search

laas / fape

FAPE (Flexible Acting and Planning Environment) is a constraint-based planner for the ANML language.
BSD 2-Clause "Simplified" License
31 stars 7 forks source link

FAPE in the ANML landscape? #8

Open d-vyd opened 2 years ago

d-vyd commented 2 years ago

Hi Arthur,

I read through the issues and understand that you are not continuing to develop FAPE. Can it currently function as an embedded tool in a larger JVM application, and if so is there a way to hand the planner a domain/problem without saving and loading files to disk? Also, are there other advanced ANML planners you can recommend that are still being developed?

Thank you and congrats on completing your PhD, -david

arbimo commented 2 years ago

Hi David,

it is possible to completely embed FAPE in a JVM application, handing out the problem as strings instead of files. It should be possible to construct the problem directly in Java but it is tricky as it would then be your responsibility to maintain the invariants when constructing the problem.

Also FAPE is currently unmaintained so I would not recommend diving to much into the implementation as I most likely won't have the time to provide support for it. Regarding other ANML planner, Tamer (e.g. notably the paper "Temporal Planning with Intermediate Conditions and Effects") is a planner for the non-hierarchical part of ANML. It is however closed-source.

There is also the Aries planner that supports the same kind of problems as FAPE but has not parser for ANML.

What are the features of ANML that you are interested in and which kind of use case?

Best, Arthur

d-vyd commented 2 years ago

Hi Arthur,

Thanks for writing. I'm interested in computational narrative (stories!) achieved through simulation of "thinking" agents. I've primarily worked with PDDL planners, but most have licenses that restrict commercial use and/or do not handle temporality. Arbimo wrote me that the Aries planner is more focused on hierarchical planning. I will primarily move in the direction of event calculus (or similar) realized through Prolog, answer set programming, etc. These approaches seem very flexible, but are likely much slower.

Are you still researching approaches to AI planning or have you moved on to other topics?

-david

From: Arthur Bit-Monnot @.> Sent: Wednesday, April 13, 2022 7:54 AM To: laas/fape @.> Cc: d-vyd @.>; Author @.> Subject: Re: [laas/fape] FAPE in the ANML landscape? (Issue #8)

Hi David,

it is possible to completely embed FAPE in a JVM application, handing out the problem as strings instead of files. It should be possible to construct the problem directly in Java but it is tricky as it would then be your responsibility to maintain the invariants when constructing the problem.

Also FAPE is currently unmaintained so I would not recommend diving to much into the implementation as I most likely won't have the time to provide support for it. Regarding other ANML planner, Tamer (e.g. notably the paper "Temporal Planning with Intermediate Conditions and Effects") is a planner for the non-hierarchical part of ANML. It is however closed-source.

There is also the Arieshttps://github.com/plaans/aries planner that supports the same kind of problems as FAPE but has not parser for ANML.

What are the features of ANML that you are interested in and which kind of use case?

Best, Arthur

— Reply to this email directly, view it on GitHubhttps://github.com/laas/fape/issues/8#issuecomment-1097674163, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AGJMWF5IROQWY3AOSM75O6LVEZ4RZANCNFSM5RNA25IA. You are receiving this because you authored the thread.Message ID: @.***>