Search

Collaborative robots (Cobots) are robotic systems designed to physically interact with humans during task execution in a shared industrial workspace. In this paper, we consider a scenario in which a human operator can be supported by a lightweight robotic arm in order to accomplish complex manipulation tasks. Specifically, we assume that manipulation tasks are explicitly represented as hierarchical task networks to be interactively executed exploiting the human physical guidance. In this setting, the human interventions are continuously interpreted by the robotic system in order to infer whether the human guidance is aligned or not with respect to the planned activities. The interpreted human interventions are then exploited by the robotic system to adapt its cooperative behavior during the execution of the shared plan. Depending on the estimated operator intentions, the robotic system can adjust tasks or motions, while regulating the robot compliance in order to follow or lead the human co-worker. The proposed approach is demonstrated in a testing scenario consisting of a human operator that interacts with a Kuka LBR iiwa arm in order to perform a cooperative manipulation task. The collected experimental results show the feasibility and the effectiveness of the approach.

A passivity-based control of Vertical Take-off and Landing (VToL) Unmanned Aerial Vehicles (UAVs) is presented in this paper. An estimator of unmodeled dynamics and external wrench (forces plus moments) acting on the aerial vehicle and based on the momentum of the system is employed to compensate such disturbances effects. This arrangement allows VToL UAVs to perform hovering, tracking and aerial manipulation tasks in unstructured environments. A rigorous stability proof is provided under certain assumptions. Experiments are presented to evaluate the performance of the proposed control design. Passivity-based framework is employed within a classical control scheme for UAVs.Control of angular dynamics does not rely upon exact model cancellation.Estimation is directly fed back to the passivity-based controller.Theoretical proofs and experiments are presented to validate the approach.