Systematic analysis of the mechanisms of physical human-human interaction (pHHI) requires adequate computational modelling framework. Recently differential game theory models—a multi-agent counterpart of optimal control—have been used to analyse sensorimotor collaborative strategies. A task can be defined by a pair of quadratic cost functionals (one per partner). The ‘plant’ is constituted by the two partners’ body dynamics plus their mechanical links (if any). Every partner has his/her own sensory system. In analogy with single-agent dynamics, we assume that each partner maintains an internal model or state observer of own and partner’s dynamics. The framework naturally incorporates the effects of noisy or incomplete sensory information about own body state. Different interaction strategies can be simulated, ranging from ‘optimal’ collaboration (Nash equilibrium) to no collaboration (two separate LQG controllers). We compared the model predictions with an experimental scenario in which two partners have partly conflicting goals—a reaching task with different via-points. This framework also reproduces behaviours—like ‘slacking’—that are typical of the robot-human interaction in robot-assisted adaptation or rehabilitation trials.

Modelling collaborative strategies in physical human-human interaction

THEKKEDATH CHACKOCHAN, VINIL;Sanguineti, Vittorio
2017

Abstract

Systematic analysis of the mechanisms of physical human-human interaction (pHHI) requires adequate computational modelling framework. Recently differential game theory models—a multi-agent counterpart of optimal control—have been used to analyse sensorimotor collaborative strategies. A task can be defined by a pair of quadratic cost functionals (one per partner). The ‘plant’ is constituted by the two partners’ body dynamics plus their mechanical links (if any). Every partner has his/her own sensory system. In analogy with single-agent dynamics, we assume that each partner maintains an internal model or state observer of own and partner’s dynamics. The framework naturally incorporates the effects of noisy or incomplete sensory information about own body state. Different interaction strategies can be simulated, ranging from ‘optimal’ collaboration (Nash equilibrium) to no collaboration (two separate LQG controllers). We compared the model predictions with an experimental scenario in which two partners have partly conflicting goals—a reaching task with different via-points. This framework also reproduces behaviours—like ‘slacking’—that are typical of the robot-human interaction in robot-assisted adaptation or rehabilitation trials.
978-3-319-46668-2
978-3-319-46669-9
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/893698
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? 4
social impact