Merge pull request #562 from aiplan4eu/fix-proto-explicit-init

fix: proto explicit init

setup.py

@@ -29,7 +29,7 @@
         "fast-downward": ["up-fast-downward==0.4.1"],
         "lpg": ["up-lpg==0.0.10"],
         "fmap": ["up-fmap==0.0.13"],
-        "aries": ["up-aries>=0.3.2"],
+        "aries": ["up-aries>=0.3.3"],
         "symk": ["up-symk>=1.0.1"],
         "engines": [
             "tarski[arithmetic]",

unified_planning/grpc/proto_writer.py

@@ -603,7 +603,7 @@ def _convert_problem(self, problem: model.Problem) -> proto.Problem:
             actions=[self.convert(a) for a in problem.actions],
             initial_state=[
                 proto.Assignment(fluent=self.convert(x), value=self.convert(v))
-                for (x, v) in problem.initial_values.items()
+                for (x, v) in problem.explicit_initial_values.items()
             ],
             timed_effects=timed_effects,
             goals=goals,

unified_planning/grpc/unified_planning.proto

@@ -116,7 +116,6 @@ message Fluent {
     // If non-empty, then any state variable using this fluent that is not explicitly given a value in the initial state
     // will be assumed to have this default value.
     // This allows mimicking the closed world assumption by setting a "false" default value to predicates.
-    // Note that in the initial state of the problem message, it is assumed that all default values are set.
     Expression default_value = 4;
 }
 
@@ -539,7 +538,8 @@ message Problem {
     // features: ACTION_BASED
     repeated Action actions = 6;
 
-    // Initial state, including default values of state variables.
+    // Explicit assignments to state variables in the initial state.
+    // State variables not assigned there and will take the default value of their fluent, if any.
     repeated Assignment initial_state = 7;
 
     // Facts and effects that are expected to occur strictly later than the initial state.

unified_planning/test/examples/minimals.py

@@ -363,6 +363,31 @@ def get_example_problems():
     )
     problems["basic_with_costs"] = basic_with_costs
 
+    # basic with defaults
+    problem = Problem("basic_with_default_values")
+    object = UserType("object")
+    objects = [problem.add_object(f"o{i}", object) for i in range(0, 5)]
+    available = Fluent("available", BoolType(), a=object)
+
+    on = Fluent("on", object, a=object)
+    problem.add_fluent(available, default_initial_value=True)
+    problem.add_fluent(on, default_initial_value=objects[0])
+    goal = problem.add_fluent("g", default_initial_value=False)
+    for i in [0, 1, 3, 4]:  # override default for all but objects[2]
+        problem.set_initial_value(on(objects[i]), objects[4])
+    act_a = InstantaneousAction("a", obj=object)
+    act_a.add_precondition(available(objects[0]))
+    act_a.add_precondition(Equals(on(act_a.obj), objects[0]))
+    act_a.add_effect(goal, True)
+    problem.add_action(act_a)
+    problem.add_goal(goal)
+
+    problems["basic_with_default_values"] = TestCase(
+        problem=problem,
+        solvable=True,
+        valid_plans=[up.plans.SequentialPlan([act_a(objects[2])])],
+    )
+
     # counter
     counter_1 = Fluent("counter_1", IntType(0, 10))
     counter_2 = Fluent("counter_2", IntType(0, 10))