Human Robot co-existence

To ensure co-existence and space sharing between human and robot, collision avoidance is one of main strategies for interaction between them without contact. Since, the human is not controlled, the robot is controlled to achieve this objective while ensuring as much as possible its curent task. To do so, 3D depth camera sensor (Microsoft Kinect V2) is used to estimate online the distance between human (obstacle) and robot's end-effector. This distance is used to apply a repulsive virtual forces in task space. This approach explored well in a previous works (Flacco et al. 2012), is improved here to consider cases where the human hide the robot with respect to the camera view. Hence, the robot's model and its proprioceptive data are integrated in with 3D image perception to estimate the online human-robot distance even when the robot is hidden.

For human safety, a principle of "Safety contour" was introduced to keep distance between human and robot's end effector. This strategy was compared with a previously applied method based on an "infinite depth" approach, and its purpose is to be less conservative in terms of safety.

Human-Robot co-manipulation of objects

To achieve a complex task, it is very important to ensure a physical human-robot interaction (pHRI) and collaboration (pHRC). One of the main key task that highlights pHRC is the co-manipulation of object simultaneously by the human and robot. In our research, we invesitigate the case where the human leads the collaboration and is unpredictable and its actions and intention are unknown by the robot.

• Objects and tools co-manipulation (dynamical objects)

• impedance and admittance control

• Objects with parameters uncertainties

• Collaboration quality criteria

• Intercation stability and passivity-based control

Exoskeleton with muscle-based activation

• Design of active exoskeleton with one degree of freedom

• Control of exoskeleton through activation of arm's antagonist muscles (biceps and triceps)

• The muscle activities are estimated using ElectroMyoGraphy (EMG)