pddl.Increase(fluent=fluent, expression=expression) # Can also be None
domain.actions.append(pddl.Action(name='call-{}'.format(stream.name),
parameters=parameters,
num_external_parameters=len(parameters),
precondition=precondition, effects=effects, cost=cost))
stream.certified = tuple(set(stream.certified) |
set(map(rename_future, stream.certified)))
def __init__(self, head, fn, domain, info):
if info is None:
info = FunctionInfo(p_success=self._default_p_success,
overhead=self._default_overhead)
super(Function, self).__init__(get_prefix(head), info, get_args(head),
domain)
self.head = head
opt_fn = lambda *args: self._codomain()
if fn == DEBUG:
fn = opt_fn
self.fn = fn
self.opt_fn = opt_fn if (
self.info.opt_fn is None) else self.info.opt_fn
def parse_common(lisp_list, stream_map, stream_info):
assert (2 <= len(lisp_list) <= 3)
head = tuple(lisp_list[1])
assert (is_head(head))
name = get_prefix(head)
if stream_map == DEBUG:
fn = DEBUG
else:
if name not in stream_map:
raise ValueError('Undefined external function: {}'.format(name))
fn = stream_map[name]
domain = []
if len(lisp_list) == 3:
domain = list_from_conjunction(lisp_list[2])
info = stream_info.get(name, None)
return head, fn, domain, info
def visualize_constraints(constraints,
filename='constraint_network.pdf',
use_functions=True):
from pygraphviz import AGraph
graph = AGraph(strict=True, directed=False)
graph.node_attr['style'] = 'filled'
graph.node_attr['shape'] = 'circle'
graph.node_attr['fontcolor'] = 'black'
graph.node_attr['colorscheme'] = 'SVG'
graph.edge_attr['colorscheme'] = 'SVG'
functions = set()
heads = set()
for fact in constraints:
if get_prefix(fact) == EQ:
functions.add(fact[1])
else:
heads.add(fact)
heads.update(functions)
objects = {a for head in heads for a in get_args(head)}
optimistic_objects = filter(lambda o: isinstance(o, OptimisticObject),
objects)
for opt_obj in optimistic_objects:
graph.add_node(str(opt_obj), shape='circle', color=PARAMETER_COLOR)
for head in heads:
if not use_functions and (head in functions):
continue
# TODO: prune values w/o free parameters?
name = str_from_tuple(head)
color = COST_COLOR if head in functions else CONSTRAINT_COLOR
graph.add_node(name, shape='box', color=color)
for arg in get_args(head):
if arg in optimistic_objects:
graph.add_edge(name, str(arg))
graph.draw(filename, prog='dot')
return graph
def next_results(self, accelerate=1, verbose=False):
start_time = time.time()
assert not self.enumerated
self.enumerated = True
input_values = self.get_input_values()
try:
value = self.external.fn(*input_values)
except TypeError:
raise TypeError('Function [{}] expects {} inputs'.format(
self.external.name, len(input_values)))
self.value = self.external._codomain(value)
# TODO: cast the inputs and test whether still equal?
#if not (type(self.value) is self.external._codomain):
#if not isinstance(self.value, self.external._codomain):
if self.value != value:
raise ValueError(
'Function [{}] produced a nonintegral value [{}]. '
'FastDownward only supports integral costs. '
'To "use" real costs, scale each cost by a large factor, '
'capturing the most significant bits.'.format(
self.external.name, self.value))
if self.value < 0:
raise ValueError(
'Function [{}] produced a negative value [{}]'.format(
self.external.name, self.value))
if verbose:
print('{}{}={}'.format(get_prefix(self.external.head),
str_from_tuple(self.get_input_values()),
self.value))
results = [self.external._Result(self, self.value)]
if isinstance(self, PredicateInstance) and (
self.value != self.external.opt_fn(*input_values)):
self.update_statistics(start_time,
[]) # TODO: do this more automatically
else:
self.update_statistics(start_time, results)
return results
def get_stream_instances(stream_plan):
import pddl
# TODO: something that inverts the negative items
stream_instances = [] # TODO: could even apply these to the state directly
for result in stream_plan:
name = result.instance.external.name
precondition = list(map(fd_from_fact, result.instance.get_domain()))
effects = [([], fd_from_fact(fact)) for fact in result.get_certified() if not get_prefix(fact) == EQ]
cost = None # TODO: effort?
instance = pddl.PropositionalAction(name, precondition, effects, cost)
stream_instances.append(instance)
return stream_instances
def rename_atom(atom, mapping):
name = get_prefix(atom)
if name not in mapping:
return atom
return (mapping[name],) + get_args(atom)
def get_constraints(objectives):
return [o for o in objectives if get_prefix(o) != MINIMIZE]
def get_costs(objectives):
return [o for o in objectives if get_prefix(o) == MINIMIZE]
def get_parameters(expression):
head = get_prefix(expression)
if head in [NOT, MINIMIZE]:
return get_parameters(get_args(expression)[0])
return list(filter(is_parameter, get_args(expression)))
def add_fact(fact):
head, color = (fact[1], COST_COLOR) if get_prefix(fact) == EQ else (
fact, CONSTRAINT_COLOR)
s_fact = str_from_tuple(head)
graph.add_node(s_fact, shape='box', color=color)
return s_fact
