Interactions dynamics, Rhythmicity, action and communication – DIRAC

The acceptability of Human Machine Interfaces (HMI) is a central issue for numerous modern defense systems. The observed difficulty of complex computer systems (monitoring systems, drones, robots …) to sense the human world and interact with it in ways that emulate collaborative human-human work, has founded the DIRAC vision. Our interdisciplinary proposal addresses the question of understanding and mastering the development of “pleasant” yet efficient interactions using a very non-conventional approach. Instead of elaborating on existing complicated solutions, we take a radical simplification route taking advantages of recent discoveries in low-level human interactions and dynamical motor control. We argue that exploiting natural stability and adaptability properties of unintentional synchronizations and rhythmic activities can solve several of the acceptability problems of HMIs, and allow rethinking the current approaches to design them. In early communication among humans, synchrony was found to be a fundamental mechanism relying on very low-level sensory-motor networks, inducing the synchronization of inter-individual neural populations from sensory flows (vision, audition, or touch). Yet, to become a partner in a “working together” scenario, the machine needs also a minimal level of autonomy and adaptation. Predicting the rhythmic structure of the interaction will be used to build reinforcement signals to adapt the machine behavior: synchrony is caused by the interaction but also sustains the interaction itself in a circular way, as promoted by the enaction approach. For more long-term interaction, the challenge is to maintain the interest of the user in the interaction. In DIRAC, our proposition is to add proactive rhythmic interaction capabilities on top of our low-level reflex and passive synchronization system. If the interface can autonomously switch from the leader mode to the follower mode according to the current goal of the interaction (being imitated vs. imitating the human), the self-esteem, engagement (or pleasure), and efficacy of the human when interacting with the machine should be increased, by virtue of having the control over the machine (i.e., the machine recognizes me as an agent). The efficiency of our approach will be evaluated on a robot sentinel application — the guard's tour learning — involving learning and recognition of a visual place, learning the place/action association and building a cognitive map). In addition, simple manipulation tasks (object taking and deposit) will also be tested. The human-like shape of one of our robots (TINO) will allow a smooth translation from human-to-human low-level interactions to HMI. Also, the use of simpler mobile platform built for home application, which uses a tablet PC for managing the interaction, will allow the comparison of interactions mediated by a tactile screen with more human-like visual interactions. In our complementary psychological studies, our task will use a cover story allowing the impact study of our synchronization model for unintentional synchronization during the teaching of the task. Autonomous robot learning will be performed through imitation games (the robot imitates the human). Conversely, imitation will be used as a communication tool contributing to improve teaching (the human imitates the robot to provide it with a feedback on its actions efficiency, in a self evaluation mode). Hence, we plan to evaluate in DIRAC how the interaction dynamics can modify the robot ability to learn and work efficiently with a human. Our evaluation will be both qualitative — answers to a questionnaire related to the pleasure or effortless of the interaction — and quantitative — measure of the precision of the guard ‘s tour reproduction. Finally, our neuromimetic control architecture will constitute a synthetic model of the user to analyze a selection of other HMI.