Spiderplan is a constraint-based automated planner.
cd spiderplan
sbt clean; compile; test; publishM2
cd spiderplan
sbt run test/robot-move-pick-place/domain.aiddl test/robot-move-pick-place/problem-01.aiddl
Temporal intervals consist of two time points for their start-time and end-time. SpiderPlan uses flexible temporal intervals which means that start- and end-time both use a range of possible values from an earliest to a latest value. As a result we have four relevant values for each time point: the Earliest Start Time (EST), Latest Start Time (LST), Earliest End Time (EET), and Latest End Time (LET).
Spider plan represents both problems and solutions as Constraint Databases (CDBs). A constraint database is a set of constraints grouped by type. It has a form
CDB = {
t1:{a b c}
t2:{e f g}
}
As an AIDDL term, a CDB is a set of key-value pairs whose keys are types and whose values are sets of constraints of the given type.
A statement
(I x:v)
links a state-variable x with value v to a temporal interval I. As an
AIDDL term, a statement is a tuple consistng of two elements: the interval and a
key-value pair whose key is x and whose value us v.
In a CDB, statements use the key statement.
statement:{
(s0_1 (at r1):loc1)
(s0_2 (object-at o1):loc2)
(s0_3 (adjacent loc1 loc2):true)
(s0_4 (adjacent loc2 loc1):true)
}
Temporal constraints restrict temporal intervals by imposing limitations on
their start- and end-times. These constraints mainly come in two flavors: unary
(e.g., duration, release, deadline) and binary (e.g., before, equals,
overlaps, during). In addition, there is one special temporal constraint
intersectio-possible which is satisfied if a set of intervals may have a
non-empty intersection.
In a CDB, temporal constraints use the key temporal.
temporal:{
(duration s0_3 (+INF +INF))
(duration s0_4 (+INF +INF))
(duration s0_2 (1 +INF))
(release s0_2 (0 0))
(duration s0_1 (1 +INF))
(release s0_3 (0 0))
(equals s0_1 G1)
(release s0_1 (0 0))
(release s0_4 (0 0))
}
Open goals are statements that the planner should achieve. They have exactly the
same form as operators, but appear in their own constraint type goal.
Open goals can be satisfied by setting their interval equal to a statement in the CDB which uses the same variable and value. If no such statement exists, operators allow for more complex ways to change a CDB.
An operator is a tuple of key-value pairs:
namename and possibly parameters of the operatorsignaturetype map for each parameteridvariable used to make the operator unique where neede (e.g., making sure it uses unique temporal intervals)intervalterm that represents the interval of the operator itself (usually depends on the ID term)preconditionscollection of statements that must exist in a CDB for the operator to be applicableeffectscollection of statements that will be added to a CDB if the operator is appliedconstraintsa CDB that will be added to a CDB when the operator is applied. So adding an operator may lead to the addition of any type of constraint. Typically, constraints at least connect preconditions and effects to the operator's main interval
To signal that an open-goal has been achieved, its interval will be added to a
set under a constraint type goal.sat.
In addtition to the regular set of constraints, Spiderplan offers libraries for constraints solved by external solvers. Licenses may be different from Spiderplan due to the usage of external libraries. For this reason and to avoid cluttering the main SpiderPlan library wth dependencies, these solvers are in their own libraries.
Allows to put constraints on Prolog background knowledge. In some cases, these constraints can be removed during preprocessing. Operators, for instance, can be replaced by operators that fulfill their Prolog constraints.
Allows combined task and motion planning and multi-robot coordination by attaching motion constraints to planning operators.