Is it possible to get intermediate states rather than just action sequences through the API

[jack2200]

As stated above, please let me know which function can be implemented if possible, thanks

[pellierd]

From the coded problem, you can obtain the initial state with the method getInit(). It returns a BitExp. You have to create a BitState from this BitExp. Then, you have to apply the actions of the plan to get the intermediate states with method apply().

[jack2200]

Is this the case with the latest PDDL4j 4.0.0? Is there an easy way to do this? I think there is a big difference between 4.0.0 and 3.8.3

[pellierd]

I didn't change this point. The intermediate states are not stored in the plan.

[khanetor]

From the coded problem, you can obtain the initial state with the method getInit(). It returns a BitExp. You have to create a BitState from this BitExp. Then, you have to apply the actions of the plan to get the intermediate states with method apply().

Could you provide an example? I couldn't effectively navigate the code to do this.

[khanetor]

In case anyone else need to get the intermediate states, here is the solution in Scala:

val planner = new FF()
// ...
val problem = planner.instantiate(parsedProblem)
val plan = planner.solve(problem)
val state = new State(problem.getInitialState())

val states = plan.actions.asScala.toList.foldLeft(List(problem.toString(state)))((iStates, action) =>
  state.apply(action.getConditionalEffects())
  problem.toString(state) :: iStates
)

You can adapt this to Java as needed.