Research on the physical human-robot interaction is very important for the creation of welfare and medical care robotic systems. To achieve this goal, the integration of mechanical systems, sensory and actuating devices, and control algorithms, should be targeted at safety-dependability issues. To guarantee the intrinsic safety of the physical interaction, robotic systems need to be equipped with compliant actuators. The compliant behavior of the robot leads to a more natural physical interaction and reduces the risk of damages in the case of unwanted collisions. Thus, design of variable stiffness actuation schemes and interaction control strategies is one of the key research problems for ensuring safe and stable interaction.

 

 

 

 

 

 

 

 

 

 

 

Fig.2. An elliptic manipulandum (left), a haptic interface for testing cooperative movements (middle), and a scenario for cooperative human-robot interaction (right).

 

We have practical experiences of using haptic devices and designing specialized manipulanda for measuring and evaluating human movements. In the past, we have also studied control strategies for two-arm robots in cooperative motions. It is of a great research interest to undertake a study of sharing control principles when one of the robot arms is replaced by a human. This is, perhaps, one of the most complex research problems as the presence of a human in the control loop is extremely hard to model. In this aspect, the model-predictive control approach, combined with the on-line detection of motion features and machine learning of motion patterns, is worth looking at. For simplicity and safety, sharing control techniques and experimental scenarios can be tested by using virtual reality technologies and Phantom haptic interfaces (see Fig. 2).