concepts.dm.crow.planners.regression_planning_impl.crow_regression_planner_iddfs_v1.CrowRegressionPlannerIDDFSv1

class CrowRegressionPlannerIDDFSv1

Bases: CrowRegressionPlanner

Methods

dfs(planning_state, depth)	The main entry of the depth-first search.
dfs_impl(planning_state, depth)	The actual implementation of the depth-first search.
dfs_inner(planning_state, stmt, ...)
evaluate(expression, state[, csp, ...])	Evaluate an expression and return the result.
hash_state(planning_state)
hash_state_human(planning_state)	Hash the planning state for human-readable output.
main()
main_entry(program[, minimize])
print_visited_states_count()
set_enable_human_control_interface([...])
set_enable_visited_states_count([...])
set_results(results)

Attributes

domain
results
search_stat
always_commit_skeleton	Whether to always commit the skeleton of the program.
min_search_depth	The minimum search depth.
include_effect_appliers	Whether to include the effect appliers in the search result.
enable_state_hash	Whether to enable the state hash.
executor	The executor.
state	The initial state.
goal_expr	The goal expression.
simulation_interface	The simulation interface.
is_goal_ordered	Whether the goal is ordered.
is_goal_serializable	Whether the goal is serializable.
is_goal_refinement_compressible	Whether the refinement for each component of the goal is compressible.
enable_csp	Whether to enable the CSP solver.
max_csp_trials	The maximum number of CSP trials (for solving a single CSP).
max_global_csp_trials	The maximum number of global CSP trials.
max_csp_branching_factor	The maximum CSP branching factor.
use_generator_manager	Whether to use the generator manager.
store_generator_manager_history	Whether to store the generator manager history.
include_dependency_trace	Whether to include the dependency graph in the search result.
verbose	Whether to output verbose information.

__init__(executor, state, goal_expr, *, simulation_interface=None, is_goal_ordered=True, is_goal_serializable=True, is_goal_refinement_compressible=True, enable_csp=True, max_csp_trials=1, max_global_csp_trials=100, max_csp_branching_factor=5, use_generator_manager=True, store_generator_manager_history=False, include_effect_appliers=False, include_dependency_trace=False, verbose=False, **kwargs)

Initialize the planner.

Parameters:

• executor (CrowExecutor) – the executor.

• state (CrowState) – the initial state.

• goal_expr (str | ValueOutputExpression | None) – the goal expression.

• simulation_interface (CrowSimulationControllerInterface | None) – the simulation interface.

• enable_reordering – whether to enable reordering.

• is_goal_ordered (bool) – whether the goal is ordered.

• is_goal_serializable (bool) – whether the goal is serializable.

• is_goal_refinement_compressible (bool) – whether the goal refinement is compressible.

• enable_csp (bool) – whether to enable the CSP solver.

• max_csp_trials (int) – the maximum number of CSP trials.

• max_global_csp_trials (int) – the maximum number of global CSP trials.

• max_csp_branching_factor (int) – the maximum CSP branching factor.

• use_generator_manager (bool) – whether to use the generator manager.

• store_generator_manager_history (bool) – whether to store the generator manager history.

• include_effect_appliers (bool) – whether to include the effect appliers in the search result. The effect appliers are of type CrowEffectApplier.

• include_dependency_trace (bool) – whether to include the dependency graph in the search result.

• verbose (bool) – whether to output verbose information.

__new__(**kwargs)

dfs(planning_state, depth)

The main entry of the depth-first search. This function is a generator of possible planning results. Internally, it calls dfs_impl() to perform the actual search but wraps around it to handle visited states (so that we can avoid redundant search).

Parameters:

• planning_state (_IDDFSNode) – the current planning state.

• depth (int) – the current search depth.

Return type:

Iterator[CrowPlanningResult3]

dfs_impl(planning_state, depth)

The actual implementation of the depth-first search. This function is a generator of possible planning results.

Parameters:

• planning_state (_IDDFSNode) – the current planning state.

• depth (int) – the current search depth.

Return type:

Iterator[CrowPlanningResult3]

dfs_inner(planning_state, stmt, middle_program, scope_id, depth)

Parameters:

• planning_state (_IDDFSNode)

• stmt (CrowBehaviorOrderingSuite | CrowBehaviorForeachLoopSuite | CrowBehaviorWhileLoopSuite | CrowBehaviorConditionSuite | CrowBindExpression | CrowAssertExpression | CrowRuntimeAssignmentExpression | CrowControllerApplicationExpression | CrowAchieveExpression | CrowUntrackExpression | CrowBehaviorApplicationExpression | CrowBehaviorEffectApplicationExpression | CrowBehaviorCommit)

• middle_program (CrowBehaviorOrderingSuite | None)

• scope_id (int)

• depth (int)

Return type:

Iterator[CrowPlanningResult3]

evaluate(expression, state, csp=None, bounded_variables=None, clone_csp=True, force_tensor_value=False)

Evaluate an expression and return the result.

