Skip to Main content Skip to Navigation

Cooperative POMDPs for human-Robot joint activities

Fabio Valerio Ferrari 1
1 Equipe MAD - Laboratoire GREYC - UMR6072
GREYC - Groupe de Recherche en Informatique, Image, Automatique et Instrumentation de Caen
Abstract : This thesis presents a novel method for ensuring cooperation between humans and robots in public spaces, under the constraint of human behavior uncertainty. The thesis introduces a hierarchical and flexible framework based on POMDPs. The framework partitions the overall joint activity into independent planning modules, each dealing with a specific aspect of the joint activity: either ensuring the human-robot cooperation, or proceeding with the task to achieve. The cooperation part can be solved independently from the task and executed as a finite state machine in order to contain online planning effort. In order to do so, we introduce a belief shift function and describe how to use it to transform a POMDP policy into an executable finite state machine.The developed framework has been implemented in a real application scenario as part of the COACHES project. The thesis describes the Escort mission used as testbed application and the details of implementation on the real robots. This scenario has as well been used to carry several experiments and to evaluate our contributions.
Document type :
Complete list of metadatas

Cited literature [98 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Monday, March 12, 2018 - 12:38:11 PM
Last modification on : Tuesday, February 11, 2020 - 1:38:21 AM
Long-term archiving on: : Wednesday, June 13, 2018 - 1:33:41 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01729083, version 1


Fabio Valerio Ferrari. Cooperative POMDPs for human-Robot joint activities. Human-Computer Interaction [cs.HC]. Normandie Université, 2017. English. ⟨NNT : 2017NORMC257⟩. ⟨tel-01729083⟩



Record views


Files downloads