#pddl — Public Fediverse posts

A well-rounded paper on how to translate symbolic statements into actionable constraints for #robotics control and #motionplanning: https://www.frontiersin.org/articles/10.3389/frobt.2023.917637/full

In short: symbolic statements are produced by higher-level decision-making systems (e.g. from a #pddl planner working with #semanticmapping) and given to lower-level actions.

Last week I attended #sido and the #OpenSourceExperience in #paris, and the meeting of the French research group on #robotics. I hope to share what I learned little by little.

It was also an opportunity to explain what I was offering: intrinsically #explainableai using planning (#pddl #ai) for #hri, including dialogues. I shared this video to explain the why, the how, and a demo of it in the wild on #pepperrobot : https://www.youtube.com/watch?v=1ktgn33Zbe4

Please get in touch if that inspires you.

#PDDL #ai applied for #motionplanning of several #robots is shown here:
https://www.researchgate.net/profile/Antony-Thomas-3/publication/363740023_A_Framework_for_Multi-Robot_Task_and_Motion_Planning/links/632c35ee70cc936cd329a826/A-Framework-for-Multi-Robot-Task-and-Motion-Planning.pdf
It can be used for coarser-grained motion planning in semantic maps. In practice a finer motion planner would be needed.

That's the way I'd go, because PDDL also supports logical reasoning.

In this paper, an affordance model (combining objects and actions) called FOON is translated into a #pddl domain to perform #taskplanning on robots: https://doi.org/10.48550/arXiv.2207.05800
#ai #robotics
This is how I wish we did robotics in the industry.
It is already commonplace in the research.