Parameters:

• expression (ObjectOrValueOutputExpression | VariableAssignmentExpression) – the expression to evaluate.

• state (CrowState) – the current state.

• csp (ConstraintSatisfactionProblem | None) – the current CSP.

• bounded_variables (Dict[str, ObjectConstant | TensorValue] | None) – the bounded variables.

• clone_csp (bool) – whether to clone the CSP.

• force_tensor_value (bool) – whether to force the result to be a tensor value.

Returns:

the evaluation result and the updated CSP.

Return type:

Tuple[None | StateObjectReference | StateObjectList | TensorValue | OptimisticValue, ConstraintSatisfactionProblem | None]

hash_state(planning_state)

Parameters:

planning_state (_IDDFSNode)

Return type:

Tuple[str, …]

hash_state_human(planning_state)

Hash the planning state for human-readable output. Note that this function is not toally consistent with hash_state. This function is a “stronger” version of hash_state. For example, the following two states with different hash_state corresponds to the same human-readable hash.

State1:

unordered{ achieve@4 clear(V::x) }
Scopes:
    0: {}
    1: {}
    2: {'x': ObjectConstant<A: block>, 'y': ObjectConstant<B: block>}
    3: {'x': ObjectConstant<B: block>, 'y': ObjectConstant<C: block>}
    4: {'x': ObjectConstant<B: block>}

State2:

unordered{ achieve@2 clear(V::y) }
Scopes:
    0: {}
    1: {}
    2: {'x': ObjectConstant<A: block>, 'y': ObjectConstant<B: block>}
    3: {'x': ObjectConstant<B: block>, 'y': ObjectConstant<C: block>}
    4: {'x': ObjectConstant<B: block>}

Parameters:

planning_state (_IDDFSNode) – the planning state to hash.

Returns:

the hash of the planning state. It is a tuple of strings.

Return type:

Tuple[str, …]

main()

Return type:

Tuple[List[Tuple[CrowControllerApplier, …]], dict]

main_entry(program, minimize=None)

Parameters:

• program (CrowBehaviorOrderingSuite)

• minimize (ValueOutputExpression | None)

Return type:

List[Tuple[CrowControllerApplier, …]]

print_visited_states_count()

set_enable_human_control_interface(human_control_interface=True)

Parameters:

human_control_interface (bool)

set_enable_visited_states_count(enable_visited_states_count=True)

Parameters:

enable_visited_states_count (bool)

set_results(results)

Parameters:

results (List[CrowPlanningResult])

Return type:

None

always_commit_skeleton: bool

Whether to always commit the skeleton of the program.

property domain: CrowDomain

enable_csp: bool

Whether to enable the CSP solver.

enable_state_hash: bool

Whether to enable the state hash. When there are not a lot of duplicated states during search, turning this off can improve the performance.

executor: CrowExecutor

The executor.

goal_expr: ValueOutputExpression | None

The goal expression. If this is None, the goal will be extracted from the behavior named ‘__goal__’.

include_dependency_trace: bool

Whether to include the dependency graph in the search result.

include_effect_appliers: bool

Whether to include the effect appliers in the search result. The effect appliers are of type CrowEffectApplier.

is_goal_ordered: bool

Whether the goal is ordered. This only takes effect when the goal is a set of AND-connected expressions.

is_goal_refinement_compressible: bool

Whether the refinement for each component of the goal is compressible. This only takes effect when the goal is a set of AND-connected expressions.

is_goal_serializable: bool

Whether the goal is serializable. This only takes effect when the goal is a set of AND-connected expressions.

max_csp_branching_factor: int

The maximum CSP branching factor.

max_csp_trials: int

The maximum number of CSP trials (for solving a single CSP).

max_global_csp_trials: int

The maximum number of global CSP trials.

min_search_depth: int

The minimum search depth.

property results: List[CrowPlanningResult]

property search_stat: dict

simulation_interface: CrowSimulationControllerInterface | None

The simulation interface.

state: CrowState

The initial state.

store_generator_manager_history: bool

Whether to store the generator manager history.

use_generator_manager: bool

Whether to use the generator manager.

verbose: bool

Whether to output verbose information.