Stefan Glasauer

Human-robot interaction in handing-over tasks

In many future joint-action scenarios, humans and robots will have to interact physically in order to successfully cooperate. Ideally, seamless human-robot interaction should not require training for the human, but should be intuitively simple. Nonetheless, seamless interaction and cooperation involve some degree of learning and adaptation. Here, we report on a simple case of physical human-robot interaction, a hand-over task. Even such a basic task as manually handing over an object from one agent to another requires that both partners agree upon certain basic prerequisites and boundary conditions. While some of them are negotiated explicitly, e.g. by verbal communication, others are determined indirectly and adaptively in the course of the cooperation. In the present study, we compared human-human hand-over interaction with the same task done by a robot and a human. To evaluate the importance of biological motion, the robot human interaction was tested with two different velocity profiles: a conventional trapezoidal velocity profile in joint coordinates and a minimum-jerk profile of the end-effector. Our results show a significantly shorter reaction time for minimum jerk profiles, which decreased over the first three hand-overs. The results of our comparison provide the background for implementing effective joint-action strategies in humanoid robot systems.

Biological movement increases acceptance of humanoid robots as human partners in motor interaction

The automatic tendency to anthropomorphize our interaction partners and make use of experience acquired in earlier interaction scenarios leads to the suggestion that social interaction with humanoid robots is more pleasant and intuitive than that with industrial robots. An objective method applied to evaluate the quality of human–robot interaction is based on the phenomenon of motor interference (MI). It claims that a face-to-face observation of a different (incongruent) movement of another individual leads to a higher variance in one’s own movement trajectory. In social interaction, MI is a consequence of the tendency to imitate the movement of other individuals and goes along with mutual rapport, sense of togetherness, and sympathy. Although MI occurs while observing a human agent, it disappears in case of an industrial robot moving with piecewise constant velocity. Using a robot with human-like appearance, a recent study revealed that its movements led to MI, only if they were based on human prerecording (biological velocity), but not on constant (artificial) velocity profile. However, it remained unclear, which aspects of the human prerecorded movement triggered MI: biological velocity profile or variability in movement trajectory. To investigate this issue, we applied a quasi-biological minimum-jerk velocity profile (excluding variability in the movement trajectory as an influencing factor of MI) to motion of a humanoid robot, which was observed by subjects performing congruent or incongruent arm movements. The increase in variability in subjects’ movements occurred both for the observation of a human agent and for the robot performing incongruent movements, suggesting that an artificial human-like movement velocity profile is sufficient to facilitate the perception of humanoid robots as interaction partners.

Handing over a cube: spatial features of physical joint-action.

Even though joint action is highly developed in humans, not much is known about motor control in physical joint‐action tasks. Here we investigated a physical handover task: one subject sequentially passed wooden cubes to another without communicating verbally. Temporal parameters such as reaction time decreased on a trial‐to‐trial basis, showing that the efficiency of the task is optimized on‐line by implicit negotiation between the partners. In contrast, the spatial position of the handover was found to be invariant and trial‐independent. Thus, our results suggest that physical joint‐action is guided by on‐line adaptation and a priori assumptions.

Evaluation of a novel biologically inspired trajectory generator in human-robot interaction

In many future joint-action scenarios, humans and robots will have to interact physically in order to cooperate successfully. Ideally, human-robot interaction should not require training on the human side, but should be intuitive and simple. Previously, we reported on a simple case of physical human-robot interaction, a hand-over task [1]. Even such a basic task as manually handing over an object from one agent to another requires that both partners agree upon certain basic prerequisites and boundary conditions. While some of them are negotiated explicitly, e.g. by verbal communication, others are determined indirectly and adaptively in the course of the cooperation. In the previous study we compared a human-human hand-over interaction with the same task performed by a human and a robot. However, the trajectories used for the robot, a conventional trapezoidal velocity profile in joint coordinates and a minimum-jerk profile of the end-effector, have little resemblance to the natural movements of humans. In this study we introduce a novel trajectory generator that is a variation of the traditional minimum-jerk profile, the ‘decoupled minimum-jerk’ profile. Its trajectory is much closer to those observed in human-human experiments. We evaluated its performance concerning human comfort and acceptance in a simple hand-over experiment by using a post-test questionnaire. The evaluation of the questionnaire revealed no difference with respect to comfort, human-likeness, or subjective safety of the new planner compared to the minimum-jerk profile. Thus, the ‘decoupled minimum-jerk’ planner, which offers important advantages with respect to target approach, proved to be a promising alternative to the previously used minimum-jerk profile.

Spatiotemporal movement planning and rapid adaptation for manual interaction.

Many everyday tasks require the ability of two or more individuals to coordinate their actions with others to increase efficiency. Such an increase in efficiency can often be observed even after only very few trials. Previous work suggests that such behavioral adaptation can be explained within a probabilistic framework that integrates sensory input and prior experience. Even though higher cognitive abilities such as intention recognition have been described as probabilistic estimation depending on an internal model of the other agent, it is not clear whether much simpler daily interaction is consistent with a probabilistic framework. Here, we investigate whether the mechanisms underlying efficient coordination during manual interactions can be understood as probabilistic optimization. For this purpose we studied in several experiments a simple manual handover task concentrating on the action of the receiver. We found that the duration until the receiver reacts to the handover decreases over trials, but strongly depends on the position of the handover. We then replaced the human deliverer by different types of robots to further investigate the influence of the delivering movement on the reaction of the receiver. Durations were found to depend on movement kinematics and the robot’s joint configuration. Modeling the task was based on the assumption that the receiver’s decision to act is based on the accumulated evidence for a specific handover position. The evidence for this handover position is collected from observing the hand movement of the deliverer over time and, if appropriate, by integrating this sensory likelihood with prior expectation that is updated over trials. The close match of model simulations and experimental results shows that the efficiency of handover coordination can be explained by an adaptive probabilistic fusion of a-priori expectation and online estimation.