Separating PathSearchSolution from Forward and from Backward solvers.

[pevnak]

Hi,

I wonder if it is possible at the moment to identify, from which solver the PathSearchSolution comes from. I guess the only way is to test, if first (start) and last(goal) state in trajectory have parents, but it is a bit hacky. Since the search_tree is effectively in different order, would not make sense to add flag to PathSearchSolution (or flag to the type of PathSearchSolution) or use separate types?

Tomas

[ztangent]

That's certainly one way of testing it! In the future I think it could be made clear from creating a PartialState type that the backward planner will use and store. But what is the use case for making this distinction?

[pevnak]

I am playing a bit with bootstrapping, where I am optimizing heuristics (realized by NN) from problems it has solved and for construction, I need to know the direction of the search tree, such that I know, where is the goal state.

I think for now, I will use the heuristic I have figured out when I was asking my questions above.