Home » News » COMPUTATIONAL MODELLING OF PHYSICAL PATIENT-ROBOT COOPERATION: HDR DEFENSE BY LUDOVIC SAINT-BAUZEL

COMPUTATIONAL MODELLING OF PHYSICAL PATIENT-ROBOT COOPERATION: HDR DEFENSE BY LUDOVIC SAINT-BAUZEL

Category: Defense

Ludovic Saint-Bauzel will defend his habilitation to direct research (HDR) at Sorbonne University on Wednesday 08 December at 2pm, in the Durand amphitheatre, Esclanglon building.

Title of the work: “Computational modelling of the physical patient-robot cooperation: From the prediction of pathological movement for the control of assistance robots to the study of human-robot interaction”.

The composition of the jury is as follows:

Summary :

This work presents how computational modelling can be used at different levels of human-robot interaction. In particular, we see how anticipation can be used to improve this interaction at different time horizons, but also how modelling can be done at different cognitive levels of the interaction.

Short-term anticipatory modelling allows the control loop to be improved by decoupling the non-linearities giving information about the state of the interaction to adapt the control parameters. This short-term modelling also allows, as a predictor, to judge or supervise the quality of execution of the modelled movement whether it is healthy or pathological.

When the anticipation is in the medium term, the modelling makes it possible to observe the subject’s intention. Among other things, this describes the interaction as a sequence of robot objectives, with the model guessing the sequence of intentions. From the human’s point of view, this collaboration with the machine seems natural and intuitive, as it takes into account changes of mind and errors.

Finally, long-term anticipation allows for controls on the interaction. It allows for dynamic role changes (leader/follower) between the user and the robot. These models were used in kinesthetic negotiation scenarios of direction (going left or right).

The interaction modelling also led to studies on the evocative power of the programmed action in the robot. This work presents encouraging early results on trust and agentivity during interaction. Trust, for example, has been shown to be transmissible through the kinesthetic channel.

This work opens up perspectives on how such models can be used to improve the intelligibility of robots during their interactions with humans.


Contact : Ludovic Saint-Bauzel, Lecturer