Planning Concurrent Actions and Decisions in Human-Robot Joint Action Context

Anthony Favier (1) , Rachid Alami (1)

1 LAAS-RIS - Équipe Robotique et InteractionS

We address the task planning problem for robots collaborating with humans to achieve a shared task. In this context, humans often want to be free to choose what they want to do, as well as how according to their internal preferences. In this work, we propose a new human-aware task planner that explicitly takes into account a model of execution inspired by joint action schemes while capturing the uncontrollability factor of the human operators and social signals. Its output is the robot's behavioral policy, which describes the robot's action to perform in every state such that the action is congruent and compliant with the human's decision and action, including being passive, to be executed in parallel. The generation of the robot policy using our proposed framework is described, highlighting the planning process and the joint action model it is based on. We provide empirical evidence and brief user study results through simulated BlocksWorld scenarios, validating its ability to produce compliant concurrent robot behaviors.

Mots clés: Human-centered computing → Collaborative interaction • Computing methodologies → Robotic planning collaboration and coordination Human-Robot Collaboration; Planning for Joint Action; Robot Planning; Dynamic Execution; Human-centered computing → Collaborative interaction; • Computing methodologies → Robotic planning; collaboration and coordination Human-Robot Collaboration

Domaines: Robotique [cs.RO]; Interface homme-machine [cs.HC]; Système multi-agents [cs.MA]; Recherche opérationnelle [math.OC]

Anthony Favier, Rachid Alami. Planning Concurrent Actions and Decisions in Human-Robot Joint Action Context. HRI'24 Workshop Symbiotic Society with Avatars (SSA), Mar 2024, Boulder, United States. ⟨hal-04511507⟩