Add an alternative planning approach using Evrete

[russmiles]

The Evrete planning approach provides an alternative to the current GOAP approach in Embabel.

It might be interesting to look at the relative advantages or disadvantages of supporting an alternative planning approach, and then consider whether we could implement such a thing perhaps with libraries like https://www.evrete.org/

[johnsonr]

This definitely makes sense to look at. Planner is pluggable to allow alternatives to GOAP.

Something like Rete makes sense. We could of course also use a reasoning model, surfacing the plan and even checking it, but I think having planning based on facts is an important tenet of Embabel.

[igordayen]

perhaps planning algorithm execution engine itself might be decoupled theoreticlally from rules / condition execution.

As for planning system models there is around so called PDDL (planning domain definition model). Below is partial list of known planning models (please refer to http://pddl4j.imag.fr/running_planners_from_command_line.html):

FF (FastForward). It is based on Enforced Hill Climbing algorithm and A* search and the relaxed planning graph heuristic devised by J. Hoffmann.

HSP (Heuristic Search Planner). It is based on A search* that can be combined with any goal cost heuristics functions developed in the library.

GSP (Generic Search Planner). This planner is a generic state space planner. It is possible to choose the search strategy and the goal cost heuristic function.

TFD (Total-order Forward Decomposition). It is based on dept first search search strategy and can only deal with total-order task decomposition.

PFD (Partial-order Forward Decomposition). It is based on dept first search search strategy and can deal with partial-order task decomposition.

[johnsonr]

I added a planner field taking an enum to the Action annotation. Currently the only allowed value is GOAP, but this will allow alternatives in future.