concepts.pdsketch.planners.optimistic_search

CSP-based PDSketch planner. At the high-level, this planner defers the decision of all continuous parameters to a CSP solver. During forward search, the planner only make decisions about discrete parameters (mostly object-typed parameters), and record all preconditions and goal tests that involve continuous parameters to an accumulated CSP data structure.

At each step, the planner will internally call a CSP solver to solve the accumulated CSP. If the CSP solver returns a solution, the planner will generate the corresponding fully-grounded trajectory based on the discrete parameters and the CSP solution.

Functions

apply_action(executor, state, action, csp)	Apply an action to a state.
construct_csp_from_optimistic_plan(executor, ...)	Construct a CSP from an optimistic plan.
goal_test(executor, state, goal_expr, csp, ...)	Test whether a state satisfies a goal expression with DPLL+Sampling CSP solver.
ground_action(executor, action, assignments)	Ground a single action with a given assignment to the underlying CSP.
ground_actions(executor, actions, assignments)	Ground a list of actions with a given assignment to the underlying CSP.
instantiate_action(csp, action)	Instantiate an action by replacing all placeholder values with a new optimistic variable.
optimistic_search(executor, state, goal_expr, *)	Perform brute-force search with DPLL+Sampling algorithm for mixed discrete-continuous domains.
optimistic_search_domain_check(domain)	Check whether a domain is suitable for optimistic search.
optimistic_search_strips(executor, state, ...)	Perform heuristic search with DPLL+Sampling algorithm and STRIPS-based heuristics, for mixed discrete-continuous domains.
prepare_optimistic_search(func_name, ...[, ...])	Prepare for optimistic search.
solve_optimistic_plan(executor, state, ...)	Solve an optimistic plan using the DPLL+Sampling algorithm.

Classes

OptHeuristicSearchState

The state for heuristic search.